With the increasing use of screening mammography and advances in neoadjuvant therapy, tumours at the time of surgery are more often non-palpable.1 2 3 4 5 6 7 Accurate image-guided localisation is the key to successful excision of these lesions.

Hookwire localisation has been the traditional standard method of localising non-palpable breast lesions for decades. It has many inherent limitations and challenges. Wire placement has to be done on the day of surgery to minimise the risk of wire dislodgment, which limits the flexibility of radiology appointments and scheduling of surgery, therefore potentially resulting in delayed surgery.8 Wire displacement and wire transection with retained fragments have also been reported.5 9 10 The track of the wire limits the surgical approach, causing additional healthy breast tissue to be dissected along the course of the wire.5 9 10 These can affect cosmetic outcome.5 9 10

More recently, radioguided occult lesion localisation (ROLL) has gained popularity, as it overcomes many disadvantages of wire localisation and is reported to be equally effective compared to hookwire.11 However, it also needs to be performed on the same day or a day before surgery due to the half-life of the radiotracer.12 Moreover, radiation safety precautions and the need of Nuclear Medicine unit support limit its widespread use.

Recently, non-radioactive non-wire techniques have started to emerge and address many of these issues. A magnetic marker system (Magseed, Endomagnetics, Cambridge, United Kingdom) is one of these techniques and received clearance for longterm breast implantation from United States Food and Drug Administration in February 2018. It was introduced in Hong Kong in 2019. Our study aimed to evaluate the efficacy and safety of magnetic seed marker localisation of non-palpable breast lesions. To the best of our knowledge, there is no prior publication on magnetic seed marker localisation in a Chinese population.

A retrospective review of all Chinese women who underwent magnetic seed marker localisation for non-palpable breast lesions from June 2019 to February 2020 in a single institution was conducted. Patients were selected by breast surgeons and breast radiologists in consensus by reviewing images on the basis of target visibility and target depth. Patients who had a magnetic seed marker placed but surgery performed out of the study period were excluded.

The magnetic seed marker (Magseed, Endomagnetics, Cambridge, United Kingdom) is made of non-radioactive paramagnetic low-nickel stainless steel. The seed is 5 mm × 0.9 mm, which is the smallest non-wire non-radioactive localisation device available. The magnetic seed marker is preloaded in a sterile 7- or 12-cm 18-gauge deployment needle.

The magnetic seed marker is intended to be placed at a depth up to 3 cm from the skin according to the manufacturer’s instructions13 due to limitations of signal transmission from a greater depth. It is localised with a detector probe (Sentimag, Endomagnetics), which generates an alternating magnetic field to transiently magnetise the seed.14 A visual numerical value and audio feedback are produced according to the strength of the magnetic field, thus signalling the distance of the seed from the detector probe.14

Magnetic seed marker placement was percutaneously performed under image guidance by one of four breast radiologists with 3 to 19 years of experience performing image-guided breast localisation, or by a breast radiology trainee who was directly supervised by one of the breast radiologists. During ultrasound-guided placement, the patient lies supine and rolled slightly with a wedge put under the shoulder on the ipsilateral side to spread the breast evenly. The ipsilateral arm is raised over the patient’s head to facilitate a larger sterilisation field. During stereotactically guided placement, the patient lies on either side or sits up to facilitate breast compression by the stereotactic table.

Target-to-seed distance was evaluated in real time for magnetic seed markers placed under sonographic guidance and was measured on post-procedure mammograms in mediolateral and craniocaudal projections for magnetic seed markers placed under stereotactic guidance. If multiple magnetic seed markers were placed in one breast, the minimum distance between the markers was measured. For patients with magnetic seed markers inserted before the day of surgery, target ultrasound and/or mammography were performed on the day of surgery to evaluate for any delayed magnetic seed marker migration, which was defined by any difference between the initial target-to-seed distance after the localisation procedure and the final target-to-seed distance on the day of surgery. If the final target-to-seed distance was >=10 mm, signifying significant migration, alternative localisation was performed on the day of surgery. Lesions with acceptable marker position underwent marker-guided excision as planned with the depth of the marker from the skin evaluated by preoperative ultrasound, followed by intraoperative guidance with the use of the probe. The presence of the markers in the specimens was confirmed with the probe by surgeons and by specimen radiographs with the radial margins evaluated.

Rates of placement success and retrieval success with a 95% confidence interval (CI) were calculated using the Wilson score method.15 Placement success was defined as a final target-to-seed distance <10 mm in any plane on images on the day of surgery, with reference to guidelines from the National Health Service Breast Screening Programme16 and previous studies.14 17 For degrees of magnetic seed marker placement success, the final target-to-seed distances were further subdivided into ≤1 mm, 2 to 5 mm, and 5 to 9 mm. Retrieval success was determined by the presence of the magnetic seed marker(s) in the specimen radiograph.

Electronic patient records were reviewed for patients’ demographics, preoperative pathology (if any), and indications for surgery. Specimen radiographs and pathology reports were reviewed to verify excision of target lesions and to evaluate the resection margins.

The target lesions were divided into two groups according to the indications for surgery. The surgery was considered to be of therapeutic intent if the target lesion had been proven to be malignant from preoperative pathology. Otherwise, the surgery was considered to be of diagnostic intent. Among the surgeries with therapeutic intent, margin clearance, defined as at least 1-mm disease-free margins, was assessed. The re-excision rate due to inadequate margin clearance was analysed. Complications related to magnetic seed marker deployment and surgeries were recorded.

There were 22 Chinese patients with magnetic seed markers placed during the study period; one patient was excluded due to deferred surgery out of the study period (Fig 1a). A total of 21 patients, with mean age 60.0 years (range, 38-73 years) were included (Table 1). Thirteen patients (61.9%) each had one magnetic seed marker placed on the day of surgery, which were performed during the initial learning period of this new technique. Eight patients (38.1%) had nine magnetic seed markers inserted before the day of surgery in out-patient setting, ranging from 6 to 56 days from surgery with a median of 8 days (interquartile range=6.25-13.75) [Fig 1b]. Fifteen out of 22 magnetic seed markers (68.2%) were placed under ultrasound guidance, and seven magnetic seed markers (31.8%) were placed under stereotactic guidance. The most common type of target lesion was a solid mass (15 of 21, 71.4%), all of which had markers placed under ultrasound guidance. The other six lesions had magnetic seed markers placed by stereotactic guidance, including three groups of microcalcifications, one biopsy marker, one architectural distortion, and one focal asymmetry. One group of calcifications required two magnetic seed markers for bracketing due to its extensive distribution.

Two magnetic markers (9.1%) migrated >=10 mm away from their targets. Both had been placed under stereotactic guidance and migrated along the direction of breast compression (Fig 2). One of these magnetic seed markers was aimed for bracketing initially. No delayed migration was detected in all of the nine magnetic seed markers placed before the day of surgery, and there was no further migration of the two with initial migration. Among the 22 magnetic seed markers, 17 (77.3%) and three (13.6%) were ≤1 mm and 2 to 5 mm from their target, respectively (Figs 3 and 4). Therefore, placement success was achieved in 20 out of 22 magnetic seed markers, with a success rate of 90.9% (95% CI=72.2%-97.5%). The mean final target-to-seed distance was 3.1±9.8 mm (Table 2). The final distance between the two bracketing magnetic seed markers was 29 mm. All 22 magnetic seed markers were able to be localised by the probe intraoperatively and removed successfully (100%; 95% CI=85.1%-100%).

Two out of 21 lesions (9.5%) required alternative localisation performed on the day of surgery to guide lesion excision due to significant magnetic seed marker migration of >=10 mm. One of the lesions was a mammographic architectural distortion that could be visualised on ultrasound. The magnetic seed marker had migrated 13 mm laterally on mammogram. Ultrasound-guided skin marking was performed on the day of surgery with the magnetic seed marker detected and removed together with successful removal of the target lesion (Fig 2). Another lesion was a wide distribution of microcalcifications that required two magnetic seed markers for bracketing under stereotactic guidance. One of the magnetic seed markers migrated 45 mm along the direction of breast compression, with no significant associated haematoma. Salvage hookwire localisation was performed on the day of surgery. The target lesion and the non-migrated magnetic seed marker were first removed by hookwire guidance, and the migrated magnetic seed marker was then detected by the probe and removed.

Nineteen lesions (90.5%) had magnetic seed marker–guided excision as planned, with sonographic depth of the magnetic seed markers from skin ranging from 3 to 21 mm with a mean of 10.8±4.8 mm. Among these 19 lesions, 16 (84.2%) were excised with diagnostic intent and three (15.8%) were excised with therapeutic intent.

For the 16 lesions excised with diagnostic intent, preoperative biopsies or fine needle aspiration had been performed in 14 (87.5%) lesions. Core needle biopsy of 12 lesions, resulted in two with non-diagnostic findings, four with benign pathologies and six with high-risk findings; including four papillary lesions, one atypical ductal hyperplasia, and one with scanty atypical ductal cells. Fine needle aspiration was performed in two lesions, detecting one fibroadenoma and one papillary lesion. In final surgical pathology, two of these 16 lesions (12.5%) had a malignant upgrade from the core biopsy findings including one low-grade and one high-grade ductal carcinoma in situ (DCIS).

For the three lesions excised with therapeutic intent, both preoperative biopsy and final surgical pathology showed DCIS. The subtype of these lesions included a high-grade DCIS, a low-grade DCIS, and an intermediate-grade DCIS with atypical lobular hyperplasia. One of them had close (0.5 mm) margins and required re-excision, for a margin clearance rate of 66.7% and a re-excision rate of 33.3%. There were no reported complications related to magnetic seed marker localisation or lesion excision.

Successful localisation of breast lesions by magnetic seed markers was achieved in 19 out of 21 (90.5%) Chinese patients with a high placement success rate (90.9%) in our study. The majority of the magnetic seed markers were accurately placed with a mean final target-to-seed distance of 3.1 mm. All of the successfully placed magnetic seed markers were <5 mm of the target, with 85% of them ≤1 mm. In all, 100% marker retrieval was achieved without any reported complications. Such results were similar to several recent studies which revealed 100% successful magnetic seed marker retrieval14 17 18 19 20 and high placement success (96.7%-100%).14 17 18 Our re-excision rate for therapeutic intent surgery was found to be 33.3%, which was apparently higher than that reported in previous studies, ranging from 14.8% to 21.9%.17 18 19 This could be attributable to our small sample size with only three lesions excised for therapeutic intent. In fact, a prospective non-randomised control study by Zacharioudakis et al19 with 100 patients in each arm demonstrated that the outcome of magnetic seed marker localisation was comparable to hookwire localisation for breast conservation surgery in terms of re-excision rate. A systemic review by Fusco et al21 demonstrated that the successful localisation and margin clearance rates were 65% to 100% and 58% to 84%, respectively for hookwire localisation, and 93% to 100% and 60% to 100%, respectively for ROLL, while the margin clearance rates from other previous studies9 10 22 ranged from 57% to 87.4% for hookwire localisation and 75% to 93.5% for ROLL. All these suggest that magnetic seed marker is a feasible alternative localisation method.

Thirteen magnetic seed markers (59.1%) were placed on the day of surgery during our initial experience. The purpose was to ensure safety and to allow radiologists’ and surgeons’ familiarisation with the new device and workflow. Among all of the nine magnetic seed markers placed before the day of surgery (range, 6-56 d), none of them showed delayed migration on the day of surgery. This illustrates that delayed migration is unlikely to occur and it may not be necessary for patients to come back to the Radiology Department on the day of surgery to confirm magnetic seed marker position prior to the operation. Similar results were reported by a multi-centre open-label cohort study on mastectomy patients, which showed no migration of magnetic seed markers between placement and surgery, which were up to 30 days apart.20 This reassures the feasibility of decoupling of surgery and radiology appointments, which can potentially reduce localisation-related delay in surgery. Prolonged fasting before surgery and the associated increased risk of vasovagal syncope can therefore be avoided.9 Magnetic seed markers are approved to be placed up to 30 days prior to surgery in Hong Kong at the time this article was written. However, successful retrieval was achieved in one of our patients who had had her surgery deferred to 56 days after magnetic seed marker placement due to personal reasons. This suggests that magnetic seed markers may be applicable for long-term implantation, which has already been approved in the United States and Europe.

Due to limitations of signal transmission, magnetic seed markers are intended to be placed at a depth up to 3 cm from skin according to the manufacturer’s instructions.13 It is challenging to estimate the true lesion depth as the intraoperative breast position varies from the position during breast examinations, particularly when the breast is under compression during mammographic examinations. The distance from skin on the image does not necessarily reflect the shortest distance to the lesion and can be overestimated. Therefore, for lesions visible only on mammography, we selected those near the skin or at middle depth on mammography. For ultrasound-detected lesions, the sonographic depth of the lesion from the skin would be measured. We performed sonographic measurements for magnetic seed marker depth for all patients as the sonographic breast position should best resemble its intraoperative position. In our study, the depth of magnetic seed marker placement in successfully localised lesions ranged from 3 to 21 mm with a mean of 10.8 mm. All magnetic seed markers were able to be localised by the probe intraoperatively. The depth limitation of magnetic seed markers is probably not a major issue in the Chinese population, since Chinese females tend to have thinner breasts.23 Further study is warranted to validate this postulation.

Because of potential signal interference, two magnetic seed markers should not be placed at close proximity (<2 cm apart) within the breast.14 20 This can potentially limit its use in bracketing a target or in localising multiple target lesions in one breast. We had one case requiring two magnetic seed markers placed in the same breast for bracketing a group of microcalcifications. Although one of them showed significant migration from the initial target, the final distance between the two magnetic seed markers was 29 mm. Since there could be potential interference to the probe from hookwires, the target lesion and the non-migrated magnetic seed marker were first removed by hookwire guidance, and the migrated magnetic seed marker was then detected by the probe and removed. The utility of multiple magnetic seed markers in one breast should be further evaluated in future studies with larger sample sizes.

In our study, two magnetic seed markers (9.1%) were found to have undergone significant migration of >=10 mm from the target on post-insertion images. Both of them migrated along the direction of breast compression after the compression was released, with no significant hematoma detected radiologically or clinically. We postulate such migration to be resulting from the accordion effect, which is a well-known cause for clip migration after stereotactically guided biopsy. Fatty breasts are known to be more susceptible to accordion effect–related migration as they are more compressible and are usually compressed to a greater degree.24 The migrated biopsy marker can move in the direction of compression either proximal or distal to the needle track when the breast expands to its original size and shape after compression.25 26 27 28 It is best evaluated in the plane orthogonal to the direction of compression used.25 Such migration was also recognised in 5.9% of tomosynthesis-guided magnetic seed marker localisation procedures by a previous study.17 For prevention, it is suggested to hold and release the breast slowly from the compression pad after marker placement.17 Chinese patients probably have a lower risk of accordion effect–related migration, as they tend to have denser breasts.23 However, it could not be analysed in our study given our small sample size with only seven magnetic seed markers placed under stereotactic guidance. Future research with a larger sample size is needed to evaluate the association between breast density and seed migration.

There are several other drawbacks of magnetic seed marker localisation. Cost is a major concern as it is more expensive compared with hookwire or ROLL. Extra costs are needed for the initial purchase of the probes and instruments,17 as specialised non-ferromagnetic surgical instruments must be used to avoid magnetic interaction between magnetic seed marker and sensor. However, minimising localisation-related delay in surgery may reduce the operational cost and improve workflow efficiency. A full cost analysis is necessary in the future. In addition, magnetic seed markers could not be placed under magnetic resonance imaging (MRI) guidance as the deployment needle is made of stainless steel. Magnetic seed marker insertion is contra-indicated for patients who have pacemakers or implanted cardiac devices due to interference of the devices with the probe.29 Magnetic seed markers are not officially indicated for use in nickel allergy patients. Bone wax, which is used as a terminal plug in the deployment needle, contains beeswax, and may cause allergic or foreign body reaction.13 Magnetic seed marker deployment is also not advised in a patient who may undergo future breast MRI prior to surgery due to its void artefact of 4 to 6 cm distance,5 9 which influences the MRI diagnostic accuracy.5

There are several limitations to this study. It is a single-institution retrospective study without direct comparison to our hookwire localisation or ROLL cases. Patients were selected for magnetic seed marker localisation in a multidisciplinary meeting involving breast radiologists and breast surgeons and this might introduce selection bias. We did not have any patients with a preoperative diagnosis of invasive carcinoma in our study, as sentinel node and occult lesion localisation with a radioisotope still remains the preferred localisation method for invasive carcinoma requiring sentinel lymph node biopsy in our centre. This can be performed in one single procedure instead of two, thus minimising patients’ discomfort and potential complications from the procedure. However, magnetic seed markers have been reported to be a safe and feasible method for image-guided excision of invasive carcinoma.14 17 18 As discussed before, our small sample size limited our analyses of migration and margin clearance rates, and the evaluation of the feasibility of using multiple seeds in one breast for bracketing a lesion or for localising multiple lesions. A prospective randomised trial with larger sample size will be necessary to fully compare wire localisation and ROLL to magnetic seed marker localisation. Patient satisfaction, the reproducibility operator dependence of magnetic seed marker deployment and intraoperative localisation, specimen weight, and cosmetic outcome can also be investigated in future studies.

The magnetic seed marker system demonstrated safety and efficacy in Chinese women to localise and excise non-palpable breast lesions and appears to overcome many of the limitations of conventional localisation techniques. It can be an alternative to hookwires or ROLL in selected patients. Future research is needed to validate the results.

Concept or design: WY Fung, T Wong, CM Chau, ELM Yu.
Acquisition of data: WY Fung.
Analysis or interpretation of data: WY Fung, T Wong, CM Chau, ELM Yu, JKF Ma.
Drafting of the manuscript: WY Fung, T Wong, CM Chau, ELM Yu.
Critical revision of the manuscript for important intellectual content: All authors.

All authors have disclosed no conflicts of interest.

We would like to express our gratitude to our breast team (Department of Surgery, Princess Margaret Hospital) for support of our research project.

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

This study was approved by Kowloon West Cluster Research Ethics Committee (Ref 146-11). The need for patient consent was waived by the Research Ethics Committee.

1. Lui CY, Lam HS, Chan LK, et al. Opportunistic breast cancer screening in Hong Kong; a revisit of the Kwong Wah Hospital experience. Hong Kong Med J 2007;13:106- 13.

2. Lau SS, Cheung PS, Wong TT, Ma MK, Kwan WH. Comparison of clinical and pathological characteristics between screen-detected and self-detected breast cancers: a Hong Kong study. Hong Kong Med J 2016;22:202-9. Crossref

3. Welch HG, Prorok PC, O’Malley AJ, Kramer BS. Breast-cancer tumor size, overdiagnosis, and mammography screening effectiveness. N Engl J Med 2016;375:1438-47. Crossref

4. Ramos M, Díez JC, Ramos T, Ruano R, Sancho M, González-Orús JM. Intraoperative ultrasound in conservative surgery for non-palpable breast cancer after neoadjuvant chemotherapy. Int J Surg 2014;12:572-7. Crossref

5. Hayes MK. Update on preoperative breast localization. Radiol Clin North Am 2017;55:591-603. Crossref

6. Chu TY, Lui CY, Hung WK, Kei SK, Choi CL, Lam HS. Localisation of occult breast lesion: a comparative analysis of hookwire and radioguided procedures. Hong Kong Med J 2010;16:367-72.

7. Dauphine C, Reicher JJ, Reicher MA, Gondusky C, Khalkhali I, Kim M. A prospective clinical study to evaluate the safety and performance of wireless localization of nonpalpable breast lesions using radiofrequency identification technology. AJR Am J Roentgenol 2015;204:W720-3. Crossref

8. Sharek D, Zuley ML, Zhang JY, Soran A, Ahrendt GM, Ganott MA. Radioactive seed localization versus wire localization for lumpectomies: a comparison of outcomes. AJR Am J Roentgenol 2015;204:872-7. Crossref

9. Kapoor MM, Patel MM, Scoggins ME. The wire and beyond: recent advances in breast imaging preoperative needle localization. Radiographics 2019;39:1886-906. Crossref

10. Cheang E, Ha R, Thornton CM, Mango VL. Innovations in image-guided preoperative breast lesion localization. Br J Radiol 2018;91:20170740. Crossref

11. Ocal K, Dag A, Turkmenoglu O, Gunay EC, Yucel E, Duce MN. Radioguided occult lesion localization versus wire-guided localization for non-palpable breast lesions: randomized controlled trial. Clinics (Sao Paulo) 2011;66:1003-7. Crossref

12. Manca G, Mazzarri S, Rubello D, et al. Radioguided occult lesion localization: technical procedures and clinical applications. Clin Nucl Med 2017;42:e498-e503. Crossref

13. Endomag Ltd. Magseed® magnetic marker system: Instructions for use. 2017. Available from: https://www. endomag.com/magseed/overview/. Accessed 30 Mar 2020.

14. Price ER, Khoury AL, Esserman LJ, Joe BN, Alvarado MD. Initial clinical experience with an inducible magnetic seed system for preoperative breast lesion localization. AJR Am J Roentgenol 2018;210:913-7. Crossref

15. Wilson EB. Probable inference, the law of succession, and statistical inference. J Am Stat Assoc 1927;22:209-12. Crossref

16. Sibbering M, Watkins R, Winstanley J, Patnick J, editors. Quality Assurance Guideline for Surgeons in Breast Cancer Screening (NHSBSP Publication No. 20). 4th ed. Sheffield: NHS Cancer Screening Programmes; 2009.

17. Lamb LR, Bahl M, Specht MC, D’Alessandro HA, Lehman CD. Evaluation of a nonradioactive magnetic marker wireless localization program. AJR Am J Roentgenol 2018;211:940-5. Crossref

18. Thekkinkattil D, Kaushik M, Hoosein MM, et al. A prospective, single-arm, multicentre clinical evaluation of a new localisation technique using non-radioactive Magseeds for surgery of clinically occult breast lesions. Clin Radiol 2019;74:974.e7-11. Crossref

19. Zacharioudakis K, Down S, Bholah Z, et al. Is the future magnetic? Magseed localisation for non palpable breast cancer. A multi-centre non randomised control study. Eur J Surg Oncol 2019;45:2016-21. Crossref

20. Harvey JR, Lim Y, Murphy J, et al. Safety and feasibility of breast lesion localization using magnetic seeds (Magseed): a multi-centre, open-label cohort study. Breast Cancer Res Treat 2018;169:531-6. Crossref

21. Fusco R, Petrillo A, Catalano O, et al. Procedures for location of non-palpable breast lesions: a systematic review for the radiologist. Breast Cancer 2014;21:522-31. Crossref

22. Nadeem R, Chagla LS, Harris O, et al. Occult breast lesions: a comparison between radioguided occult lesion localisation (ROLL) vs. wire-guided lumpectomy (WGL). Breast 2005;14:283-9. Crossref

23. Maskarinec G, Meng L, Ursin G. Ethnic differences in mammographic densities. Int J Epidemiol 2001;30:959-65. Crossref

24. Rosen LE, Vo TT. Metallic clip deployment during stereotactic breast biopsy: Retrospective analysis. Radiology 2001;218:510-6. Crossref

25. Burbank F, Forcier N. Tissue marking clip for stereotactic breast biopsy: initial placement accuracy, long-term stability, and usefulness as a guide for wire localization. Radiology 1997;205:407-15. Crossref

26. Liberman L, Dershaw DD, Morris EA, Abramson AF, Thornton CM, Rosen PP. Clip placement after stereotactic vacuum-assisted breast biopsy. Radiology 1997;205:417-22. Crossref

27. Esserman LE, Cura MA, DaCosta D. Recognizing pitfalls in early and late migration of clip markers after imaging-guided directional vacuum-assisted biopsy. Radiographics 2004;24:147-56. Crossref

28. Endomag Ltd. Sentimag®: instructions for use. 2018. Available from: https://www.endomag.com/sentimag/. Accessed 30 Mar 2020.

Health literacy is the ability to obtain, read, understand, and use healthcare information to make appropriate health decisions and follow treatment instructions. Inadequate health literacy (IHL) is a public health concern that is associated with poor health outcomes and frequent use of health services.1 Identifying groups at risk for IHL is therefore crucial. Screening tools such as the Rapid Evaluation of Adult Literacy in Medicine–Revised (REALM-R),2 Rapid Evaluation of Adult Literacy in Medicine–Short form,3 Wide Range Achievement Test (WRAT-4),4 Newest Vital Sign (NVS),5 and Single-Item Literacy Screener6 have been developed for assessment of patients’ health literacy. Nonetheless, each of those tools has limitations, hindering the extensive use of rapid health literacy screening in clinical settings.

The REALM-R, Rapid Evaluation of Adult Literacy in Medicine–Short form, and WRAT-4 focus only on word recognition,7 representing a narrow concept of health literacy and failing to address two other crucial dimensions: ‘interpretation of health information’ and ‘health decision making’.7 8 The REALM-R and WRAT-4 health literacy assessments require 2 to 3 and 3 to 5 minutes, respectively, to administer. Thus, using the REALM-R or WRAT-4 demands special arrangements in clinical settings: patients may be required to complete them prior to consultation, and they may interrupt the usual clinical workflow. An alternative is for the doctor to conduct the assessment, but the typical out-patient clinical consultation period is about 7 minutes for each patient, and adequate health literacy assessment would occupy multiple minutes of this period.

The NVS is another recommended tool for quick health literacy screening that compensates for the shortcomings of previous tools by addressing the need to understand and interpret health information from a designated nutritional label. After reading the label, the client answers six questions about it. The assessment of these capacities by the NVS is both a strength and a shortcoming, as it requires more time to complete.9 Notably, older adults take 11.7 minutes (range, 6-28 minutes) to complete the NVS, so its practicality for quick assessment of elderly patients’ health literacy is limited.10

Another rapid health literacy assessment tool is the Short-Form Test of Functional Health Literacy in Adults, a 36-item tool that assesses clients’ comprehension and numeracy abilities.11 It is recommended to allot 7 minutes to complete the assessment, and clients should stop when that time is up. However, time-limited assessments can be challenging for older adults because of their delayed cognitive processing or age-related slowness. These effects are typical but not pathological with age, rendering the Short-Form Test of Functional Health Literacy in Adults inappropriate for this population.12 Further, most of its contents were based on the US healthcare system, making its generalisability to other countries questionable.

The Single-Item Health Literacy Screening is the simplest health literacy assessment, containing only one item. Its key limitation is possible self-report bias, as it assesses clients’ perceived ability to read and understand health information from written material, which may not reflect their actual abilities.6

Given the shortcomings of existing rapid health literacy screening tools and the need to assess patients’ health literacy in clinical settings, there is a need to develop a rapid tool for non-English-speaking older adults that can be used in different healthcare systems and is based on available patient data. The project team has developed several health literacy tools for Chinese populations, including the Chinese Health Literacy Scale for Diabetes,13 Chinese Health Literacy Scale for Chronic Care (CHLCC),14 and Chinese Health Literacy Scale for Diabetes–Multiple Choice version.15 Although these tools can be used in Chinese-speaking populations, they require several minutes for clients to complete, which may not be ideal for rapid screening in busy clinics. Therefore, in this study, we aimed to develop and validate a brief practitioner-friendly health literacy screening tool, the Rapid Estimate of Inadequate Health Literacy (REIHL), which employs a multivariable prediction model to determine patients’ risk for IHL in a demanding clinical setting.

This is a cross-sectional, methodological study that was conducted from August 2010 to January 2011. The Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis guidelines were also followed.16

Older adults from one community health centre and five district elderly community centres in Hong Kong were recruited. The inclusion criteria were: (1) age ≥50 years; (2) cognitively capable (Short Portable Mental Status Questionnaire Chinese version score ≥7); and (3) able to communicate in Cantonese. The sample size calculation was derived from a receiver operating characteristic (ROC) power calculation using the ‘power.roc.test’ function under the ‘pRCO’ library in R version 3.6. Assuming that the newly developed tool’s area under the curve 0.60, type 1 error 0.05, power 0.8, and attrition 20%, at least 298 subjects should be recruited.17

Recruitment strategies included posters at community centres, monthly meetings, and in-person contact. All participants were interviewed to assess their eligibility to participate. Ethical approval was obtained from the Institutional Review Board of The University of Hong Kong/Hospital Authority Hong Kong West Cluster (Ref UW 09-033).

The newly developed REIHL screening tool was devised using model estimation. Unlike other scale development, we did not create the items for the REIHL but collected socio-demographic data (age, gender, education level, types of chronic illness) and conducted the CHLCC on the subjects. Scores on the CHLCC were used to determine which subjects had IHL. People with CHLCC scores of <36 were considered as having IHL. We then created a dummy variable representing IHL (1: IHL; 0: adequate health literacy). Socio-demographic factors (eg, age, education level, types and number of chronic illnesses) associated with IHL were identified, and these became the items of the REIHL. Chronic illnesses refer to conditions that last 1 year or more and lead to limitations in activities of daily living and/or require ongoing medical attention.18

People with IHL were more likely to have more depressive symptoms19 20 and poor self-rated health (SRH).21 22 We therefore checked the criterion validity and concurrent validity of REIHL with the following validated scales.

The CHLCC was used to check the criterion validity of REIHL. The CHLCC is a 24-item tool for measuring health literacy in Chinese populations with four subscales (remembering, understanding, applying, and analysing). It has good internal (Cronbach’s alpha, 0.91) and test-retest (intraclass correlation coefficient, 0.77; P<0.01) reliability.14

The Geriatric Depression Scale–15 (GDS-15) and the SRH scale were used to check the REIHL’s concurrent validity. The GDS-15 is used to assess older adults’ depressive symptoms,23 and it has been translated into Chinese and validated in Hong Kong with good psychometric properties (Cronbach’s alpha, 0.82; item-total correlation, 0.23-0.66).24 25 Its total score ranges from 0 to 15, with higher values representing increased depression levels. The SRH is a validated single-item scale for assessing general health status.26 It is a subjective assessment of general health, asking ‘In the last 3 months, how would you describe your health status?’ Five options are given: ‘very good’, ‘good’, ‘fair’, ‘poor’, or ‘very poor’, coded as integers from 1 (very good) to 5 (very poor).

There were a few items of missing data, which comprised about 2% of all data. Missing values were filled in using multiple imputation in SPSS (Windows version 25.0; IBM Corp, Armonk [NY], US). Chi squared tests were used to assess the bivariate relationships between demographic variables and IHL. Logistic regression analyses were used to further assess the multivariate relationships among the factors that were significantly associated with IHL. Model adequacy was evaluated by Nagelkerke’s R².27 To obtain an optimistic assessment of the model’s prediction performance and avoid overfitting, 10-fold cross-validation was used, and error mean square (EMS) was reported. To select the best model, we chose the model with the smallest EMS and lowest Bayesian information criterion (BIC) values. We derived the point scores for REIHL with reference to the Framingham Study Risk Score.16 The score for each item of the REIHL is calculated by dividing its coefficient by the smallest coefficient and then rounding up to the next highest integer. The total REIHL score is the sum of the scores of all items in the REIHL.

To test the reliability of the REIHL, we used ROC curve analysis28 to assess its sensitivity and specificity. We choose the optimum sensitivity and specificity based on maximisation of Youden’s index.29 We assessed the corresponding sensitivity and specificity of each potential cut-off point. The chosen cut-off point was the one with the largest Youden’s index (ie, sensitivity + specificity − 1). We also assessed the criterion validity and concurrent validity of REIHL. Criterion validity refers to the stated criterion, that is, the correspondence between the results of this newly developed scale and those of a validated health literacy scale. Concurrent validity is the extent to which a test relates to another previously validated metric. Here, we tested the REIHL’s criterion validity with the CHLCC and the REIHL’s concurrent validity with two health outcomes (depression and SRH). Path analysis30 was also used to examine the criterion and concurrent validity of REIHL with a validated health literacy scale (CHLCC) and two health outcomes (depressive symptoms and SRH) using MPlus version 7.31 We assessed three fit indices to determine the goodness of fit of the model: a model with non-significant Chi squared value (P>0.05), comparative fit index ≥0.95, and root mean square error of approximation ≤0.10 was considered to be a well-fitting model.32 33 We also inspected the direction and significance of the standardised estimate coefficients to determine the effects of one variable on another.

A total of 304 subjects were included in the analysis, of whom 220 (72.4%) were female. In all, 185 (60.9%) subjects were shown to have IHL when assessed by the health literacy scale (CHLCC score <36). Age, gender, education level, and number of chronic illnesses were significantly associated with CHLCC (Table 1).

Because gender was significantly correlated with education, we selected education level as the representative variable used in the regression models (Table 2). Model 1 employed a regression model that incorporated age and education, and the results were: Nagelkerke’s R² 0.39, EMS 0.15, and BIC 347. To form Model 2, we added five chronic illnesses (ie, diabetes, hypertension, stroke, heart disease, and osteoporosis) into the regression; Nagelkerke’s R² increased to 0.43, EMS to 0.16, and BIC to 359. In Model 3, the selected chronic illnesses were replaced by the number of chronic illnesses, and Nagelkerke’s R² became 0.40, EMS 0.15, and BIC 346. Because Model 3 had the lowest BIC and EMS values, and its Nagelkerke’s R² was comparable to those of the other two models, we considered Model 3 as the best and final model.

The coefficients of age, education level, and number of chronic illnesses were identified in Model 3. The smallest coefficient was 0.34, and that value was used as the denominator to calculate the score for each item. Age was categorised and scored as 0, 4, 4, or 7; education level was scored as 0, 2, 5, or 11; and chronic illnesses were scored from 0 to 5 depending on their number (Table 3). Therefore, the total REIHL score ranged from 0 to 23. The REIHL had 76.9% agreement with the CHLCC, the validated, reliable health literacy scale. The area under the ROC curve for predicting IHL was 0.82 (95% confidence interval=0.78-0.87, P<0.001; Fig 1). The curve for the REIHL showed that scoring ≥11 had a sensitivity of 77.8% and specificity of 75.6% for predicting IHL. This criterion identified 60.9% of the participants as having IHL.

All of the REIHL items had unique scores except for two items under ‘Age’ that had the same score (ie, 4) after rounding up. The actual score for those aged 65 to 74 years was 3.71 (=1.27/0.34), whereas that for those aged 75 to 84 years was 4.44 (=1.55/0.34). Because the difference between the actual scores (3.71 and 4.44) was almost 1, we considered the possibility of adjusting the score of the item ‘aged 75 to 84 years’ to 5. The sensitivity and specificity of the REIHL were 72.4% and 79.8%, respectively, when adjusted accordingly. These results were not significantly different from those before adjustment. The agreement between REIHL and CHLCC in the adjusted model was 75.3%, which was lower than that before adjustment. The area under the ROC curve of the adjusted REIHL was 0.83 (ie, very close to the corresponding value of the unadjusted version). In view of the insignificant improvement in psychometric properties, we propose to not adjust the scoring of the item ‘aged 75 to 84 years’, leaving it as 4.

The path analysis model showed excellent fit (Chi squared [2, 304] 0.16, P=0.92, comparative fit index 1.00, root mean square error of approximation <0.001, 90% confidence interval=0.00-0.04), indicating the criterion validity and concurrent validity of the REIHL (Fig 2). The path between the REIHL and CHLCC (β= −32.69, P<0.001) was statistically significant, implying that the REIHL was significantly negatively associated with the CHLCC. This shows the criterion validity of REIHL with a validated health literacy instrument. A negative association between the two scales is reasonable and expected because the REIHL measures inadequacy, unlike the CHLCC, which measures adequacy. The path between the REIHL and the GDS-15 (β = 0.13, P<0.01) was also statistically significant, but the path between the REIHL and SRH was not. This implies that there was a significant relationship between IHL and depressive symptoms. The path between the GDS-15 and SRH (β=1.02, P<0.001) was statistically significant, indicating a strong relationship between depression and poor SRH.

The newly developed REIHL is a reliable screening tool for estimating IHL among older adults in clinical settings. Although REIHL is an estimation tool, it had very good agreement with a validated health literacy measure (CHLCC). This implies that if clinicians have limited time to assess patients’ health literacy, they could estimate it using the REIHL rather than actually measuring patients’ health literacy levels.

The strengths of the REIHL are its reliability and simplicity. We used several methods to test the tool’s reliability. For instance, ROC analysis found that the area under the curve was more than 80%, and the sensitivity and specificity of the REIHL with a cut-off point of 11 reached an acceptable level, indicative of an accurate assessment tool. The relationships between REIHL and CHLCC were well illustrated in the path analysis, indicating that REIHL has reasonably good criterion validity. Previous literature showed that adults with IHL were more likely to have depressive symptoms19 20 and poor SRH.21 22 The path analysis showed that REIHL was significantly associated with GDS-15 but not with SRH; however, the GDS-15 was significantly associated with SRH, and the model showed good fit. These findings confirmed the concurrent validity of REIHL, as estimated IHL was significantly associated with depression. This result provides some added value, as IHL was indirectly associated with poor SRH via depression. This means that older adults’ poor SRH was caused not directly by IHL, but by the presence of depression.

Three of the REIHL’s items (age, education, and number of chronic illnesses) may be risk factors for depression, so testing the tool’s association with depression might be a challenge analogous to testing the relationship between risk factors and poor health outcomes. However, we are confident that the inclusion of these items in the REIHL is a good design choice to highlight the heterogeneity of older adults and remind practitioners to be sensitive to the differentiation among clients. People of advanced age and low education are more likely to have poor health outcomes (including depression), but the age and education level at which practitioners should be mindful of IHL remains unclear. The REIHL is a reminder to practitioners to pay proper attention to these important aspects so that they can communicate with patients to self-manage their health issues. Advocating the use of REIHL is not intended to replace the concept of universal precaution in health literacy and its adoption, but it highlights a population that needs special attention regarding health literacy. In clinics where most clients are older adults, practitioners could thereby direct their limited time and resources to those in the greatest need. By contrast, when encountering less-educated clients, some practitioners do not attempt to educate them, assuming they are unable to understand or apply the information. In such situations, the REIHL may encourage practitioners to adopt strategies such as referring clients with IHL to training. In one US study, people with IHL were referred to regular telephone counselling provided by health coaches (trained nurses, health educators, and diabetes educators) for 12 months.34 The health coaches delivered health advice/messages in simple sentences over the phone. In a Hong Kong study, multi-component nurse-led group meetings derived from the concept of photovoice were arranged for patients with diabetes, hypertension, and limited health literacy.35 In these meetings, participants used photos to express barriers to and facilitators of physical activity and developed plans to improve their health status.35 These two examples illustrate how people with IHL have been supported to communicate with healthcare professionals to make their health decisions.

The REIHL can be used easily by clinicians provided that they know the clients’ age, education level, and number of chronic illnesses. Its scoring system is simple, with the sum of all items forming the total score. The different levels within each item have unique scores, except for two age categories (65-74 and 75-84 years), to both of which the score 4 was assigned. We investigated the possibility of adding one more point to the latter category’s score, but this did not contribute additional sensitivity or specificity; therefore, we decided to keep the status quo.

The REIHL could contribute to the hands-on 1-minute estimation of patients’ health literacy levels that is sometimes performed in clinical areas. Such a swift assessment allows practitioners to make decisions in health education, such as avoiding the use of jargon, providing simplified information and illustrations, using the ‘teach-back’ method, and encouraging patients’ questions. These strategies can improve health behaviours among those with IHL.36 As IHL is a common phenomenon in clinical settings, the Agency for Healthcare Research and Quality and the Institute for Healthcare Improvement of the US recommend that practitioners use the teach-back method as a strategy of taking universal precautions for health literacy (ie, applying such precautions to all patients).37 In the teach-back method, patients are asked to repeat the instructions they receive from doctors and nurses, allowing healthcare professionals to check patients’ understanding of the health messages and then re-teach or modify the method of presentation if the patients do not demonstrate comprehension. Throughout the process, it is recommended that doctors and nurses have a caring attitude and use plain language in communication.38

Unlike other rapid estimation tools for health literacy, such as the REALM-R, the REIHL does not require clients to read aloud. This enables practitioners to estimate patients’ health literacy without embarrassing them, which is particularly suitable to Chinese culture in view of the concept of ‘saving face’. Its application is highly recommended in the management of geriatric patients, as such patients are a heterogeneous group in terms of health literacy adequacy. Older patients’ literacy problems may not be obvious, as some may conceal their problems out of shame or may not recognise their difficulties with reading. Such individuals may be unable to ask relevant health questions or may misunderstand healthcare providers’ recommendations. As older patients tend to have many co-morbidities, they need to navigate the health care system and interpret complex information, which are challenging for people with IHL. Understanding patients’ health literacy could allow the implementation of strategies that could potentially improve their health and reduce emergency attendance and hospital admissions. Two strategies have been proven effective to facilitate medication adherence and health literacy. A self-management education programme (two 30- to 40-minute weekly meetings followed by four phone-based educational sessions) tailored to health literacy was shown to increase adherence to antihypertensive medication.39 Another strategy is the use of a tailor-made comic book to facilitate medication counselling sessions (two 45-minute face-to-face meetings) administered by trained volunteers.40 41 Because people with IHL are more likely to have low confidence in medicine taking,33 42 health education of this kind can be beneficial to people with chronic illnesses.

The REIHL is a screening tool for health literacy. Because of its estimated nature and capacity for rapid implementation, it is best used in ambulatory care or out-patient care clinics. The REIHL cannot replace the CHLCC, which assesses health literacy levels accurately and directly. However, the REIHL is good at identifying members of the high-risk population on whom the administration of the CHLCC or other health literacy tests is warranted. The prevalence of IHL in this sample was high (61%), and this result is comparable to those found in other populations: in the Netherlands, the prevalence of IHL in patients with arterial vascular disease was 76.7%,43 whereas in Brazil, more than half of people with hypertension (54.6%) had IHL.44 As the prevalence of IHL is high across various populations, there should be no problems with the generalisation of this health literacy tool. However, to determine whether the REIHL can be applied in other populations or nations, a cross-national study should be carried out in future.45

Several limitations of this study should be acknowledged. First, the cross-sectional design did not allow us to investigate the causal relationship between health literacy and health outcomes. Second, because only Chinese subjects were included, the threshold is only valid for Chinese older adults, and whether the results can be generalised to other non-Chinese populations is not known. Future studies should investigate the scale’s psychometric properties in other populations. Third, we recruited volunteers from community district elderly centres, so there is some selection bias based on interest and motivation. Further, the tool measures the risk for IHL based on patients’ background information; thus, it is not sensitive to changes in an individual’s personal health literacy level. Previous studies have shown that cognitive impairment is strongly related to low health literacy. However, we restricted the inclusion criteria to those without impaired cognitive function.46 Fourth, the REIHL relies on self-reported items, so under-reporting or over-reporting are possible. Inaccurate reporting may be the result of stigma or the potential for embarrassment associated with low education levels or literacy abilities. Caution should be applied when interpreting REIHL scores. Finally, the present dataset is too small to be split into training and validation datasets. Future studies with larger datasets should be used to validate this scale.

The REIHL is a practitioner-friendly tool for screening older adults’ risk for IHL, which can be applied in clinical settings to identify at-risk groups. This tool is particularly useful in demanding clinical areas where older adults constitute the majority of patients. Future studies should assess how using the REIHL in a community clinical setting encourages healthcare providers to relate better to patients with lower health literacy and improves communication with them.

Concept or design of study: AYM Leung, PH Chau, I Chi.
Acquisition of data: ISH Leung, KT Cheung.
Analysis or interpretation of data: All authors.
Drafting of the manuscript: ISH Leung, KT Cheung.
Critical revision of the manuscript for important intellectual content: AYM Leung, EYT Yu, JKH Luk, D Levin-Zamir, I Chi.

All authors had full access to the data, contributed to the study, approved the final version for publication, and take responsibility for its accuracy and integrity.

As an editor of the journal, JKH Luk was not involved in the peer review process of the article. Other authors have disclosed no conflicts of interest.

The authors would like to acknowledge the invaluable contribution of the study participants. Special thanks go to the anonymous reviewers for their thoughtful review and guidance.

The findings of this study were presented in part as a poster at the 10th International Symposium on Healthy Aging, Hong Kong. Leung ISH, Leung AYM, Chau PH (2015, March 7-8). Rapid Estimate of Inadequate Health Literacy (REIHL) for community-dwelling Chinese older adults.

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available because of privacy or ethical restrictions.

This project was funded by Seed Funding for Basic Research, HKU 2010-11 (Project No: 200911159075) of the University of Hong Kong.

Approval was obtained from the Institutional Review Board of The University of Hong Kong/Hospital Authority Hong Kong West Cluster (Ref UW 09-033).

1. Berkman ND, Sheridan SL, Donahue KE, Halpem DJ, Crotty K. Low health literacy and health outcomes: an updated systematic review. Ann Int Med 2011;155:97-107. Crossref

2. Davis TC, Long SW, Jackson RH, et al. Rapid estimate of adult literacy in medicine: a shortened screening instrument. Fam Med 1993;25:391-5.

3. Arozullah AM, Yarnold PR, Bennett CL, et al. Development and validation of a short-form, rapid estimate of adult literacy in medicine. Med Care 2007;45:1026-33. Crossref

4. Wilkinson GS, Robertson GJ. Wide Range Achievement Test–Fourth Edition (WRAT-4). Lutz, FL: Psychological Assessment Resources; 2006. Crossref

5. Weiss BD, Mays MZ, Martz W, et al. Quick assessment of literacy in primary care: the newest vital sign. Ann Fam Med 2005;3:514-22. Crossref

6. Morris NS, MacLean CD, Chew LD, Littenberg B. The Single Item Literacy Screener: evaluation of a brief instrument to identify limited reading ability. BMC Fam Pract 2006;7:21. Crossref

7. Haun JN, Valerio MA, McCormack LA, Sørensen K, Passche-Orlow MK. Health literacy measurement: an inventory and descriptive summary of 51 instruments. J Health Comm 2014;19 Suppl 2:302-33. Crossref

8. Carpenter CR, Kaphingst KA, Goodman MS, Lin MJ, Melson AT, Griffey RT. Feasibility and diagnostic accuracy of brief health literacy and numeracy screening instruments in an urban emergency department. Acad Emerg Med 2014;2:137-46. Crossref

9. Shah LC, West P, Bremmeyr K, Savoy-Moore RT. Health literacy instrument in family medicine: the “newest vital sign” ease of use and correlates. J Am Board Fam Med 2010;23:195-203. Crossref

10. Patel PJ, Joel S, Rovena G, et al. Testing the utility of the newest vital sign (NVS) health literacy assessment tool in older African-American patients. Patient Educ Couns 2011;85:505-7. Crossref

11. Baker DW, Williams MV, Parker RM, Gazmararian JA, Nurss J. Development of a brief test to measure functional health literacy. Patient Educ Couns 1999;38:33-42. Crossref

12. Robinson S, Moser D, Pelter MM, Besbitt T, Paul SM, Dracup K. Assessing health literacy in heart failure patients. J Card Fail 2011;17:887-92. Crossref

13. Leung AY, Lou VW, Cheung MK, Chan SS, Chi I. Development and validation of Chinese health literacy scale for diabetes. J Clin Nurs 2013;22:2090-9. Crossref

14. Leung AY, Cheung MK, Lou VW, et al. Development and validation of the Chinese Health Literacy Scale for chronic care. J Health Comm 2013;18 Suppl 1:205-22. Crossref

15. Leung AY, Lau HF, Chau PH, Chan EW. Chinese Health Literacy Scale for Diabetes–multiple-choice version (CHLSD-MC): a validation study. J Clin Nurs 2015;24:2679-82. Crossref

16. Sullivan LM, Massaro JM, D’Agostino RB Sr. Presentation of multivariate data for clinical use: The Framingham Study risk score functions. Stat Med 2004;23:1631-60. Crossref

17. Li F, He H. Assessing the accuracy of diagnostic tests. Shanghai Arch Psychiatry 2018;30:207-12.

18. Center for Disease Control and Prevention, US Government. About chronic diseases. Available from: https://www.cdc.gov/chronicdisease/about/index.htm. Accessed 17 May 2020.

19. Rhee TG, Lee HY, Kim NK, Han G, Lee J, Kim K. Is health literacy associated with depressive symptoms among Korean adults? Implications for mental health nursing. Perspect Psychiatr Care 2017;53:234-42. Crossref

20. Puente-Maestu L, Calle M, Rodríguez-Hermonsa JL, et al. Health literacy and health outcomes in chronic obstructive pulmonary disease. Respir Med 2016;115:78-82. Crossref

21. Protheroe J, Whittle R, Bartlam B, Estacio EV, Clark L, Kurth J. Health literacy, associated lifestyle and demographic factors in adult population of an English city: a cross-sectional survey. Health Expect 2017;20:112-9. Crossref

22. Liu YB, Liu L, Li YF, Chen YL. Relationship between health literacy, health-related behaviors and health status: a survey of elderly Chinese. Int J Environ Res Public Health 2015;12:9714-25. Crossref

23. Sheik JI, Yesavage JA. Geriatric Depression Scale (GDS): recent evidence and development of a shorter version. Clin Gerontol 1986;5:165-73. Crossref

24. Lee HC, Chiu HF, Kwok WY, et al. Chinese elderly and the GDS short form: a preliminary study. Clin Gerontol 1993;14:37-42.

25. Boey KW, Chiu HF. Assessing psychological well-being of the old-old: a comparative study of GDS-15 and GHQ-12. Clin Gerontol 1998;19:65-75. Crossref

26. Lundberg O, Manderbacka K. Assessing reliability of a measure of self-rated health. Scand J Soc Med 1996;24:218-24. Crossref

27. Nagelkerke NJ. A note on a general definition of the coefficient of determination. Biometrika 1991;78:691-2. Crossref

28. Zweig MH, Campbell G. Receiver-operating characteristic (ROC) plots: a fundamental evaluation tool in clinical medicine. Clin Chem 1993;39:561-77. Crossref

29. Ruopp MD, Perkins NJ, Whitcomb BW, Schisterman EF. Youden Index and optimal cut-point estimated from observations affected by a lower limit of detection. Biom J 2008;50:419-30. Crossref

30. Garson GD. Path analysis. Asheboro (NC): Statistical Associates Publishers; 2014.

31. Muthén LK, Muthén BO. MPlus (version 6) [computer software]. Los Angeles (CA): Muthén & Muthén; 2010.

32. Bentler PM. Comparative fit indexes in structural models. Psychol Bull 1990;107:238-46. Crossref

33. Raykov T, Marcoulides GA. A First Course in Structural Equation Modeling. 2nd ed. Mahwah (NJ): Erlbaum; 2006.

34. Hadden KB, Arnold CL, Curtis LM, et al. Barriers and solutions to implementing a pragmatic diabetes education trial in rural primary care clinics. Contemp Clin Trials Commun 2020;18:100550. Crossref

35. Leung AY, Chau PH, Leung IS, et al. Motivating diabetic and hypertensive patients to engage in regular physical activity: a multi-component intervention derived from the concept of photovoice. Int J Environ Res Public Health 2019;16:1219. Crossref

36. Nouri SS, Rudd RE. Health literacy in the “oral exchange”: an important element of patient-provider communication. Patient Educ Couns 2015;98:565-71. Crossref

37. Yen PH, Leasure AR. Use and effectiveness of the teach-back method in patient education and health outcomes. Fed Pract 2019;36:284-9.

38. Warde F, Papadakos J, Papadakos T, Rodin D, Salhia M, Giuliani M. Plain language communication as a priority competency for medical professionals in a globalized world. Can Med Educ J 2018;9:e52-9. Crossref

39. Delavar F, Pashaeypoor S, Negarandeh R. The effects of self-management education tailored to health literacy on medication adherence and blood pressure control among elderly people with primary hypertension: a randomized controlled trial. Patient Educ Couns 2020;103:336-42. Crossref

40. Leung AY, Leung IS, Liu JW, Ting S, Lo S. Improving health literacy and medication compliance through comic books: a quasi-experimental study in Chinese community-dwelling older adults. Glob Health Promot 2018;25:67-78.Crossref

41. Wu SW, Tse DT, Chui JC, et al. Educational comic book versus pamphlet for improvement of health literacy in older patients with type II diabetes mellitus: a randomized controlled trial. Asian J Gerontol Geriatr 2017;12:60-4.

42. Lee YM, Yu HY, You MA, Son YJ. Impact of health literacy on medication adherence in older people with chronic diseases. Collegian 2017;24:11-8. Crossref

43. Strijbos RM, Hinnen JW, van den Haak RF, Verhoeven BA, Koning OH. Inadequate health literacy in patients with arterial vascular disease. Eur J Vasc Endovasc Surg 2018;56:239-45. Crossref

44. Costa VR, Costa PD, Nakano EY, Apolinário D, Santana AN. Functional health literacy in hypertensive elders at primary health care. Rev Bras Enferm 2019;72 Suppl 2:266-73. Crossref

45. Sharma S, Weathers D. Assessing generalizability of scales used in cross-national research. Int J Res Mark 2003;20:287-95. Crossref

46. Federman AD, Sano M, Wolf MS, Siu AL, Halm EA. Health literacy and cognitive performance among older adults. J Am Geriatr Soc 2009;57:1475-80. Crossref

Mycoplasma genitalium (MG) is capable of causing urogenital infection, especially urethritis, in men. There is increasing evidence to support its ability to cause cervicitis and pelvic inflammatory disease in women.1 Data from a recent meta-analysis showed a prevalence of 1.3% in the general populations of developed countries, with similar rates among men and women; furthermore, the prevalence ranged from 0.6% to 12.6% in clinic-based studies.2 In terms of therapy, macrolides such as azithromycin are considered first-line treatment, while fluoroquinolones (FQ) such as moxifloxacin are considered second-line treatment. However, antibiotic resistance is an emerging problem, with cited rates of macrolide and FQ resistance both reaching approximately 70% in the Asia-Pacific region.3 Consequently, in the context of increasing reports of treatment failure,4 5 multiple international guidelines indicate that all MG-positive specimens should be subjected to testing for macrolide resistance-mediating mutations and test-of-cure purposes.6 7 8

In Hong Kong, testing for MG is not routinely undertaken in the public sector. Notably, a cross-sectional study performed in 2008 in Hong Kong showed respective prevalences of 10% and 2% in symptomatic and asymptomatic men who sought sexual health services.9 Empirical treatment for non-gonococcal urethritis or non-specific genital tract infection encompassing MG infection generally comprises either a single dose of azithromycin or a 1-week course of doxycycline. Data regarding resistance profiles in Hong Kong are not available. This laboratory-based study aimed to determine the prevalences of MG in both symptomatic and asymptomatic women attending sexually transmitted infection (STI) clinics, as well as to determine the rates of macrolide and FQ resistance in MG-positive endocervical swabs, by using molecular methods.

From March to May 2019, endocervical swabs of female patients sent from two STI clinics in Hong Kong for routine molecular detection of Chlamydia trachomatis (CT) and Neisseria gonorrhoeae (NG) were also subjected to detection of MG. Information regarding the presence or absence of genitourinary symptoms was provided by attending physicians. All available laboratory information for each tested patient was reviewed to determine demographic data, history of STI, human immunodeficiency virus status, and test results from the current visit.

The cobas TV/MG assay (Roche Diagnostics, Rotkreuz, Switzerland) performed on the cobas 6800 System (Roche Diagnostics) was used for detection of MG. The cobas TV/MG assay is a CE-marked, Food and Drug Administration–cleared, commercial qualitative nucleic acid test that utilises the real-time polymerase chain reaction for dual detection of Trichomonas vaginalis (TV) and MG. Its reported overall sensitivity and specificity rates were 83.1% and 98.4%, respectively, for MG detection in endocervical swabs from female patients.10 Tests were performed in accordance with the manufacturer’s instructions.

In brief, CT and NG were detected by the cobas CT/NG assay (Roche Diagnostics), while herpes simplex virus (HSV) 1 and HSV 2 were detected by the cobas HSV 1 and 2 assay (Roche Diagnostics), both performed on the cobas 4800 System (Roche Diagnostics). In addition, NG was detected by the culture method on modified Thayer-Martin agars. Either molecular- or culture-based test indicating the presence of NG was considered a positive result. Trichomonas vaginalis was detected by the cobas TV/MG assay (Roche Diagnostics) and wet mount microscopy after enrichment in Feinberg medium. Similar to NG, either test indicating the presence of TV was considered a positive result. Syphilis was detected by serological tests of serum specimens, including enzyme immunoassay (DiaSorin, Saluggia, Italy), Venereal Disease Research Laboratory test (Becton, Dickinson and Company, Franklin Lakes [NJ], US), fluorescent treponemal antibody absorption test (ImmunoDiagnostics Limited, Hong Kong, China), or Treponema pallidum passive particle agglutination test (Fujirebio Diagnostics AB, Göteborg, Sweden).

Mutations at nucleotide positions 2071 and 2072 (2058 and 2059, respectively, by Escherichia coli numbering) of the 23S rRNA gene have been associated with macrolide resistance in MG and subsequent treatment failure11 12; mutations in the quinolone resistance-determining region of the parC gene, and possibly the gyrA gene, have been associated with FQ resistance.5 Polymerase chain reaction analysis of these genes was performed as described previously.13 14 Sequencing was performed using a 3730xl DNA Analyzer (Applied Biosystems, Foster City [CA], US), in accordance with the manufacturer’s instructions. Primers used for sequencing were the same as those used for the polymerase chain reaction. Resulting sequences were compared to the sequence of wild-type strain MG G37 by using BLAST (https://blast.ncbi.nlm. nih.gov/Blast.cgi).

Calculation of odds ratios, and the Chi squared test or Fisher’s exact test, were performed as univariate analysis to identify associations between assessed factors (ie, age, previous STI clinic visit, history of STI, symptoms, and sexually transmitted co-infections) and outcomes (ie, detection of MG and detection of resistance mutations), as well as between STI and symptoms. Because of the low outcome frequency, Firth logistic regression was employed to analyse associations between factors with P<0.25 in univariate analysis and the detection of resistance mutations. IBM SPSS Statistics Subscription (Windows version, IBM Corp, Armonk [NY], US) was used for data analysis.

The STROBE statement reporting guidelines were followed in this study.

In total, 285 non-duplicated specimens from 285 patients were included in this study. The mean patient age was 35.5 years (range, 16-76 years); 23.9% of the patients (n=68) were in the younger age-group (≤25 years). Of the 285 patients, 59.6% (n=170) were new patients without a previous relevant testing record. In all, 18.9% of the patients (n=54) had a documented history of STI. None were known human immunodeficiency virus carriers; however, human immunodeficiency virus status was not available for seven patients. Regarding the current clinic visit, 60.7% of the patients (n=173) were symptomatic, and 7.4% (n=21) had MG-positive test results (14 specimens exhibited MG alone; 7 specimens exhibited MG and another pathogen). Table 1 shows the numbers of STIs detected among these 285 patients.

Younger age, previous STI clinic visit, history of STI, symptoms, and sexually transmitted co-infections were not associated with the detection of MG (Table 2). Symptoms were only significantly associated with the detection of CT (Table 3).

Among the 21 MG-positive endocervical swabs, sequencing results for 23S rRNA were available for 19 specimens. In total, 42.1% of the specimens (n=8) harboured the macrolide resistance-mediating mutations A2071G or A2072G. Regarding parC and gyrA, sequencing results were available for 20 specimens. Overall, 65% of the specimens (n=13) harboured the FQ resistance–related mutations G248T (Ser83Ile), G259T (Asp87Tyr), or G259A (Asp87Asn) within parC; no mutations were detected in gyrA. Among the 19 specimens with sequencing results available regarding both macrolide and FQ resistance, dual resistance was detected in 42.1% of the specimens (n=8); thus, all macrolide-resistant strains were also resistant to FQ. Furthermore, C184T (Pro62Ser) in parC, for which clinical significance is unknown, was detected in one specimen without any other mutation and in one specimen with dual resistance.

In subgroup analysis of patients with MG-positive specimens, younger age, previous STI clinic visit, symptoms, and sexually transmitted co-infections were not associated with the detection of resistance mutations, similar to the findings among all specimens; however, a history of STI was negatively associated with the detection of mutations (Table 2). This association did not remain after multivariable logistic regression (odds ratio=0.151, 95% confidence interval=0.004-2.983, P=0.221).

In this study, 7.4% of endocervical swabs from women attending STI clinics exhibited MG-positive results, although no assessed factors were obviously associated with the detection of MG. Mycoplasma genitalium–positive rates did not significantly differ (P=0.131) among patients who were symptomatic (9.2%) and those who were asymptomatic (4.5%). These rates were comparable with the findings of the aforementioned 2008 cross-sectional study in Hong Kong involving male patients with STI,9 as well as with the findings of a multicentre clinical study in the US.15 Reported rates of MG detection have exhibited considerable variability. A 2018 systematic review2 revealed that higher rates were prevalent among at-risk groups (eg, commercial sex workers and men who have sex with men), in clinic-based settings, and in countries with lower economic development. In Australia, the prevalence of MG was found to range from 2.1% to 13%, depending on the population tested16; while a higher prevalence of approximately 15% has been reported in Japan.17 Nevertheless, the prevalence in the general population and asymptomatic patients remained low (1.3%),2 which did not support universal screening. In the present study, symptoms were not associated with the detection of MG. Considering the organism’s uncertain clinical significance and natural history, it is necessary to balance the need to test and the risks of unnecessary treatment, including potential aggravation of antibiotic resistance. We agree with the existing recommendation that testing should be carefully selected, reserved only for patients with increased risks or for whom treatment has failed, as well as their contacts.18

Importantly, the choice of specimen might affect the detection of MG. A recent prospective, multicentre study showed that self or clinician-collected vaginal swabs exhibited the best sensitivities (92%-98.9%), while urinary and endocervical swabs were less sensitive (81.5% and 77.8%, respectively).15 These findings were consistent with the results of prior studies19 20 21; notably, some studies found that endocervical swabs were more sensitive than urinary specimens,20 21 presumably because of the lower bacterial load in urine.22 Endocervical swabs are the routine specimens sent to our laboratory from STI clinics for molecular detection of CT and NG; we perform assays for detection of MG and CT, as recommended by European guidelines.6 Of note, the relatively low sensitivity of the test with respect to endocervical swabs might also have underestimated the prevalence of MG in our study.

Resistance-related mutations in 23S rRNA and parC genes were detected in 42.1% and 65% of MG-positive samples, respectively, among which none harboured mutations in gyrA. All macrolide-resistant strains were also FQ-resistant (42.1%). Although the populations have differed among studies, similarly high rates of macrolide resistance have been reported, while rates of FQ and dual resistance have varied among regions (Table 4).3 23 24 25 26 27 28 Several studies also demonstrated consistent increases in resistance rates over time.3 28 29

All mutations detected in this study have been described previously; while C184T (Pro62Ser) in parC is of unknown significance, others are known to confer antibiotic resistance leading to higher minimal inhibitory concentrations and treatment failure.5 11 12 23 In particular, the extent of FQ resistance is reportedly related to the presence of concurrent parC and gyrA mutations. Although MG strains with lone parC mutations had reduced susceptibilities to FQ, they were able to be eradicated by sitafloxacin and (possibly) moxifloxacin. However, concurrent gyrA mutations have been shown to further increase the FQ minimal inhibitory concentrations, leading to treatment failure.23 30

No assessed factors were significantly associated with the detection of MG in this study, possibly due to the limited number of positive samples. The authors of other studies have suggested that a syndromic approach (ie, management of a patient whereby a syndrome is used as a basis for the treatment of the causative organisms) and the use of a single dose of azithromycin for treatment of NG (as part of dual therapy), non-gonococcal urethritis/non-specific genital tract infection, or known MG infection contribute to the emergence of macrolide resistance in MG, because this regimen is suboptimal and might exert selective pressure on resistant strains.28 29 31 A similar phenomenon has been observed with respect to FQ, especially in Japan, where frequent use of the second-line antibiotic sitafloxacin caused selection of resistant strains, leading to high rates of FQ resistance in MG.3 In public clinics in Hong Kong, a single dose of azithromycin or a 1-week course of doxycycline is used as empirical treatment for non-gonococcal urethritis or non-specific genital tract infection. If no culprit pathogen is identified and the patient complains of persistent symptoms during follow-up, a 1-week course of moxifloxacin for possible MG is considered, following exclusion of other causes (eg, non-compliance). These empirical uses of macrolide and FQ regimens might explain the high rates of resistance found in this study. For other regions with lower rates of FQ resistance, the relationship between FQ use and emergence of its resistance in MG requires further investigation.

In the context of increasing drug resistance, international guidelines have suggested follow-up molecular testing for resistance determinants in MG-positive specimens.6 7 8 In particular, the most recent British and Australian guidelines include revised treatment regimens, which suggest 1 week of doxycycline followed by 3 days of azithromycin as treatment for macrolide-sensitive (or susceptibility unknown) MG, or followed by 7 to 10 days of moxifloxacin as treatment for macrolide-resistant MG.7 8 Alternative antibiotics that might be effective (eg, pristinamycin) require further evaluation.32 33

When the detection of MG and its drug resistance profile is considered after patient selection and careful review of clinical indications, testing at a private laboratory may be sought, because this service is not readily available in the public sector in Hong Kong. Our findings of high macrolide and FQ resistance rates in MG implied that the use of azithromycin and moxifloxacin as empirical first- and second-line therapies, respectively, might be ineffective; furthermore, this approach could induce greater drug resistance. These findings should be taken into consideration in future assessments of treatment guidelines for Hong Kong. The acquisition of updated treatment strategies from international guidelines may also be useful.

Because our laboratory is a reference laboratory that serves all public STI clinics in Hong Kong, our database is comprehensive in terms of laboratory testing records. Specimens in this study were unique and not duplicated for any patient. In addition, we have considerable capacity to perform arrays of confirmatory tests for various sexually transmitted pathogens. However, this study was limited by the absence of other clinical information such as sexual practices, antimicrobials prescribed, and treatment outcomes, because these data were not available to the authors. The small number of MG-positive samples also limited our ability to assess correlations with factors considered in this study.

Mycoplasma genitalium was detected in 7.4% of 285 endocervical swabs collected from both symptomatic and asymptomatic women attending STI clinics in Hong Kong. Among the MG-positive samples, macrolide resistance-mediating mutations and fluoroquinolone resistance–related mutations were detected in 42.1% and 65%, respectively. Dual resistance was also detected in all macrolide-resistant strains (42.1%). These findings suggest that both testing and treatment strategies require careful review to avoid further enhancing the prevalence of antibiotic resistance.

Concept or design: KKM Ng, PKL Leung.
Acquisition of data: All authors.
Analysis or interpretation of data: All authors.
Drafting of the manuscript: KKM Ng.
Critical revision of the manuscript for important intellectual content: All authors.

All authors had full access to the data, contributed to the study, approved the final version for publication, and take responsibility for its accuracy and integrity.

All authors have disclosed no conflicts of interest.

The authors acknowledge the excellent work and contributions by staff at the Special Investigation Laboratory of Public Health Laboratory Services Branch, Centre for Health Protection, Department of Health, Hong Kong SAR Government.

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

This study obtained ethics approval (Ref LM 424/2019) from the Ethics Committee of the Department of Health, Hong Kong SAR Government. Patients consented to testing for sexually transmitted pathogens.

1. Lis R, Rowhani-Rahbar A, Manhart LE. Mycoplasma genitalium infection and female reproductive tract disease: a meta-analysis. Clin Infect Dis 2015;61:418-26. Crossref

2. Baumann L, Cina M, Egli-Gany D, et al. Prevalence of Mycoplasma genitalium in different population groups: systematic review and meta-analysis. Sex Transm Infect 2018;94:255-62. Crossref

3. Deguchi T, Ito S, Yasuda M, et al. Surveillance of the prevalence of macrolide and/or fluoroquinolone resistance-associated mutations in Mycoplasma genitalium in Japan. J Infect Chemother 2018;24:861-7. Crossref

4. Lau A, Bradshaw CS, Lewis D, et al. The efficacy of azithromycin for the treatment of genital Mycoplasma genitalium: a systematic review and meta-analysis. Clin Infect Dis 2015;61:1389-99. Crossref

5. Murray GL, Bradshaw CS, Bissessor M, et al. Increasing macrolide and fluoroquinolone resistance in Mycoplasma genitalium. Emerg Infect Dis 2017;23:809-12. Crossref

6. Jensen JS, Cusini M, Gomberg M, Moi H. 2016 European guideline on Mycoplasma genitalium infections. J Eur Acad Dermatol Venereol 2016;30:1650-6. Crossref

7. Australasian Sexual Health Alliance. Australian STI management guidelines for use in primary care. Available from: http://www.sti.guidelines.org.au/sexually-transmissible-infections/mycoplasma-genitalium. Accessed 15 Mar 2020.

8. Soni S, Horner P, Rayment M, et al. British Association for Sexual Health and HIV national guideline for the management of infection with Mycoplasma genitalium (2018). Int J STD AIDS 2019;30:938-50. Crossref

9. Yu JT, Tang WY, Lau KH, et al. Role of Mycoplasma genitalium and Ureaplasma urealyticum in non-gonococcal urethritis in Hong Kong. Hong Kong Med J 2008;14:125-9.

10. US Food and Drug Administration. cobas TV/MG Premarket Notification 510(k). Table 34. Available from: https://www.accessdata.fda.gov/cdrh_docs/pdf19/K190433.pdf. Accessed 15 Mar 2020.

11. Jensen JS, Bradshaw CS, Tabrizi SN, Fairley CK, Hamasuna R. Azithromycin treatment failure in Mycoplasma genitalium–positive patients with nongonococcal urethritis is associated with induced macrolide resistance. Clin Infect Dis 2008;47:1546-53. Crossref

12. Bissessor M, Tabrizi SN, Twin J, et al. Macrolide resistance and azithromycin failure in a Mycoplasma genitalium– infected cohort and response of azithromycin failures to alternative antibiotic regimens. Clin Infect Dis 2015;60:1228-36. Crossref

13. Peuchant O, Ménard A, Renaudin H, et al. Increased macrolide resistance of Mycoplasma pneumoniae in France directly detected in clinical specimens by real-time PCR and melting curve analysis. J Antimicrob Chemother 2009;64:52-8. Crossref

14. Shimada Y, Deguchi T, Nakane K, et al. Emergence of clinical strains of Mycoplasma genitalium harbouring alterations in ParC associated with fluoroquinolone resistance. Int J Antimicrob Agents 2010;36:255-8. Crossref

15. Gaydos CA, Manhart LE, Taylor SN, et al. Molecular testing for Mycoplasma genitalium in the United States: results from the AMES Prospective Multicenter Clinical Study. J Clin Microbiol 2019;57:e01125-19. Crossref

16. Trevis T, Gossé M, Santarossa N, Tabrizi S, Russell D, McBride WJ. Mycoplasma genitalium in the Far North Queensland backpacker population: an observational study of prevalence and azithromycin resistance. PLoS One 2018;13:e0202428. Crossref

17. Hamasuna R. Mycoplasma genitalium in male urethritis: diagnosis and treatment in Japan. Int J Urol 2013;20:676- 84. Crossref

18. Stewart JD, Webb BQ, Francis M, Graham M, Korman TM. Should we routinely test for Mycoplasma genitalium when testing for other sexually transmitted infection? Med J Aust 2020;212:30-1. Crossref

19. Jensen JS, Björnelius E, Dohn B, Lidbrink P. Comparison of first void urine and urogenital swab specimens for detection of Mycoplasma genitalium and Chlamydia trachomatis by polymerase chain reaction in patients attending a sexually transmitted disease clinic. Sex Transm Dis 2004;31:499-507. Crossref

20. Wroblewski JK, Manhart LE, Dickey KA, Hudspeth MK, Totten PA. Comparison of transcription-mediated amplification and PCR assay results for various genital specimen types for detection of Mycoplasma genitalium. J Clin Microbiol 2006;44:3306-12. Crossref

21. Lillis RA, Nsuami MJ, Myers L, Martin DH. Utility of urine, vaginal, cervical, and rectal specimens for detection of Mycoplasma genitalium in women. J Clin Microbiol 2011;49:1990-2. Crossref

22. Murray GL, Danielewski J, Bodiyabadu K, et al. Analysis of infection loads in Mycoplasma genitalium clinical specimens by use of a commercial diagnostic test. J Clin Microbiol 2019;57:e00344-19. Crossref

23. Hamasuna R, Le PT, Kutsuna S, et al. Mutations in ParC and GyrA of moxifloxacin-resistant and susceptible Mycoplasma genitalium strains. PLoS One 2018;13:e0198355. Crossref

24. Sweeney EL, Trembizki E, Bletchly C, et al. Levels of Mycoplasma genitalium antimicrobial resistance differ by both region and gender in the state of Queensland, Australia: implications for treatment guidelines. J Clin Microbiol 2019;57:e01555-18.Crossref

25. Vesty A, McAuliffe G, Roberts S, Henderson G, Basu I. Mycoplasma genitalium antimicrobial resistance in community and sexual health clinic patients, Auckland, New Zealand. Emerg Infect Dis 2020;26:332-5. Crossref

26. Getman D, Jiang A, O’Donnell M, Cohen S. Mycoplasma genitalium prevalence, coinfection, and macrolide antibiotic resistance frequency in a multicenter clinical study cohort in the United States. J Clin Microbiol 2016;54:2278-83. Crossref

27. Dionne-Odom J, Geisler WM, Aaron KJ, et al. High prevalence of multidrug-resistant Mycoplasma genitalium in human immunodeficiency virus–infected men who have sex with men in Alabama. Clin Infect Dis 2018;66:796- 8. Crossref

28. Fernández-Huerta M, Barberá MJ, Serra-Pladevall J, et al. Mycoplasma genitalium and antimicrobial resistance in Europe: a comprehensive review. Int J STD AIDS 2020;31:190-7. Crossref

29. Martens L, Kuster S, de Vos W, Kersten M, Berkhout H, Hagen F. Macrolide-resistant Mycoplasma genitalium in southeastern region of the Netherlands, 2014-2017. Emerg Infect Dis 2019;25:1297-303. Crossref

30. Murray GL, Bodiyabadu K, Danielewski J, et al. Moxifloxacin and sitafloxacin treatment failure in Mycoplasma genitalium infection: association with parC mutation G248T (S83I) and concurrent gyrA mutations. J Infect Dis 2020;221:1017-24.

31. Horner P, Ingle SM, Garrett F, et al. Which azithromycin regimen should be used for treating Mycoplasma genitalium? A meta-analysis. Sex Transm Infect 2018;94:14-20. Crossref

32. Bradshaw CS, Jensen JS, Waites KB. New horizons in Mycoplasma genitalium treatment. J Infect Dis 2017;216:S412-9. Crossref

33. Read TR, Jensen JS, Fairley CK, et al. Use of pristinamycin for macrolide-resistant Mycoplasma genitalium infection. Emerg Infect Dis 2018;24:328-35. Crossref

Streptococcus pneumoniae causes a wide range of diseases that include middle ear infection, sinusitis, pneumonia, and meningitis. As one of the most common pathogens in community-acquired pneumonia (especially in Western countries), S pneumoniae infection contributed to 1.6 million deaths in 2010 and 3.7 million severe pneumococcal infections worldwide in 2015.1 2 3

Streptococcus pneumoniae is a gram-positive encapsulated bacterium that colonises human nasopharynx and is mainly transmitted via respiratory droplets, which cause middle ear and respiratory tract infection. Thus far, more than 90 serotypes of S pneumoniae have been identified. Streptococcus pneumoniae infection can be stratified into invasive and non-invasive disease.4 5 Invasive pneumococcal disease (IPD) is a notifiable disease in Hong Kong. In 2019, there were 187 cases; the incidence has remained similar over the past few years.6 Worldwide, there is growing concern regarding drug-resistant S pneumoniae strains (eg, strains resistant to macrolide, penicillin, and/or fluoroquinolone). However, drug-resistant strains have not been associated with higher mortality rates.7 The prevalence of drug-resistant S pneumoniae is lower in Southeast Asia than in Western countries.1 Despite inconclusive evidence in the literature regarding its association with mortality, IPD is often associated with more severe disease and requires more invasive organ support.8

In this study, we aimed to identify the predictors for 30-day mortality in patients with S pneumoniae infection, as well as predictors for IPD. We also performed subgroup analysis of patients in the intensive care unit (ICU) and identified risk factors for 30-day mortality and IPD in those patients, as well as all patients with S pneumoniae infection.

This retrospective cohort study included adults who were admitted to Pamela Youde Nethersole Eastern Hospital, Hong Kong, with pneumococcal infection from 1 January 2011 to 31 December 2018. Patients who were aged <18 years or had incomplete data were excluded.

Patient medical records and data were extracted from clinical management systems and clinical information systems (IntelliVue Clinical Information Portfolio; Philips Medical, Amsterdam, The Netherlands). Baseline demographics, clinical characteristics, and microbiological data were identified. For patients in the ICU, disease severity was quantified using APACHE (Acute Physiology and Chronic Health Evaluation) IV scores. The use of invasive organ support was recorded, including continuous renal replacement therapy, inotropes, invasive mechanical ventilation, and extracorporeal membrane oxygenation. The primary outcome was 30-day all-cause mortality; secondary outcomes were ICU and hospital mortalities, ICU and hospital length of stay (LOS), and ICU ventilator days.

Pneumococcal infection was determined by positive culture of S pneumoniae. Invasive pneumococcal disease was defined as the presence of S pneumoniae in sterile sites (eg, pleural fluid, cerebrospinal fluids, and blood).4 8 Non-invasive pneumococcal disease was defined as the presence of S pneumoniae in non-sterile sites, or a positive urinary antigen test (UAT) result. Medical co-morbidities (eg, diabetes mellitus, chronic kidney disease, heart failure, and haematological malignancies) were coded in accordance with the International Classification of Diseases, Ninth Revision, Clinical Modification. Smokers were defined as those who had ever smoked. Advanced age was defined as age >65 years.

Antibiotic resistance was determined based on Clinical and Laboratory Standards Institute testing criteria for minimal inhibitory concentrations. Breakpoints adopted for determination of parenteral penicillin resistance in non-meningitis S pneumoniae isolates were susceptible, ≤2 μg/mL; intermediate, 4 μg/mL; and resistance, ≥8 μg/mL.9 Breakpoints adopted for determination of parenteral penicillin resistance in meningitis S pneumoniae isolates were susceptible, ≤0.06 μg/mL and resistance, ≥0.12 μg/mL; breakpoints adopted for determination of levofloxacin resistance in S pneumoniae were susceptible, ≤2 μg/mL; intermediate, 4 μg/mL; and resistance, ≥8 μg/mL.9

Urinary antigen test (Alere 710-012 BinaxNOW Streptococcus) results were evaluated in accordance with the manufacturer’s instructions.

Characteristics and clinical parameters were compared between patients with IPD and those with non-invasive pneumococcal disease, as well as between 30-day survivors and non-survivors. Results were expressed as median (interquartile range) or as numbers (percentages) of cases, as appropriate. For univariate analysis, categorical variables were compared by Pearson Chi squared tests or Fisher’s exact test, as appropriate; continuous variables were compared by using the Mann-Whitney U test. Variables with P<0.2 in univariate analysis or with known clinical significance from previous studies were entered into multivariate analysis. Independent predictors for 30-day mortality and independent predictors for IPD were assessed by logistic regression analysis.8 10 11 12 Subgroup analysis was performed regarding IPD and disease severity among patients in the ICU. Hosmer-Lemeshow test was performed for goodness-of-fit for logistic regression models. Kaplan-Meier survival plots were used to compare cumulative survival between patients with IPD and those with non-invasive pneumococcal disease. SPSS (Mac version 24.0; IBM Corp, Armonk [NY], United States) was used for all statistical analyses.

Patient demographic and clinical characteristics, including co-morbidities and use of invasive organ support, are shown in Table 1. In total, 792 patients with pneumococcal disease were identified during the 8-year study period. The median age was 73 years; patients were predominantly men. Most patients exhibited respiratory tract infection (96.1%) and approximately one quarter of patients had asthma/chronic obstructive pulmonary disease (24.4%). In total, 170 patients received intensive care and 14.1% required invasive mechanical ventilation; 28% required vasopressor use. Invasive pneumococcal disease was present in 13.4% of the patients. The overall hospital mortality rate was 11.2%, while the mortality rate among patients in the ICU was 22.9%.

Invasive pneumococcal disease was associated with a higher 30-day mortality rate (28.6% vs 11.4%, P<0.001); a positive UAT result was also associated with a higher 30-day mortality rate (36.3% vs 12.7%, P<0.001). Logistic regression analysis identified statistically significant predictors for 30-day mortality, which are shown in Table 1. Patients with vasopressor use (odds ratio [OR]=4.96, P<0.001), chronic kidney disease (OR=3.62, P<0.001), a positive UAT result (OR=2.57, P=0.001), and older age (OR=2.19, P=0.010) exhibited comparatively higher 30-day mortality rates; however, asthma/chronic obstructive pulmonary disease was not an independent predictor for mortality in logistic regression analysis. The Figure depicts the results of Kaplan-Meier survival analysis comparing patients with IPD and those with non-invasive pneumococcal disease.

Table 2 shows the characteristics of patients with IPD and those with non-invasive pneumococcal disease. More patients with asthma/chronic obstructive pulmonary disease exhibited non-invasive pneumococcal disease (26.5% vs 10.4%, P<0.001). Invasive pneumococcal disease was more likely to be associated with renal failure (27.4% vs 9.6%, P<0.001) and haematological malignancy (5.7% vs 1.7%, P=0.012). Additionally, IPD was associated with higher rates of ICU admission (33.0% vs 19.7%, P=0.002), renal replacement therapy (16.0% vs 4.8%, P<0.001), and vasopressor use (93.4% vs 17.9%, P<0.001). Patients with IPD had a higher 30-day mortality rate (24.5% vs 9.5%, P<0.001) and longer hospital LOS (8 vs 4 days, P<0.001). Independent risk factors for IPD by logistic regression analysis are shown in Table 2, along with their ORs. Notably, chronic kidney disease (OR=3.10, P<0.001) was the sole independent predictor for IPD.

The results of ICU subgroup analysis are shown in Table 3. Respiratory tract infection constituted 93.5% of all S pneumoniae infections. The rate of IPD was 20.6% among patients in the ICU with S pneumoniae infection, which was higher than the rate among all patients with S pneumoniae infection. Further analysis revealed that IPD was associated with higher rates of complications and invasive organ support; in particular, more patients with IPD required renal replacement therapy (48.6% vs 24.4%, P=0.005) and vasopressor use (100% vs 88.9%, P=0.039). Additionally, more patients with IPD tended to exhibit pleural effusion/empyema, although this difference was not statistically significant. Patients who required invasive mechanical ventilation (76.0% vs 57.5%, P=0.023), extracorporeal membrane oxygenation (14.0% vs 5.0%, P=0.044), renal replacement therapy (48.0% vs 21.7%, P=0.001), and vasopressor use (98.0% vs 88.3%, P=0.043) exhibited significantly higher 30-day mortality rates. Logistic regression analysis showed that chronic kidney disease (OR=4.64, P<0.001), higher APACHE IV score (OR=3.73, P=0.016), and a positive UAT result (OR=2.94, P=0.008) were independent predictors for 30-day mortality among patients in the ICU who had IPD (Table 3).

The overall mortality rate was 28.6% for patients with IPD and 11.4% for patients without IPD. The case fatality rate in our cohort was higher than that in a previous cohort from the Netherlands, but similar to the rate in a previous study from Korea.5 13 A higher number of co-morbid diseases, worse immune function, impaired mucociliary clearance, and older age are associated with a higher risk of mortality in patients with pneumococcal infection.4

Chronic conditions such as chronic lung disease, heart failure, and diabetes, as well as smoking status, were previously shown to be associated with pneumococcal disease and IPD.10 11 Consistent with the results of prior studies, we found that patients with heart failure (8.8% vs 3.3%, P=0.011) and haematological malignancies (5.7% vs 1.7%, P=0.012) exhibited significantly higher 30-day mortality rates in univariate analysis. Surprisingly, we found a negative association between chronic lung disease and mortality. In post-hoc analysis, we found that patients with chronic lung disease (ie, asthma/chronic obstructive pulmonary disease) also had a lower rate of invasive organ support (15.0% vs 32.7%, P<0.001). This group of patients may be under constant medical surveillance; thus, they may seek medical attention and receive antibiotics earlier than patients without chronic lung disease. Importantly, we did not examine the management and status of underlying lung conditions, which may have affected mortality in these patients.

In our cohort, 122 patients were diagnosed with pneumococcal infection by using the UAT. In our hospital, the first patient was diagnosed by using the UAT in 2015. Use of the UAT in diagnosing community-acquired pneumonia has since increased; thus, in 2018, 71 of 146 patients (48.6%) were diagnosed by using the UAT. A positive UAT result was a consistent independent predictor for 30-day mortality among patients in the ICU, as well as among all patients. Post-hoc analysis showed that a positive UAT result was significantly associated with ICU admission (34.7% vs 10.1%, P<0.001). However, it was not significantly associated with ICU LOS (6.16 vs 8.43 days, P=0.515) or hospital LOS (21.46 vs 29.78 days, P=0.415).

The pneumococcal UAT assay detects the C-polysaccharide antigen of S pneumoniae, which is present in all serotypes, from urine samples.14 Fluorescence immunoassay and immunochromatographic test methods provide similar results in terms of diagnosing pneumococcal disease.15 While the UAT result remains positive for up to 3 days after initiation of antibiotic treatment, the UAT increases the diagnostic yield of pneumococcal disease relative to the yield of sputum culture of S pneumoniae; notably, the yield of such sputum cultures markedly decreases after initiation of antibiotic treatment.16 This test provides a rapid and simple method for diagnosis of patients with suspected S pneumoniae infection; it is particularly helpful in the diagnosis of patients who cannot produce sputum for cultures. The test sensitivity and specificity were approximately 60% and 99%, respectively.16 Because of the high test specificity, the UAT helps to reduce the costs of further diagnostic tests and aids in selection of empirical antibiotic treatment. It is recommended in the Infectious Diseases Society of America/American Thoracic Society guidelines for aiding the rapid identification of pneumococcal disease in adults.17 Urinary antigen tests were also found to predict the severity and outcomes of pneumonia. A Korean group found that patients with positive UAT results exhibited greater severity of disease; however, the test results were associated with rates of ICU admission and mortality.14

Counterindications for the UAT include recent pneumococcal disease within 3 months; moreover, it may cross-react with antigens from other streptococcal bacteria.16 18 Patients with acute kidney injury due to sepsis, as well as those with oliguria or anuria of various aetiologies may not be able to provide urine samples for use in the UAT.

In our cohort, IPD was associated with a higher 30-day mortality rate; however, this association did not remain statistically significant in logistic regression analysis. Consistent with the results of previous studies,8 12 we found that patients with IPD exhibited more severe disease and worse outcomes. Moreover, IPD was associated with higher rates of ICU admission, invasive organ support (ie, vasopressor use), and renal replacement therapy, as well as longer hospital LOS. The findings might be explained by the higher bacterial load in patients with IPD, which may lead to worse outcomes.

Similar to the study by Ceccato et al,8 we did not identify a positive relationship between smoking and IPD. Thus far, results regarding the relationship of smoking with IPD have been inconsistent; the association varies according to local smoking prevalence.11 With the implementation of effective smoking cessation programmes and corresponding legislation in Hong Kong, approximately 10% of individuals >15 years of age report daily cigarette consumption; this is markedly lower than the rates in other countries.19 20 In our study, smoking status information was extracted from patient records stored in the Hospital Authority Clinical Management System and nursing notes; thus, we may have underestimated the number of smokers in this cohort. Other important aspects of smoking (eg, number of pack-years and passive smoking) were not available for inclusion in this analysis.

Chronic kidney disease has been consistently associated with IPD. A large retrospective observational cohort of 36 million adults revealed a risk ratio of 21.67 for development of IPD among patients with chronic kidney disease.21 A Japanese registry showed that the relative risk for IPD among patients with chronic kidney disease ranged from 12.4 to 51.3.10 Notably, chronic kidney disease was consistently one of the most important predictors for 30-day mortality among all patients (OR=3.62, P<0.001) and among patients in the ICU (OR=4.64, P<0.001).

Patients with IPD tended to experience a higher rate of complications and require higher rates of invasive organ support. In particular, patients with IPD more frequently exhibited pleural effusion/empyema; they also more frequently required invasive mechanical ventilation, extracorporeal membrane oxygenation, renal replacement therapy, and vasopressor use. Our sample size may not have been sufficiently powered to demonstrate statistically significant results regarding the ICU subgroup; thus, future studies focused specifically on patients in the ICU may be needed. Other aspects of IPD and use of rescue therapies for acute respiratory distress syndrome (eg, prone ventilation, muscle paralytic agents, and inhaled nitrogen oxide) should be investigated in the future.

Penicillin non-susceptible S pneumoniae was not common in the present study; it was only observed in 2.4% of patients. Non-susceptibility to levofloxacin was observed in 0.9% of patients. Drug non-susceptible S pneumoniae were not significantly associated with 30-day mortality (penicillin non-susceptible S pneumoniae was present in two non-survivors and 14 survivors, P=0.641; levofloxacin non-susceptible S pneumoniae was present in zero non-survivors and six survivors, P=1.000). However, these results should be carefully interpreted, because of the small number of drug non-susceptible S pneumoniae in our cohort. According to a recent study in Hong Kong, the penicillin resistance rate was approximately 7% and the levofloxacin resistance rate was 0%.22

Viral-bacterial interactions have been described with respect to pneumococcal disease.23 An epidemiological study regarding the 2009 H1N1 influenza pandemic period showed a significant increase in the number of pneumococcal pneumonia hospitalisations.24 However, viral co-infection was not associated with IPD or mortality in our findings. Notably, an age-specific interaction was described between influenza and IPD; specifically, patients aged 5 to 19 years were significantly more frequently affected, compared with other age-groups.24 25

Thus far, this is the first and largest study regarding pneumococcal disease in adults in Hong Kong; it provides clinical and outcome data in both general ward and intensive care subgroups to allow a comprehensive overview of pneumococcal disease in the locality. It is a standard practice in our centre to check urinary antigens and perform blood cultures for nearly all patients with suspected pneumonia to facilitate accurate diagnosis and avoid missed diagnoses. By including data regarding invasive organ support and ICU admission, we were able to identify and describe complications of pneumococcal disease and determine the broader clinical characteristics of affected patients.

However, because of changes in vaccination programmes, the influenza and pneumococcal vaccination statuses were not available for analysis in the current study. Because of the limited number of patients with drug non-susceptible S pneumoniae in the present cohort, further robust analyses regarding antibiotic sensitivity patterns and appropriateness of antimicrobial treatment could not be performed. Furthermore, capsular serotypes of S pneumoniae among patients in our cohort were not available for analysis. Future studies focused on capsular subtypes of S pneumoniae will facilitate understanding of pneumococcal disease. Because this was a retrospective study, it was subject to potential confounding factors. Finally, the results of this single-centre study may not be generalisable to other countries with higher prevalences of drug non-susceptible S pneumoniae infection.

Pneumococcal disease is associated with high rates of morbidity and mortality. In this cohort, vasopressor use, chronic kidney disease, advanced age, and positive UAT results were predictors for 30-day mortality.

Concept or design: MY Man, HP Shum.
Acquisition of data: MY Man, HP Shum.
Analysis or interpretation of data: MY Man, HP Shum.
Drafting of the manuscript: MY Man, A Wu.
Critical revision of the manuscript for important intellectual content: All authors.

All authors had full access to the data, contributed to the study, approved the final version for publication, and take responsibility for its accuracy and integrity.

All authors have disclosed no conflicts of interest.

The abstract of this study was accepted as an oral presentation at the Annual Scientific Meeting of the Hong Kong Society of Critical Care Medicine on 8 December 2019.

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

This study was approval by the Hospital Authority Hong Kong East Cluster Research Ethics Committee (Ref HKECREC- 2019-065). The requirement for written informed consent was waived.

1. Aliberti S, Cook GS, Babu BL, et al. International prevalence and risk factors evaluation for drug-resistant Streptococcus pneumoniae pneumonia. J Infect 2019;79:300-11. Crossref

2. Black RE, Cousens S, Johnson HL, et al. Global, regional, and national causes of child mortality in 2008: a systematic analysis. Lancet 2010;375:1969-87. Crossref

3. Wahl B, O’Brien KL, Greenbaum A, et al. Burden of Streptococcus pneumoniae and Haemophilus influenzae type b disease in children in the era of conjugate vaccines: global, regional, and national estimates for 2000-15. Lancet Glob Health 2018;6:e744-57. Crossref

4. Drijkoningen JJ, Rohde GG. Pneumococcal infection in adults: burden of disease. Clin Microbiol Infect 2014;20 Suppl 5:45-51. Crossref

5. Song JY, Choi JY, Lee JS, et al. Clinical and economic burden of invasive pneumococcal disease in adults: a multicenter hospital-based study. BMC Infect Dis 2013;13:202. Crossref

6. Centre for Health Protection, Department of Health, Hong Kong SAR Government. Number of notifiable infectious diseases by month. 2019. Available from: https://www.chp. gov.hk/en/statistics/data/10/26/43/6830.html. Accessed 5 May 2020.

7. Cillóniz C, de la Calle C, Dominedò C, et al. Impact of cefotaxime non-susceptibility on the clinical outcomes of bacteremic pneumococcal pneumonia. J Clin Med 2019;8:1150. Crossref

8. Ceccato A, Torres A, Cilloniz C, et al. Invasive disease vs urinary antigen-confirmed pneumococcal community-acquired pneumonia. Chest 2017;151:1311-9. Crossref

9. Clinical and Laboratory Standards Institute. M100 Performance standards for antimicrobial susceptibility testing. 29th ed. CLSI supplement M100. Available from: https://clsi.org/media/2663/m100ed29_sample.pdf. Accessed 22 Feb 2020.

10. Imai K, Petigara T, Kohn MA, et al. Risk of pneumococcal diseases in adults with underlying medical conditions: a retrospective, cohort study using two Japanese healthcare databases. BMJ Open 2018;8:e018553. Crossref

11. Torres A, Blasi F, Dartois N, Akova M. Which individuals are at increased risk of pneumococcal disease and why? Impact of COPD, asthma, smoking, diabetes, and/or chronic heart disease on community-acquired pneumonia and invasive pneumococcal disease. Thorax 2015;70:984-9. Crossref

12. Heo JY, Seo YB, Choi WS, et al. Incidence and case fatality rates of community-acquired pneumonia and pneumococcal diseases among Korean adults: catchment population-based analysis. PLoS One 2018;13:e0194598. Crossref

13. van Mens SP, van Deursen AM, de Greeff SC, et al. Bacteraemic and non-bacteraemic/urinary antigen-positive pneumococcal community-acquired pneumonia compared. Eur J Clin Microbiol Infect Dis 2015;34:115-22. Crossref

14. Kim B, Kim J, Jo YH, et al. Prognostic value of pneumococcal urinary antigen test in community-acquired pneumonia. PLoS One 2018;13:e0200620. Crossref

15. Olofsson E, Özenci V, Athlin S. Evaluation of the sofia S. pneumoniae FIA for detection of pneumococcal antigen in patients with bloodstream infection. J Clin Microbiol 2019;57:e01535-18. Crossref

16. Molinos L, Zalacain R, Menéndez R, et al. Sensitivity, specificity, and positivity predictors of the pneumococcal urinary antigen test in community-acquired pneumonia. Ann Am Thorac Soc 2015;12:1482-9. Crossref

17. Mandell LA, Wunderink RG, Anzueto A, et al. Infectious Diseases Society of America/American Thoracic Society consensus guidelines on the management of community-acquired pneumonia in adults. Clin Infect Dis 2007;44 Suppl 2:S27-72.

18. Blaschke AJ. Interpreting assays for the detection of Streptococcus pneumoniae. Clin Infect Dis 2011;52 Suppl 4: S331-7.Crossref

19. Census and Statistics Department, Hong Kong SAR Government. Thematic Household Survey Report—Report No. 70—Pattern of smoking. Available from: https://www.censtatd.gov.hk/hkstat/sub/sp453. jsp?productCode=C0000047. Accessed 5 May 2020.

20. Wang TW, Asman K, Gentzke AS, et al. Tobacco product use among adults—United States, 2017. MMWR Morb Mortal Wkly Rep 2018;67:1225-32. Crossref

21. Zhang D, Petigara T, Yang X. Clinical and economic burden of pneumococcal disease in US adults aged 19-64 years with chronic or immunocompromising diseases: an observational database study. BMC Infect Dis 2018;18:436. Crossref

22. Chan KC, Ip M, Chong PS, Li AM, Lam HS, Nelson EA. Nasopharyngeal colonisation and antimicrobial resistance of Streptococcus pneumoniae in Hong Kong children younger than 2 years. Hong Kong Med J 2018;24 Suppl 6:4-7.

23. Blasi F, Mantero M, Santus P, Tarsia P. Understanding the burden of pneumococcal disease in adults. Clin Microbiol Infect 2012;18 Suppl 5:7-14. Crossref

24. Weinberger DM, Simonsen L, Jordan R, Steiner C, Miller M, Viboud C. Impact of the 2009 influenza pandemic on pneumococcal pneumonia hospitalizations in the United States. J Infect Dis 2012;205:458-65. Crossref

25. Chiavenna C, Presanis AM, Charlett A, et al. Estimating age-stratified influenza-associated invasive pneumococcal disease in England: a time-series model based on population surveillance data. PLoS Med 2019;16:e1002829. Crossref

Warfarin, an oral vitamin K antagonist, has been widely used as anticoagulant therapy for the treatment and prophylaxis of thromboembolic disease. Patients with atrial fibrillation (AF) exhibit elevated risks of mortality and morbidity, including fivefold greater risk of stroke and threefold greater risk of heart failure, compared with individuals without AF.1 2 In patients with prosthetic heart valve (PHV), the incidence of PHV thrombosis was 0.5% to 6% per patient-year, depending on the prosthesis site.3 Warfarin has been shown to significantly reduce the risk of stroke in patients with non-valvular AF and the risk of embolism in patients with PHV.4 5

To ensure the efficacy and safety of warfarin therapy, strict control of the international normalised ratio (INR) is required. One measurement of INR does not indicate whether warfarin dose is appropriate for a given patient. Instead, time in therapeutic range (TTR) is commonly used in clinical practice. According to the European Society of Cardiology Guidelines for the management of AF, the ideal TTR is regarded as 70%.6 However, warfarin control in clinical practice is reportedly unsatisfactory worldwide.7 8 Poor TTR has been associated with elevated risks of major haemorrhage, ischaemic stroke, and all-cause mortality.9

Hong Kong is currently following the European Society of Cardiology Guidelines for the Management of Atrial Fibrillation with respect to warfarin; these guidelines recommend INR control between 2.0 and 3.0 in patients with normal heart valve and between 2.5 and 3.5 in patients with PHV.6 In contrast, the Japanese Guidelines for Pharmacotherapy of Atrial Fibrillation (JCS 2013) recommend INR control between 2.0 and 3.0 in patients aged <70 years or patients with PHV, and between 1.6 and 2.6 in patients aged ≥70 years.10 This recommendation is based on the findings of a study in which the incidence rate of major haemorrhagic complications was determined to be lower at INR between 1.6 and 2.6.11 It remains unknown whether additional benefits would be obtained by application of Japanese guidelines in Hong Kong.

There are extensive drug-drug interactions, drug-herb interactions, and drug-food interactions of warfarin, which may affect anticoagulation control.12 13 To assure the efficacy and safety of warfarin, patient education concerning warfarin is needed.14 15 However, a study in 2008 showed that only one in six patients with AF underwent regular INR examinations in China; patients with AF also commonly exhibited minimal knowledge concerning the importance of regular INR examinations.16

The study aimed to investigate the adequacy of warfarin control in clinical practice in Hong Kong by means of the TTR; it compared warfarin outcome prediction using European and Japanese INR therapeutic ranges as concurrent primary endpoints. Predictors for poor warfarin control were analysed as secondary endpoints. The impacts of TTR on both clinical and economic outcomes were investigated, using the European therapeutic range. Patient knowledge concerning warfarin therapy was also assessed, as were predictors of this knowledge.

The single-centre cohort study was conducted in the Prince of Wales Hospital, which is a regional acute public hospital in Hong Kong. Patients who received warfarin therapy in both the acute coronary syndrome registry and warfarin clinic for at least 1 year and who had their last visit from 1 January 2010 to 31 August 2015 were included. One year of warfarin therapy was presumed to be necessary for patients to develop stable INR.8 Patients aged <41 years and >90 years were excluded, due to the infrequency of warfarin therapy in both age-groups based on hospital records. Data for patient recruitment and subsequent patient review were retrieved through the Clinical Management System, which is a computerised patient medical record system.

Time in therapeutic range was defined as the fraction of INRs in range, with the percentage derived by dividing number of INRs within the therapeutic range by the total number of INRs recorded.17 Ideal TTR was defined as 70%.6 Warfarin indications for individual patients were categorised in four groups: AF, PHV, both AF and PHV, and neither AF nor PHV (eg, deep vein thrombosis and pulmonary embolism). Associations of outcomes and adaptions of either guidelines were subsequently determined.

Predictors of poor warfarin control, using the European therapeutic range, were regarded as secondary endpoints in our study. Patients were stratified into four quartiles according to TTR. Patients with TTR in Quartile 1 were considered to have poor warfarin control. Patients were compared across the four quartiles to identify predictors. Factors included were age, sex, co-morbidities, medication profile, and patient knowledge concerning warfarin therapy. Co-morbidities comprised hypertension, heart failure, thyroid disorder, liver dysfunction, and diabetes mellitus. Ten commonly prescribed medications were chosen for medication profile comparison, based on a pilot study of the first 20 recruited patients. The pilot study was conducted using the same recruitment criteria and the 20 patients were selected at random. All prescribed medications were recorded for these 20 patients. The 10 most commonly prescribed medications included aspirin, hydrochlorothiazide, metoprolol, diltiazem, diclofenac, famotidine, senna, simvastatin, lisinopril, and pantoprazole. For other cardiovascular medications, the potential impact was suspected with their high-frequency use in the cohort and further investigation was performed. The potential impact was detected using ongoing data collection based on low TTR and high thrombotic and bleeding events of patients with certain medications that were not included in the list of 10 medications previously. The investigators evaluated each additional medication carefully and its impact on the clinical outcomes.

Impacts of TTR on clinical outcomes were investigated; patient TTR values were stratified into four quartiles. Thrombotic events, bleeding complications, and overall incidences of complications were assessed. Stroke, pulmonary embolism, acute coronary syndrome, and arterial embolism were included as thrombotic events in our study. Severity of bleeding complications was classified based on discussion at the Control of Anticoagulation Subcommittee of the International Society of Thrombosis and Haemostasis.18 Major bleeding included: (1) fatal bleeding; and/or (2) symptomatic bleeding in a critical area or organ (eg, intracranial, intraspinal, intraocular, retroperitoneal, intraarticular or pericardial, or intramuscular with compartment syndrome); and/or (3) bleeding causing a decline in haemoglobin level of ≥2 g/dL (1.24 mmol/L), or leading to transfusion of ≥2 units of whole blood or red cells. Otherwise, all non-major bleeds were regarded as minor bleeds.

Impacts of TTR, using the European therapeutic range, on economic outcomes were investigated. Costs were calculated per day of warfarin therapy, such that patients’ direct healthcare costs could be calculated regardless of the length of warfarin therapy. Direct healthcare costs related to warfarin (from the healthcare provider perspective) were calculated using the Hong Kong government gazette.19 Costs for INR examinations, procedures (eg, surgery and diagnostic tests, excluding INR examinations), hospitalisation, clinic visits, and overall costs were compared separately.

Patient knowledge concerning warfarin therapy was assessed using the Oral Anticoagulation Knowledge (OAK) test.20 Question 14 of the original test was omitted from our study, because the frequencies of INR tests and follow-up visits were determined by local physicians in Hong Kong. A “Do not know” option was included to minimise random guessing. The assessment was translated into Chinese and performed via phone interviews from 2 January 2016 to 1 April 2016. Patient knowledge was considered satisfactory if a score of ≥75% was achieved.21 Predictors for OAK score performance were identified.

For descriptive statistics, frequencies and percentages were used for categorical variables; means ± standard deviations were used for continuous variables. The Wilcoxon signed rank test, Chi squared test, Fisher’s exact test, and one-way analysis of variance (pairwise comparison with the Tukey method) were used for comparisons of TTR with European and Japanese therapeutic ranges. Fisher’s exact test and Mann-Whitney U test were used to determine the impacts of TTR on clinical and economic outcomes, respectively. An ordinal regression model with stepwise selection was used to identify independent predictors for poor warfarin control. Multiple linear regression with stepwise selection for variables was used to determine predictors for OAK score. Two-sided P values <0.05 were considered statistically significant. All statistical analysis was performed by SPSS (Windows version 22.0; IBM Corp, Armonk [NY], US) and R (version 3.5.3; https://www.r-project.org/).

In total, 259 patients were included in the study; among them, 126 (48.6%) were men. The mean patient age was 67.9±10.4 years. The detailed demographic characteristics of the patients are shown in Table 1.

The overall mean INR was 2.3±0.3. The median follow-up time for included patients was 2065 days (interquartile range=1556-2065). The median number of INR examinations was 46 (interquartile range=33-73). Using the European therapeutic range, 34.5% of all measured INR values were within the therapeutic range. The overall TTR was 40.2±17.1%; 7.7% of patients had ideal TTR during the study period. Using the Japanese therapeutic range, 44.1% of all measured INR values were within the therapeutic range. The overall TTR was 49.1±16.1%; this was significantly higher than the TTR when using the European therapeutic range (P<0.001). Notably, 12.4% of all patients had ideal TTR during the study period.

Mean TTR values for different indications were compared, as shown in Table 2. When using the European therapeutic range, the mean TTR with an indication for AF was significantly higher than both the mean TTR with an indication for PHV (P<0.001) and the mean TTR with an indication for AF and PHV (P<0.001). When using the Japanese therapeutic range, the mean TTR with an indication for AF was also significantly higher than the mean TTR with an indication for both AF and PHV (P<0.001). The mean TTR values were significantly higher when using the Japanese therapeutic range than when using the European therapeutic range within each indication category.

Patients were divided into four quartiles according to their TTR, using the European therapeutic range (Table 3). Predictors were determined by performing regression across the four quartiles. Adjusted odds ratios (ORs) for poor TTR were calculated. The results showed that younger age was associated with worse TTR, as were concurrent use of furosemide, famotidine, or simvastatin.

Clinical outcomes were compared between the two therapeutic ranges (Table 4). Of the 259 patients, 35.9% experienced complications. Of the 39 patients with thrombotic events, 41.0% had recurrent non-ST-elevation myocardial infarction and 33.3% had stroke. Among patients with bleeding complications, 68.8% experienced minor bleeding. Patients with ideal TTR had significantly fewer overall complications and bleeding complications, compared with patients with non-ideal TTR, in both European and Japanese therapeutic ranges. All patients who had complications were those with non-ideal TTR, using the European therapeutic range. When patients were further stratified into quartiles based on TTR using the European therapeutic range, TTR exhibited statistically significant associations with each tested clinical outcome (Table 5).

Healthcare costs are expressed in terms of US$ per year (US$1=HK$7.8), as shown in Table 4. When including all services related to warfarin, average patient costs were US$809.9/year. In terms of economic outcomes, the INR examination, clinical visit, and total healthcare costs were significantly lower for patients with ideal TTR when using either European or Japanese therapeutic ranges. Using the Japanese therapeutic range, patients with ideal TTR also had lower hospitalisation costs. When using the European therapeutic range, healthcare provider costs increased by US$530.1/year for each patient with non-ideal TTR.

In total, 174 patients completed the OAK test, with a mean score of 54.1% correct for the 19 questions used in our version of the test. The mean duration of warfarin therapy for this subgroup of patients during the study period was 4.8±1.4 years. Only 24 (13.8%) patients achieved the satisfactory overall test score of ≥75%. Of the 19 questions in the test, only four were answered correctly by ≥70% of respondents (Table 6).

Multiple linear regression revealed that respondents with older age (adjusted β=-0.17; 95% confidence interval [CI]=-0.23 to -0.11; P=0.001) or co-morbid diabetes (adjusted β=-1.21; 95% CI=-2.29 to -0.12; P=0.03) were more likely to have low scores on the OAK test. In contrast, respondents with co-morbid hypertension (adjusted β=1.68; 95% CI=0.56-2.80; P=0.004) or co-morbid thyroid dysfunction (adjusted β=2.38; 95% CI=0.80-3.97; P=0.003) were more likely to have high scores on the OAK test. Respondents with better TTR tended to be more likely to have high scores on the OAK test, although this difference was not statistically significant (adjusted β=2.73; 95% CI=-0.21-5.68; P=0.069).

The mean TTR observed in our study was lower than that observed in studies performed in Western nations. A meta-analysis of 40 studies using the European therapeutic range identified a mean TTR of 75.2% after 4 to 12 months of warfarin management.22 A study focusing on warfarin use in Japanese patients using the Japanese therapeutic range showed an overall TTR of 69.7% in patients with non-valvular AF.23 Studies in Hong Kong showed that the mean TTR for target INR of 2.0 to 3.0 in patients with AF improved from 24.2% to 39.7% in the past decade.24 25 Our study showed better warfarin control in patients with AF (mean TTR=48.0%), compared with past local data; however, the rate of control remains unsatisfactory. A prior retrospective study demonstrated a mean TTR of 72.5% in Swedish patients with mechanical heart valve prosthesis; another study showed that the mean TTR was 47.48% in Malaysian patients with mechanical heart valve(s) replacement.26 27 The mean TTR in patients with PHV in this study was 30.5%, which was lower than the previously reported rate. Our study also demonstrated that warfarin control was worse in patients with PHV than in patients with AF.

The lower TTR in Hong Kong, compared with that in Western nations, could be attributed to ethnicity. Geographical differences in the genetic polymorphism profile between Hong Kong and Western nations could lead to differences in warfarin metabolism and warfarin dosing.28 Moreover, previous evidence suggests that individuals of East Asian ethnicity are more likely to experience intracranial haemorrhage, compared with individuals of Caucasian ethnicity (in that study, “white race/ethnicity”) who exhibit comparable levels of warfarin control.29 Notably, the possibility that physicians targeted a lower INR range in Hong Kong could not be ruled out in this study.

The overall predictive abilities of European and Japanese therapeutic ranges were similar. The calculated ORs for each economic outcome across European and Japanese therapeutic ranges were similar, with the exception of procedural and hospitalisation costs. For clinical outcomes, ORs could not be calculated to compare ideal TTR with non-ideal TTR, given that there were no complications in the ideal TTR group. However, there were complications in the group with ideal TTR based on the Japanese therapeutic range. The ORs calculated showed that the Japanese therapeutic range could be used to predict clinical outcomes. Notably, a lower INR target can be established in Hong Kong. However, a larger, well-designed randomised controlled trial is needed to establish non-inferiority in terms of clinical outcomes, as well as superiority in terms of economic outcomes, when using the Japanese therapeutic range.

The level of warfarin control has been associated with clinical outcomes. A systematic review of 47 studies revealed that TTR was negatively correlated with major bleeding and thromboembolic events.30 Our results were consistent in demonstrating an association of TTR with clinical outcome, which indicated that patients with worse TTR were more likely to experience overall complications, thrombotic events, and bleeding complications. Moreover, TTR has been associated with economic outcomes. A previous study in the US showed that patients with AF whose TTR was <60% had higher total healthcare and stroke-related costs.31 Our study demonstrated similar results, using a TTR cut-off of 70%. With better warfarin control, corresponding healthcare expenses can be reduced; many such expenses are borne by the government.

Predictors for suboptimal TTR have been investigated in previous studies. Notably, heart failure has been highly associated with poor warfarin control8 23; however, this association was not supported by our findings. In contrast, our study showed that younger patients were more likely to have poor TTR. This association might be related to improved medication adherence in older patients, because of better health consciousness among those individuals.32 33 Concurrent use of furosemide, famotidine, or simvastatin (in combination with warfarin) was associated with poor TTR. Despite common concurrent use of simvastatin and warfarin, the anticoagulant effect of warfarin is reportedly 8% to 15% stronger in simvastatin-treated patients, due to the CYP 2C9*3 polymorphism.34 Regarding concurrent use of warfarin and pantoprazole, altered warfarin absorption and metabolism have been observed during in vitro studies of proton pump inhibitor treatment35; however, there is a lack of supporting clinical evidence.36 Our study showed a tendency for enhanced likelihood of poor TTR control in patients with concurrent use of pantoprazole, although this association was not statistically significant. Thus, the influence of proton pump inhibitor use on warfarin control remains unclear. Patients with concurrent use of aspirin and warfarin exhibited a tendency for enhanced risk of poor TTR; this association was also not statistically significant. We noted a considerable reduction in the number of patients in the fourth TTR quartile (14.1%), compared with the other three groups (range, 33.8-49.2%). Concurrent use of aspirin and warfarin is known to enhance the risk of major bleeding, which could cause physicians to approach anticoagulation control more conservatively.37 Furthermore, the use of aspirin and poor TTR have both been independently associated with higher bleeding risk, while poor TTR has been regarded as an independent contributor to all-cause mortality.38 Therefore, regardless of the concurrent use of aspirin, optimal TTR should be achieved with regard to the appropriate INR therapeutic range to reduce complications in patients receiving warfarin therapy.

According to validation studies performed by Zeolla et al,20 the mean OAK score among long-term warfarin users was 72%. A study in Malaysia revealed that only 11.2% of patients achieved a satisfactory score, with a mean OAK score of 48% for the cohort.39 Similar results were achieved in our study; the mean score was 54.1% and 13.8% of patients achieved a score of ≥75%. Poor OAK score could be attributed to restricted medical consultation time, leading to a lack of knowledge concerning respective diseases and medications.40 Patients with older age were more likely to have low OAK scores, which was consistent with the findings of a previous study that demonstrated a negative correlation between age and warfarin knowledge.41 Nonetheless, the observed relationships of co-morbidities with warfarin knowledge require further analyses to establish underlying explanations.

This study had several important limitations. This was a single-centre study with limited sample size and study population distribution skewed towards AF patients concerning warfarin indications. The target INR range for included patients was unknown. Notably, some physicians may have set a lower goal of 1.5 to 2.5 in patients with higher risk of bleeding. Patient TTR could have been affected by medication delay or refusal due to medical procedures. The impacts of TTR on medication costs were not investigated because differences in available strengths of warfarin led to various combinations of warfarin prescriptions. Moreover, we could not adjust for diet, use of traditional Chinese medicine or complementary alternative medications, and medication non-compliance as factors that may influence warfarin control. The OAK test was amended to fit our local practices and was not administered to some of the recruited patients in this study. Further validation is needed concerning the Chinese version of the amended OAK test.

Warfarin use in Hong Kong patients was poorly controlled, regardless of indication. Patients with indications for AF had better warfarin control. Using the Japanese therapeutic range, the level of warfarin control remained unsatisfactory. Our study showed that TTR could be a predictor for both economic and clinical outcomes. Younger age was found to be an independent predictor of poor warfarin control, as were concurrent use of aspirin or simvastatin.

Patients had poor knowledge concerning INR value and interpretation. More education is needed regarding drug-drug interactions of warfarin and consequences of missed doses.

Concept or design: VWY Lee, BPY Yan.
Acquisition of data: IMH Lee, SKS Mak.
Analysis or interpretation of data: ASM Lam, VWY Lee, BPY Yan.
Drafting of the manuscript: ASM Lam, VWY Lee, BPY Yan.
Critical revision for important intellectual content: All authors.

All authors had full access to the data, contributed to the study, approved the final version for publication, and take responsibility for its accuracy and integrity.

As an editor of the journal, BPY Yan was not involved in the peer review process. Other authors have disclosed no conflicts of interest.

This manuscript was posted on Research Square as a registered online preprint (https://doi.org/10.21203/rs.2.15276/v1).

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

The study was approved by the Joint Chinese University of Hong Kong–New Territories East Cluster Clinical Research Ethics Committee (Ref CRE 2013.667). Informed verbal consent was obtained from patients participating in knowledge assessment, which was conducted via phone interview. The need for patient consent was waived by the Ethics Committee for the retrospective cohort study because no personal identifiers or related information were obtained during the data collection process.

1. Jabre P, Roger VL, Murad MH, et al. Mortality associated with atrial fibrillation in patients with myocardial infarction a systematic review and meta-analysis. Circulation 2011;123:1587-93. Crossref

2. Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an independent risk factor for stroke: the Framingham Study. Stroke 1991;22:983-8. Crossref

3. Cáceres-Lóriga FM, Pérez-López H, Santos-Gracia J, Morlans-Hernandez K. Prosthetic heart valve thrombosis: pathogenesis, diagnosis and management. Int J Cardiol 2006;110:1-6. Crossref

4. Aguilar MI, Hart R. Oral anticoagulants for preventing stroke in patients with non-valvular atrial fibrillation and no previous history of stroke or transient ischemic attacks. Cochrane Database Syst Rev 2005;(3):CD001927. Crossref

5. Cannegieter SC, Rosendaal FR, Briët E. Thromboembolic and bleeding complications in patients with mechanical heart valve prostheses. Circulation 1994;89:635-41. Crossref

6. Kirchhof P, Benussi S, Kotecha D, et al. 2016 ESC Guidelines for the management of atrial fibrillation developed in collaboration with EACTS. Eur Heart J 2016;37:2893-962. Crossref

7. Asarcıklı LD, Şen T, İpek EG, et al. Time in Therapeutic Range (TTR) value of patients who use warfarin and factors which influence TTR. J Am Coll Cardiol 2013;62:C127-8. Crossref

8. Nelson WW, Choi JC, Vanderpoel J, et al. Impact of co-morbidities and patient characteristics on international normalized ratio control over time in patients with nonvalvular atrial fibrillation. Am J Cardiol 2013;112:509- 12. Crossref

9. Cancino RS, Hylek EM, Reisman JI, Rose AJ. Comparing patient-level and site-level anticoagulation control as predictors of adverse events. Thromb Res 2014;133:652-6. Crossref

10. JCS Joint Working Group. Guidelines for Pharmacotherapy of Atrial Fibrillation (JCS 2013). Circ J 2014;78:1997-2021. Crossref

11. Yasaka M, Minematsu K, Yamaguchi T. Optimal intensity of international normalized ratio in warfarin therapy for secondary prevention of stroke in patients with non-valvular atrial fibrillation. Intern Med 2001;40:1183-8. Crossref

12.Holbrook AM, Pereira JA, Labiris R, et al. Systematic overview of warfarin and its drug and food interactions. Arch Intern Med 2005;165:1095-106. Crossref

13.Leite PM, Martins MAP, Castilho RO. Review on mechanisms and interactions in concomitant use of herbs and warfarin therapy. Biomed Pharmacother 2016;83:14- 21. Crossref

14. Wofford JL, Wells MD, Singh S. Best strategies for patient education about anticoagulation with warfarin: a systematic review. BMC Health Serv Res 2008;8:40. Crossref

15. Kagansky N, Knobler H, Rimon E, Ozer Z, Levy S. Safety of anticoagulation therapy in well-informed older patients. Arch Intern Med 2004;164:2044-50. Crossref

16. Zhou Z, Hu D. An epidemiological study on the prevalence of atrial fibrillation in the Chinese population of mainland China. J Epidemiol 2008;18:209-16. Crossref

17. Schmitt L, Speckman J, Ansell J. Quality assessment of anticoagulation dose management: comparative evaluation of measures of time-in-therapeutic range. J Thromb Thrombolysis 2003;15:213-6. Crossref

18. Schulman S, Kearon C, Subcommittee on Control of Anticoagulation of the Scientific and Standardization Committee of the International Society on Thrombosis and Haemostasis. Definition of major bleeding in clinical investigations of antihemostatic medicinal products in non-surgical patients. J Thromb Haemost 2005;3:692-4. Crossref

19.Government Logistics Department, Hong Kong SAR Government. Hospital Authority Ordinance (Chapter 113). Revision to list of public charges. Available from: http://www.gld.gov.hk/egazette/pdf/20172124/ egn201721243884.pdf. Accessed 6 Dec 2017.

20. Zeolla MM, Brodeur MR, Dominelli A, Haines ST, Allie ND. Development and validation of an instrument to determine patient knowledge: the Oral Anticoagulation Knowledge Test. Ann Pharmacother 2006;40:633-8. Crossref

21.Rahmani P, Guzman CL, Blostein MD, Tabah A, Muladzanov A, Kahn SR. Patients’ knowledge of anticoagulation and its association with clinical characteristics, INR Control and warfarin-related adverse events. Blood 2013;122:1738. Crossref

22. Erkens PM, ten Cate H, Büller HR, Prins MH. Benchmark for time in therapeutic range in venous thromboembolism: a systematic review and meta-analysis. PLoS One 2012;7:e42269. Crossref

23. Tomita H, Kadokami T, Momii H, et al. Patient factors against stable control of warfarin therapy for Japanese non-valvular atrial fibrillation patients. Thromb Res 2013;132:537-42. Crossref

24. Leung CS, Tam KM. Antithrombotic treatment of atrial fibrillation in a regional hospital in Hong Kong. Hong Kong Med J 2003;99:179-85.

25. Li WH, Huang D, Chiang CE, et al. Efficacy and safety of dabigatran, rivaroxaban, and warfarin for stroke prevention in Chinese patients with atrial fibrillation: the Hong Kong Atrial Fibrillation Project. Clin Cardiol 2017;40:222-9. Crossref

26. Grzymala-Lubanski B, Svensson PJ, Renlund H, Jeppsson A, Själander A. Warfarin treatment quality and prognosis in patients with mechanical heart valve prosthesis. Heart 2017;103:198-203. Crossref

27. Tan CS, Fong AY, Jong YH, Ong TK. INR control of patients with mechanical heart valve on long-term warfarin therapy. Glob Heart 2018;13:241-4. Crossref

28. Gaikwad T, Ghosh K, Shetty S. VKORC1 and CYP2C9 genotype distribution in Asian countries. Thromb Res 2014;134:537-44. Crossref

29. Shen AY, Yao JF, Brar SS, Jorgensen MB, Chen W. Racial/ ethnic differences in the risk of intracranial hemorrhage among patients with atrial fibrillation. J Am Coll Cardiol 2007;50:309-15. Crossref

30. Wan Y, Heneghan C, Perera R, et al. Anticoagulation control and prediction of adverse events in patients with atrial fibrillation: a systematic review. Circ Cardiovasc Qual Outcomes 2008;1:84-91. Crossref

31. Deitelzweig S, Evans M, Hillson E, et al. Warfarin time in therapeutic range and its impact on healthcare resource utilization and costs among patients with nonvalvular atrial fibrillation. Curr Med Res Opin 2016;32:87-94. Crossref

32. Skeppholm M, Friberg L. Adherence to warfarin treatment among patients with atrial fibrillation. Clin Res Cardiol 2014;103:998-1005. Crossref

33. Kang CD, Tsang PP, Li WT, et al. Determinants of medication adherence and blood pressure control among hypertensive patients in Hong Kong: a cross-sectional study. Int J Cardiol 2015;182:250-7. Crossref

34. Andersson ML, Mannheimer B, Lindh JD. The effect of simvastatin on warfarin anticoagulation: a Swedish register-based nationwide cohort study. Eur J Clin Pharmacol 2019;75:1387-92. Crossref

35. Li XQ, Andersson TB, Ahlström M, Weidolf L. Comparison of inhibitory effects of the proton pump-inhibiting drugs omeprazole, esomeprazole, lansoprazole, pantoprazole, and rabeprazole on human cytochrome P450 activities. Drug Metab Dispos 2004;32:821-7. Crossref

36. Henriksen DP, Stage TB, Hansen MR, Rasmussen L, Damkier P, Pottegård A. The potential drug-drug interaction between proton pump inhibitors and warfarin. Pharmacoepidemiol Drug Saf 2015;24:1337-40. Crossref

37. Dans AL, Connolly SJ, Wallentin L, et al. Concomitant use of antiplatelet therapy with dabigatran or warfarin in the Randomized Evaluation of Long-Term Anticoagulation Therapy (RE-LY) trial. Circulation 2013;127:634-40. Crossref

38. Proietti M, Lip GY. Impact of quality of anticoagulation control on outcomes in patients with atrial fibrillation taking aspirin: an analysis from the SPORTIF trials. Int J Cardiol 2018;252:96-100. Crossref

39. Matalqah LM, Radaideh K, Sulaiman SA, Hassali MA, Kader MA. An instrument to measure anticoagulation knowledge among Malaysian community: a translation and validation study of the Oral Anticoagulation Knowledge (OAK) Test. Asian J Biomed Pharm Sci 2013;3:30-7.

40. Lee VW, Tam CS, Yan BP, Yu CM, Lam YY. Barriers to warfarin use for stroke prevention in patients with atrial fibrillation in Hong Kong. Clin Cardiol 2013;36:166-71. Crossref

41. Hasan SS, Shamala R, Syed IA, et al. Factors affecting warfarin-related knowledge and INR control of patients attending physician- and pharmacist-managed anticoagulation clinics. J Pharm Pract 2011;24:485-93. Crossref