Enhanced recovery after surgery (ERAS) practices were developed in the 1990s whereby multiple modalities of intervention1 were introduced perioperatively to improve postoperative recovery,2 reduce length of stay (LOS),3 and lower the incidence of perioperative morbidity.4 These practices have been widely adopted in many surgical fields,5 6 7 including orthopaedics.8 Further enhancements of postoperative pain management, venous thromboembolism prophylaxis, and early mobilisation have led to encouraging results in primary total hip arthroplasty (THA) and total knee arthroplasty (TKA); such results have included earlier recovery,9 LOS reduction,10 improved function,11 and lower venous thromboembolism incidence12 without declines in patient satisfaction, postoperative complication rate,13 or cost.14 The development of ERAS practices has matured with the progressive introduction of standardised clinical pathways for all patients receiving THA or TKA,15 also referred as fast-track hip and knee arthroplasty.16 These ERAS practices have become the standard of care in most joint replacement centres.17 18

Because of the ageing population and increasing incidence of degenerative joint disease,19 the wait time for elective TKA in a public joint replacement service can reach 5 years in Hong Kong20; thus, many patients visit private orthopaedic surgeons for earlier surgery. Despite the presence of robust joint replacement options in the Hong Kong orthopaedic community, there remain differences between public and private hospital settings in terms of the environment, perioperative medical care, and service availability. To our knowledge, no studies have been published regarding the effects of ERAS practices on LOS after lower limb total joint arthroplasty, or whether ERAS practices could be implemented as a standardised protocol by private hospitals in Hong Kong.

Therefore, this study investigated whether the promising results of ERAS practices could be reproduced in private hospitals by comparison of LOS among patients who underwent unilateral primary THA or TKA by a single surgeon at Hong Kong Sanatorium & Hospital, a private hospital in Hong Kong, before and after the implementation of ERAS.

Patients who had undergone unilateral primary THA or TKA by the senior author (KYC) at Hong Kong Sanatorium & Hospital, Hong Kong, were included in the study cohort. Patients with revision arthroplasty, one-stage bilateral arthroplasty, and unicompartmental knee arthroplasty were excluded. Because ERAS practices were progressively implemented from 2015 to 2016, we allocated patients who were treated from 2012 to 2014 into the conventional group and patients who were treated from 2017 to 2018 into the ERAS group.

Our ERAS practices were generally similar to conventional practices. Most patients underwent surgery on the morning after an evening admission. Most patients received spinal anaesthesia unless contra-indicated (eg, ankylosing spondylitis, severe spinal deformity, coagulopathy, or fixed cardiac output state); routine sedation (using intravenous midazolam and propofol) was also conducted to improve patient comfort. Cementless THA systems, either a Pinnacle acetabular cup with a Summit femoral stem (DePuy, Warsaw [IN], US) or an R3 acetabular cup with a Synergy femoral stem (Smith & Nephew, Auckland, New Zealand), were implemented by means of a posterolateral approach. Total knee arthroplasty was performed using a medial parapatellar approach with thigh tourniquet and conventional instruments. Cemented rotating platform TKA systems were used: either a Legacy Posterior Stabilised Flex Mobile prosthesis (Zimmer, Warsaw [IN], US) or an Attune prosthesis (DePuy). A Foley urinary catheter was inserted only on urinary retention with bladder volume of >800 mL.21 Prophylactic antibiotics were administered on the induction of anaesthesia, then continued for 2 days after surgery. Prophylaxis against venous thromboembolism, both pharmacological (with subcutaneous enoxaparin) and mechanical (with a sequential compression device), was routinely implemented. Patients were discharged home when they could safely exit their beds without assistance and stably walk using an assistive device without any sign of complications.

When using ERAS practices, a higher dose of intravenous steroid is administered on the induction of anaesthesia for both THA and TKA. In the conventional group, 4 to 8 mg of dexamethasone was administered; in the ERAS group, 125 mg of methylprednisolone (equivalent to 25 mg of dexamethasone) was administered instead.22 Notably, high-dose glucocorticoids before arthroplasty are reportedly safe and recommended for routine use.23

In the conventional group, no standard pain control regimen was established. Pain medications were prescribed at the discretion of anaesthetists or surgeons, including the use of femoral nerve block and postoperative patient-controlled analgesia pump. Pain management was standardised and optimised in the ERAS group, particularly for patients undergoing TKA. Pre-emptive analgesia was implemented, such that patients routinely began oral pregabalin and transdermal buprenorphine patch treatments before surgery. Preventive analgesia was also employed both intra- and post-operatively. A periarticular “cocktail” injection of local infiltrative analgesia24—consisting of ropivacaine, ketorolac, and 1:1000 adrenaline—was injected into the posterior joint capsule before implantation of the prosthesis; it was also injected into the subcutaneous layer anteriorly and intra-articularly during wound closure. After surgery, patients received multimodal oral analgesia including regular cyclooxygenase-2 inhibitors or non-steroidal anti-inflammatory drugs, pregabalin, and paracetamol. Buprenorphine patch treatment was maintained for 5 to 7 days. Patient-controlled analgesia was omitted when using ERAS practices. In contrast, the pain control requirement was lower for patients undergoing THA. In both conventional and ERAS groups, local anaesthetic (bupivacaine) was injected into the subcutaneous plane before skin closure, while oral paracetamol was prescribed after surgery. After discharge from the hospital, patients who underwent TKA were administered non-steroidal anti-inflammatory drugs, pregabalin, and paracetamol for up to 5 weeks after surgery; most patients undergoing THA were prescribed paracetamol alone.

Prophylactic intravenous palonosetron was routinely administered to prevent postoperative nausea and vomiting; intravenous metoclopramide was used to manage breakthrough symptoms.

Tranexamic acid was routinely used in the ERAS group to minimise bleeding and the need for transfusion. For patients undergoing TKA, 1 g of tranexamic acid was injected intra-articularly after deep layer closure. No routine use of tranexamic acid was adopted in conventional practices. A deep drain was used when adhering to conventional practices but not when adhering to ERAS practices. For patients undergoing THA, intravenous tranexamic acid was administered at the same time as induction of anaesthesia; a deep drain was also used and removed the next morning.

While hypnotics were only administered on request when in the conventional group, patients in the ERAS group were routinely prescribed hypnotics the night before surgery and the first 2 to 3 days after surgery. This helped patients in the ERAS group to comply with the rehabilitation programme after surgery.

Same-day or day-zero rehabilitation was implemented in the ERAS group. Patients in the conventional group had bed rest on the day of surgery, then began mobilisation on postoperative day 1. Conversely, patients in the ERAS group who underwent morning surgery were encouraged to mobilise in the afternoon or evening on the same day, under physiotherapist supervision.

The outcome was postoperative LOS, which was denoted by the number of days after surgery when the patient was discharged from the hospital. The day of surgery was regarded as postoperative day 0. Discharge criteria remained consistent throughout the study period (ie, safe exit from bed without assistance and stable walking using an assistive device), as described above. The proportion of patients discharged on or before postoperative day 3 in each group was compared. We also investigated the effects of age, sex, nature of surgery (THA versus TKA), and class of hospital bed (general versus private ward) on LOS.

Patient data were anonymously entered into an encrypted file to ensure privacy. Data analysis was performed using SPSS (Window version 26.0; IBM Corp, Armonk [NY], US). The Chi squared test, independent samples t test with two-tailed significance, and one-way analysis of variance were used for comparisons. A P value of <0.05 was considered statistically significant.

In total, 228 patients were included: 117 in the conventional group and 111 in the ERAS group. The mean and median ages did not significantly differ between the conventional and ERAS groups (Table 1). Most patients were aged between 50 and 89 years (89.7% in the conventional group and 97.2% in the ERAS group); however, the distribution of ages did not significantly differ between groups. The distributions of sex, nature of surgery, and class of hospital bed also did not significantly differ between the conventional and ERAS groups (Table 1).

The mean LOS significantly improved from 5.16 ± 2.06 days to 3.28 ± 1.04 days (P<0.001) after ERAS implementation (Table 2). Patients discharged on or before postoperative day 3 comprised 13.7% of the conventional group and 77.5% of the ERAS group (P<0.001).

Subgroup analysis was performed to examine the effects of sex, nature of surgery, class of hospital bed (Table 3), and age-group (Fig) on LOS.

In the conventional group, there were no significant differences in mean LOS between female and male patients, patients receiving TKA and patients receiving THA, or general ward and private ward patients (Table 3). One-way analysis of variance showed a significant difference in mean LOS among age-groups (F [7, 109]=2.58, P=0.017) [Fig]. The mean LOS generally increased as age increased from the third decade (3 days) to the ninth decade (7 days); however, two patients in the 20-29 age-group had exceptionally long hospital stays.

In the ERAS group, there were no significant differences in mean LOS between female and male patients or between patients receiving TKA and patients receiving THA (Table 3). One-way analysis of variance showed that age did not have a significant effect on the mean LOS (F [5, 105]=1.13, P=0.348) [Fig]. However, the mean LOS significantly differed between general ward and private ward patients (3.06 ± 0.59 vs 3.66 ± 1.46 days, P=0.003) [Table 3].

No postoperative complications or instances of 30-day re-admission were observed among patients who underwent TKA. Among patients who underwent THA, three (two from the conventional group, one from the ERAS group) had complications. In the conventional group, one patient with spondyloepiphyseal dysplasia experienced dislocation during in-patient stay, which required closed reduction; one patient experienced dislocation during postoperative week 3, which required re-admission and revision to offset the liner and a longer neck hip ball to improve soft tissue tension. In the ERAS group, one patient had periprosthetic femoral fracture after an accidental fall on postoperative day 13, which required re-admission with revision to the long cementless stem, as well as cable fixation. No patients in either group experienced postoperative wounds or periprosthetic infections.

Despite more efficient service provision, postoperative LOS in private hospitals might be limited by confounders that surgeons cannot control (eg, patient preference and financial factors).25 Nevertheless, it was unsurprising that our results were consistent with previous literature: ERAS practices are effective for reducing the LOS after unilateral primary arthroplasty.

Regarding factors that affect postoperative LOS, a significant difference in the mean LOS was observed between general ward and private ward patients in the ERAS group. In public hospitals, the LOS among patients with worse socio-economic backgrounds is often limited by inadequate social support from family after discharge26 or a suboptimal home environment (eg, non-lift landing flats in older urban buildings).27 While placement issues are rarely problematic for patients in private hospitals,28 a possible explanation for the difference in LOS between general ward and private ward patients, where the cost difference is on average 5 times higher, is that patients with better socio-economic backgrounds may have higher expectations for surgical outcomes29; thus, they may tolerate longer hospital stays for rehabilitation, despite the higher costs of such stays. Private insurance is also reportedly an independent predictor of discharge delay despite objective readiness for discharge30; however, we presumed that the effect of insurance was not applicable in the present study because fewer than 10% of patients in our cohort had no insurance coverage. Furthermore, no significant differences in the mean LOS were noted with regard to the nature of surgery, sex, or age in the ERAS group. These findings may be related to the use of standardised ERAS practices and perioperative protocols, which have minimised variation in patient management.31

The implementation of ERAS practices in private hospitals is potentially beneficial to all stakeholders (including hospital administrators) because it facilitates hospital bed availability, while reducing costs via shorter convalescence duration and reduced morbidity.32 However, there are some important challenges for surgeons who wish to initiate ERAS practices in private centres. These challenges include occasional requirements for minor alterations in ward environments, changes in anaesthesia technique, rapid turnover of in-house surgical staff, and noncompliance with ERAS practices.33 Furthermore, a large caseload might be necessary to attract a dedicated multidisciplinary team for the sustainable development of ERAS practices in private centres. While it may be challenging to achieve full ERAS implementation in a short period of time, the stepwise addition of ERAS components might improve patient outcomes in private hospitals.34

There were some limitations in this study. First, this study used a retrospective design without randomisation, which may have led to imbalance and bias in the results. Second, this study only involved patients from a single surgeon; thus, the sample size was small. Third, differences in functional status and co-morbidities were not considered in the analysis, as the electronic health record sharing system between public and private hospitals was only established in 2016 so complete acquisition of patient’s parameters was not possible. Finally, other clinical outcome parameters and patient satisfaction were not investigated; these will be examined in a future study, where a thorough documentation in patient reported outcome measure and clinician-based outcome measure will improve the validity of results.

In conclusion, ERAS practices produced significant improvement in mean postoperative LOS, compared to conventional practices, for patients who underwent unilateral primary THA or TKA in a private hospital. Specifically, a significantly greater proportion of patients in the ERAS group were able to return home on or before postoperative day 3. The findings indicate that the good outcomes of ERAS practices in public joint replacement centres can be reproduced in private hospitals with sufficient caseloads and consistent implementation of ERAS practices.

Concept or design: KY Chiu.
Acquisition of data: MMT Chung.
Analysis or interpretation of data: MMT Chung.
Drafting of the manuscript: MMT Chung, JKF Ng, FY Ng, PK Chan.
Critical revision of the manuscript for important intellectual content: KY Chiu.

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 results of this study were presented in the Hong Kong Orthopaedic Association 40th Annual Congress in Hong Kong (31 October to 1 November 2020).

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

This study was approved by the Hong Kong Sanatorium & Hospital Medical Group Research Committee (Ref RC-2019- 25). The requirement for patient consent was waived for this retrospective study.

1. Kehlet H. Multimodal approach to postoperative recovery. Curr Opin Crit Care 2009;15:355-8. Crossref

2. Kehlet H, Dahl JB. Anaesthesia, surgery, and challenges in postoperative recovery. Lancet 2003;362:1921-8. Crossref

3. Kehlet H. Fast-track hip and knee arthroplasty. Lancet 2013;381:1600-2. Crossref

4. Rogers LJ, Bleetman D, Messenger DE, et al. The impact of enhanced recovery after surgery (ERAS) protocol compliance on morbidity from resection for primary lung cancer. J Thoracic Cardiovasc Surg 2018;155:1843-52. Crossref

5. Geltzeiler CB, Rotramel A, Wilson C, Deng L, Whiteford MH, Frankhouse J. Prospective study of colorectal enhanced recovery after surgery in a community hospital. JAMA Surg 2014;149:955-61. Crossref

6. Melnyk M, Casey RG, Black P, Koupparis AJ. Enhanced recovery after surgery (ERAS) protocols: time to change practice? Can Urol Assoc J 2011;5:342-8. Crossref

7. Scheib SA, Thomassee M, Kenner JL. Enhanced recovery after surgery in gynecology: a review of the literature. J Minim Invasive Gynecol 2019;26:327-43. Crossref

8. Andersen LØ, Gaarn-Larsen L, Kristensen BB, Husted H, Otte KS, Kehlet H. Subacute pain and function after fast-track hip and knee arthroplasty. Anaesthesia 2009;64:508-13. Crossref

9. Wu CL, Raja SN. Treatment of acute postoperative pain. Lancet 2011;377:2215-25. Crossref

10. Auyong DB, Allen CJ, Pahang JA, Clabeaux JJ, MacDonald KM, Hanson NA. Reduced length of hospitalization in primary total knee arthroplasty patients using an updated enhanced recovery after surgery (ERAS) pathway. J Arthroplasty 2015;30:1705-9. Crossref

11. Pua YH, Ong PH. Association of early ambulation with length of stay and costs in total knee arthroplasty: retrospective cohort study. Am J Phys Med Rehabil 2014;93:962-70. Crossref

12. Pearse EO, Caldwell BF, Lockwood RJ, Hollard J. Early mobilisation after conventional knee replacement may reduce the risk of postoperative venous thromboembolism. J Bone Joint Surg Br 2007;89:316-22. Crossref

13. Zhu S, Qian W, Jiang C, Ye C, Chen X. Enhanced recovery after surgery for hip mand knee arthroplasty: a systematic review and meta-analysis. Postgrad Med J 2017;93:736-42. Crossref

14. Duncan CM, Hall Long K, Warner DO, Hebl JR. The economic implications of a multimodal analgesic regimen for patients undergoing major orthopaedic surgery: a comparative study of direct costs. Reg Anesth Pain Med 2009;34:301-7. Crossref

15. Duggal S, Flics S, Cornell CN. Introduction of clinical pathways in orthopedic surgical care: the experience of the hospital for special surgery. In: Ronald MacKenzie C, Cornell CN, Memtsoudis SG, editors. Perioperative Care of the Orthopedic Patient. New York: Springer; 2014: 365-71. Crossref

16. Husted H. Fast-track hip and knee arthroplasty: clinical and organizational aspects. Acta Orthop Suppl 2012;83:1-39. Crossref

17. Christelis N, Wallace S, Sage CE, et al. An enhanced recovery after surgery program for hip and knee arthroplasty. Med J Aust 2015;202:363-8. Crossref

18. Soffin EM, YaDeau JT. Enhanced recovery after surgery for primary hip and knee arthroplasty: a review of the evidence. Br J Anaesth 2016;117(Suppl 3):iii62-72. Crossref

19. Yan CH, Chiu KY, Ng FY. Total knee arthroplasty for primary knee osteoarthritis: changing pattern over the past 10 years. Hong Kong Med J 2011;17:20-5.

20. Hospital Authority, Hong Kong SAR Government. Elective total joint replacement surgery. 2019. Available from: https://www.ha.org.hk/visitor/ha_visitor_index.asp?Content_ID=221223&Lang=EN%20G&Dimension=100&Parent_ID=214172&Ver=HTML. Accessed 25 Feb 2020. Crossref

21. Bjerregaard LS, Hornum U, Troldborg C, Bogoe S, Bagi P, Kehlet H. Postoperative urinary catheterization thresholds of 500 versus 800 ml after fast-track total hip and knee arthroplasty: a randomized, open-label, controlled trial. Anesthesiology 2016;124:1256-64. Crossref

22. Lunn TH, Kristensen BB, Andersen LØ, et al. Effect of high-dose preoperative methylprednisolone on pain and recovery after total knee arthroplasty: a randomized, placebo-controlled trial. Br J Anaesth 2011;106:230-8. Crossref

23. Kehlet H, Lindberg-Larsen V. High-dose glucocorticoid before hip and knee arthroplasty: to use or not to use-that's the question. Acta Orthop 2018;89:477-9. Crossref

24. Ng FY, Ng JK, Chiu KY, Yan CH, Chan CW. Multimodal periarticular injection vs continuous femoral nerve block after total knee arthroplasty: a prospective, crossover, randomized clinical trial. J Arthroplasty 2012;27:1234-8. Crossref

25. Badham J, Brandrup J. Length of stay comparisons for private and public hospitals. Aust Health Rev 2000;23:162-70. Crossref

26. Freitas A, Silva-Costa T, Lopes F, et al. Factors influencing hospital high length of stay outliers. BMC Health Serv Res 2012;12:265. Crossref

27. Waring J, Marshall F, Bishop S, et al. An ethnographic study of knowledge sharing across the boundaries between care processes, services and organisations: the contributions to ‘safe’ hospital discharge. Southampton (UK): NIHR Journals Library; Sep 2014. Crossref

28. Perelman J, Closon MC. Impact of socioeconomic factors on in-patient length of stay and their consequences in per case hospital payment systems. J Health Serv Res Policy 2011;16:197-202. Crossref

29. Li X, Galvin JW, Li C, Agrawal R, Curry EJ. The impact of socioeconomic status on outcomes in orthopaedic surgery. J Bone Joint Surg Am 2020;102:428-44. Crossref

30. Celio DA, Poggi R, Schmalzbauer M, Rosso R, Majno P, Christoforidis D. ERAS, length of stay and private insurance: a retrospective study. Int J Colorectal Dis 2019;34:1865-70. Crossref

31. Barbieri A, Vanhaecht K, Van Herck P, et al. Effects of clinical pathways in the joint replacement: a meta-analysis. BMC Med 2009;7:32. Crossref

32. Stowers MD, Lemanu DP, Hill AG. Health economics in enhanced recovery after surgery programs. Can J Anaesth 2015;62:219-30. Crossref

33. Kahokehr A, Sammour T, Zargar-Shoshtari K, Thompson L, Hill AG. Implementation of ERAS and how to overcome the barriers. Int J Surg 2009;7:16-9. Crossref

34. Tan NL, Hunt JL, Gwini SM. Does implementation of an enhanced recovery after surgery program for hip replacement improve quality of recovery in an Australian private hospital: a quality improvement study. BMC Anesthesiol 2018;18:64. Crossref

Labour is regarded as a time of suffering in a woman’s life, during which she may experience intensive pain that lasts for many hours. Ineffective labour pain management could create a negative life experience for a woman, which may negatively impact postpartum sexual and marital satisfaction.1 2 Labour pain involves both physical and psychological elements such as uterine contractions, tension, fear, anxiety, and the sensations of powerlessness and a loss of control.3 Current remedies for labour pain include pharmacological and non-pharmacological interventions. The most common pharmacological interventions include nitrous oxide inhalation, the injection of narcotic analgesics (eg, pethidine), and epidural analgesia. However, these methods are associated with adverse effects such as nausea and vomiting, longer first and second stages of labour, hypotension, motor blockade, fever, and urinary retention; they can also lead to neonatal respiratory depression and newborn sleepiness that affects breastfeeding.4 5 6 7 8 9 Hence, women prefer safer and simpler non-pharmacological pain relief methods.10 11

A notable non-pharmacological remedy is massage, which may provide pain relief to the site of application, along with overall psychological relaxation.12 The pressure applied during massage is presumed to block the transmission of pain impulses to the brain, while stimulating local release of endorphins.13 Randomised controlled trials concerning intrapartum massage have been conducted in various countries over the past two decades.12 14 15 16 17 18 19 20 21 22 23 However, there have been conflicting findings concerning beneficial effects (ie, reductions in pain score or the use of pharmacological analgesia)12 14 15 16 17 21 because of small sample sizes which ranged from 28 women12 to 176 women.21 Furthermore, the duration of intrapartum massage was either unspecified15 21 or lasted for only 30 to 40 minutes.12 14 18 19 22 23 In addition, intrapartum massage was performed by various types of people: student midwives,22 therapists17 19 or partners who had received training by therapists immediately before labour.12 14 These factors probably influenced the effectiveness, consistency, and duration of the application of massage. A recent Cochrane review concluded that the current quality of evidence regarding intrapartum massage is low to very low.24 Therefore, this randomised controlled study investigated the efficacy of a comprehensive massage programme, combined with controlled breathing and visualisation—all initiated during the antenatal period—as a non-pharmacological pain relief method during labour, with the goal of reducing pethidine or epidural analgesia use.

This randomised controlled study was conducted in two public hospitals in Hong Kong, where the midwives were responsible for intrapartum management and natural vaginal delivery of low-risk pregnancies. The respective annual childbirth rates were approximately 5000 and 7000; the caesarean section rates were 21% and 23%.25 Recruitment commenced in September 2016 and completed in December 2017. The recruitment of women was conducted at 32 to 36 weeks of gestation during their routine antenatal visit by a team of research midwives. The inclusion criteria were low-risk nulliparous Chinese women aged ≥18 years, who could communicate in Cantonese, and who carried a singleton pregnancy without known contraindications for vaginal delivery. Exclusion criteria were the use of massage among women in the control group, the absence of a partner to learn the massage technique, planned delivery in hospitals other than the study sites, and planned caesarean delivery. There was no exclusion of recruited women who attempted vaginal delivery or induction of labour but eventually required intrapartum caesarean delivery.

Randomisation was conducted via two-by-two blocking with a block size of 4; a computer-generated number indicating either the study or control group was sealed in an opaque envelope. After a woman had provided written informed consent to participate, the midwife revealed the group allocation by opening the envelope. Because there were multiple midwifery staff responsible for participant recruitment at different occasions, none of the staff were aware of the allocation of previous participants; hence, they were unable to guess the group allocation.

Couples (ie, participating women and their partners) randomised to the massage group were invited to attend a 2-hour childbirth massage programme class at 36 weeks of gestation. This programme was based on the United Kingdom’s Royal College of Midwives accredited course ‘Towards Natural Childbirth and Beyond’.26 It included a 30-minute theoretical explanation of the evidence underpinning the childbirth massage programme, followed by a 90-minute practicum. During the 90-minute practicum, the couples received training by accredited midwifery trainers with respect to the massage technique, controlled breathing, and visualisation, in accordance with the methods used in previous studies.16 20 The massage areas included the lower back and four limbs. Couples were taught how to synchronise the massage strokes with slow rhythmic breathing. Visualisation (ie, a mind mapping component) was also taught.26 In this process, the woman was asked to imagine something comfortable, which could bring her to a relaxed state. Subsequently, the couples were encouraged to practise the massage technique regularly at home in the evening, in a dimly lit and quiet environment, with the aim of encouraging relaxation and improving the quality and duration of sleep.27 The control group received standard antenatal education without instruction concerning massage, controlled breathing, or visualisation techniques.

When a woman in the massage group was admitted to the study hospital at onset of labour or for planned labour induction, her partner was first asked to demonstrate massage technique to the research team midwives to ensure that the partner could perform the procedure properly. If labour was not yet established, each woman was encouraged to relax through self-massage on her abdomen and legs. When labour commenced, the partner stayed to provide arm and shoulder massage for relaxation or lateral sacral massage for pain relief, according to the woman’s preference. There was no time limit for massage as long as the couple was happy and felt comfortable to continue the procedure throughout the labour. The partner could take a break in times of fatigue, or when the woman fell asleep. The partners of women in the control group were also encouraged to accompany the women during labour and delivery. Women in both groups otherwise received the same intrapartum care. They received explanations concerning the effectiveness of various analgesic methods according to the ranking of reported efficacy4 7: epidural was ranked highest, followed by pethidine, then nitrous oxide and other non-pharmacological analgesia methods (including transcutaneous nerve stimulation, birthing ball, and warm pads). Women could choose various methods or a combination of methods according to their pain tolerance and acceptance, using a step-up approach or direct implementation of the most effective methods. The degree of labour pain was assessed using the visual analogue scale for pain (ranging from 0 [no pain] to 10 [most painful]) at different stages of labour: latent phase (cervical dilatation of 1-3 cm), active phase (cervical dilatation of 4-7 cm), late active phase (cervical dilatation of 8-9 cm), and second stage (cervical dilatation of 10 cm); it was also assessed when the women first requested pethidine or epidural analgesia.

The primary outcome of this study was the use of the two most effective pharmacological methods (as described above): intramuscular pethidine injection or epidural analgesia. Women were also categorised according to the type of analgesia that they eventually received: none of the analgesic methods; non-pharmacological methods only; nitrous oxide ± non-pharmacological methods; pethidine ± other pain relief except epidural; or epidural ± above methods. The proportions of women that received each type of analgesia were also compared as one of the secondary outcomes. Other secondary outcomes included intrapartum caesarean rate, duration of labour, the pain score at the point when the women first requested pethidine or epidural analgesia, the interval between the onset of labour to the time of making such a request, and the cervical dilatation at which such a request was made.

A previous study reported a reduction of 60% in the epidural rate with the use of intrapartum massage when compared with the control group.2 Therefore, our study sample size was calculated based on the assumption that the requirement for pethidine injection or epidural analgesia could be reduced by 60% (ie, from the current 15% according to Hospital Authority data to 6%) in the study group. Using an 80% power (beta) threshold and a two-tailed alpha value of 5%, we calculated that 181 participants were required in each arm. The method of calculation was obtained from the website of Department of Obstetrics and Gynaecology, the Chinese University of Hong Kong (http://www.obg.cuhk.edu.hk/ResearchSupport/StatTools/index.php). Because we anticipated that 40% of the recruited participants would be excluded (eg, because of a shift to a private hospital for delivery, change to elective caesarean section, or withdrawal from the study), we planned to recruit 300 participants for each arm.

The Chi squared test and t test were used to assess differences in baseline characteristics, obstetric outcomes, neonatal outcomes, and the proportions of women using specific pharmacological pain relief methods between the massage and control groups. Linear-by-linear association was used to assess trends regarding the use of different types of analgesia. The t test was used to compare between-group differences in the stage of labour, cervical dilatation, and pain score among participants who used pethidine/epidural, as well as the mean pain scores in different phases of labour among participants who did not use any pain relief modalities. A P value of <0.05 was considered statistically significant. All analyses used a per-protocol approach. Intention-to-treat analysis (including all participants recruited at baseline) could not be conducted because information collected during labour (eg, the use of pain relief modalities) was not available for participants who delivered in other hospitals, required caesarean section before pain labour commenced, or withdrew from the study. Statistical analyses were performed using SPSS software (Windows version 22.0; IBM Corp, Armonk [NY], United States).

Of the 1130 women eligible for this study, 528 were excluded for reasons shown in the Figure; thus, 602 women were randomised to the massage group (n=302) and control group (n=300). Furthermore, 69 (22.8%) and 54 (18.0%) women were subsequently excluded from the massage and control groups, respectively, for reasons such as delivery in private hospitals, planned caesarean section, development of complications before labour, or withdrawal from the study. Finally, 479 pregnant women (233 in the massage group and 246 in the control group) were included in the per-protocol analysis.

There were no significant differences between groups in terms of maternal age, height, or demographic characteristics nor in the proportions of women who underwent induction of labour, augmentation of labour, or delivered by caesarean section (Table 1). The mean duration of labour did not differ between the massage and control groups. No significant differences were found in gestational age at delivery, birthweight, or the proportion of babies with Apgar score ≥8 at 5 minutes (Table 1). All women in the massage group practised massage during labour (n=233). The duration of massage ranged from 35 minutes to 195 minutes (median, 100 minutes).

The proportion of women who used pethidine or epidural did not significantly differ between the massage and control groups (12.0% vs 15.9%; P=0.226). However, linear-by-linear association analysis showed a significant shift in the massage group, from using stronger analgesics (eg, epidural analgesia: 2.1% in the massage group vs 5.7% in the control group) to weaker analgesics. Thus, more women in the massage group required none of the analgesics, compared with women in the control group (29.2% vs 21.5%; P=0.041) [Table 2].

Among women who needed pethidine or epidural for pain control, there was no difference between the two groups in terms of the pain score at the point when they requested these pain relief modalities, or the interval between the onset of labour to the time of requesting these modalities (Table 3). However, the cervical dilatation at which pethidine or epidural was first requested was significantly greater in the massage group (3.8 ± 1.7 cm) than in the control group (2.3 ± 1.0 cm; P<0.001) [Table 3].

Among women who needed none of the pain relief modalities, the pain scores progressively increased with cervical dilatation, although there were no differences between the two groups (Table 4).

Although our study did not show a statistically significant reduction in the number of women who used either pethidine or epidural analgesia with the practice of massage (12.0% vs 15.9%), linear-by-linear analysis revealed that there was a statistically significant overall shift in the pattern of analgesics use in the massage group: a smaller proportion of women requested epidural analgesia (2.1% vs 5.7%) and a larger proportion of women requested none of the pain relief methods (29.2% vs 21.5%). Our results suggest that the pain perceptions of labouring women were improved by the training and practice of massage, controlled breathing, and visualisation. Thus, some women who initially requested the stronger methods (eg, epidural analgesia) might have achieved satisfactory pain control with weaker analgesic methods (eg, pethidine or nitrous oxide). Similarly, women who initially requested pethidine or nitrous oxide might have shifted to non-pharmacological methods only; this led to a greater proportion of women in the massage group who requested no analgesia.

Although pain scores are commonly used to compare analgesic effectiveness, such comparisons are often difficult on the basis of a single pain score during labour. This is because the labour process is generally long and its characteristics are variable; labouring women might use more than one method of pain relief at different stages of labour. Nonetheless, if the pain is intolerable, labouring women require a stronger analgesic method.28 Hence, we used a pattern of analgesic utility (rather than a single pain score) as an indicator for the effectiveness of the massage programme; our linear-by-linear association findings indicated that the massage programme may reduce pain perception among labouring women. Furthermore, the mean cervical dilatation at the time of pethidine or epidural analgesia request was higher in the massage group than in the control group (3.8 ± 1.7 cm vs 2.3 ± 1.0 cm). Notably, among women who requested pethidine or epidural analgesia, the pain score at the point of first pethidine or epidural analgesia request was very similar between the massage and control groups (7.6 ± 2.2 vs 7.6 ± 2.8). Although the midwives were not blinded to the allocation in this study, women in both groups received the same intrapartum care and could choose pain relief methods according to their pain tolerance and acceptance. These results further support the notion that the practice of massage might have modulated the pain perception among labouring women, such that they only requested stronger pharmacological pain relief during later phases of labour; additional studies are required to confirm the underlying biological mechanism.

Janssen et al17 also reported a delay in epidural insertion by 1 cm of cervical dilatation (5.9 cm in the massage group vs 4.9 cm in the control group) in their randomised controlled trial. However, they failed to show a significant reduction in the rate of epidural analgesia use (81.1% in the massage group vs 65.0% in the control group). Importantly, their participants only learned and practised massage at the time of labour, while our participants began learning the massage programme during the antenatal period.

In another randomised controlled trial, Levett et al21 showed that the incidence of epidural analgesia was significantly reduced (from 68.2% to 23.9%) in a cohort of 176 Australian patients. However, their control group had a baseline epidural analgesia rate of 68.2%, which was much higher than the rate in our control group (ie, 5.7%, which is similar to the 6.6% reported previously in Hong Kong29) Possible reasons for the large difference in epidural rates between Australia and Hong Kong include variations in midwifery practices, pain tolerance among labouring women, and limited resources in Hong Kong public hospitals. Other obstetric practice differences include an overall (massage and control groups combined) higher normal vaginal delivery rate in our cohort than in the cohort reported by Levett et al21 (76.4% vs 57.9%); our overall cohort also exhibited a lower instrumental delivery rate (10.9% vs 17.0%) and a lower caesarean section rate (12.7% vs 25.1%). Regardless of our low baseline epidural rate, we found a 60% reduction in epidural use in the massage group (2.1%), compared with the control group (5.7%). Finally, Levett et al21 only reported the incidence of simultaneous use of pethidine and nitrous oxide, whereas we stratified analgesic methods according to the strength of pain relief; this allowed us to detect an overall shift towards weaker analgesics among women in the massage group.

Importantly, neither study (this study or the study by Levett et al21) demonstrated any reduction in the overall duration of labour in nulliparous women, although the cohort reported by Levett et al21 exhibited a marked reduction in the rate of epidural use. In contrast, Bolbol-Haghighi et al22 showed that massage practice was associated with significantly shorter durations in both the first stage (9.0 hours vs 11.5 hours) and the second stage of labour (49 minutes vs 64 minutes) among a cohort of Iranian women. However, their study included multiparous women with an overall vaginal delivery rate over 95%; they did not describe the availability of epidural analgesia for their participants. In summary, it remains unclear whether the practice of massage has consistent effects on labour progression; the underlying mechanisms of such effects are unknown.

This study had several strengths. First, it involved a large number of nulliparous labouring women. To our knowledge, this is the largest number of such women among similar published studies; it allowed us to identify any changes in the utilisation of different levels of analgesic methods, without any confounding effects related to multiparity.20 22 Second, this study involved a team of accredited and dedicated midwife trainers, which enabled us to ensure that a consistent high-quality massage technique was applied to women in the study. Third, training at 36 weeks of gestation allowed each couple (ie, a participating woman and her partner) to have sufficient time to practise massage at home and refine their technique before the woman began labour. Fourth, on admission prior to delivery, an accredited midwife trainer was available to verify each couple’s massage technique and ensure quality. Finally, there was no limit to the duration of intrapartum massage; women could receive their preferred amount of massage to achieve optimal results.

There were some notable limitations in this study. First, because of the pain relief methods used, we were unable to incorporate blinding in the trial design. However, the midwives providing intrapartum care were not involved in data collection. Second, approximately one-fifth of the participants in each group had changes to their childbirth plan, including shift to a private hospital or to planned caesarean delivery; thus, they were excluded from the final analysis. Third, we could only assess the intrapartum massage provided by the participating women’s partners; we could not assess the breathing and visualisation practice at home, which are also essential components of the overall massage programme. Finally, because continuous foetal heart monitoring was the standard method of intrapartum foetal surveillance in Hong Kong during the study period, women were unable to move freely during labour; this restriction might have limited the ability to perform certain massage techniques. Nevertheless, the shifts towards less epidural analgesia use and higher rates of analgesic-free labour, in the absence of adverse labour outcomes, support the efficacy of our massage programme.

This study demonstrated an overall shift towards using weaker pain relief modalities among women participating in an intrapartum massage programme. The findings imply that massage, in combination with controlled breathing and visualisation, may modulate pain perception among labouring women, leading to higher rates of analgesic-free labour.

Concept or design: CY Lai.
Acquisition of data: MKW Wong, WH Tong, SY Chu, KY Lau, AML Tam.
Analysis or interpretation of data: CY Lai, LL Hui.
Drafting of the manuscript: CY Lai, TTH Lao, TY Leung.
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 thank Ms Linda Kimber, Midwife/Director, and Ms Mary McNabb, Scientific Advisor and Academic Midwife, both of Childbirth Essentials, Banbury, United Kingdom, for training the midwives involved in the present study and for advising the authors on the study design.

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 (CREC Ref: 2016.332). The study has been registered at the Centre for Clinical Research and Biostatistics, The Chinese University of Hong Kong, (Unique Trial Number: CUHK_CCRB00525; https://www2.ccrb.cuhk.edu.hk/registry/public/393). All participants were informed about the nature of the study and provided written consent to participate before randomisation into study groups.

1. Kabeyama K, Miyoshi M. Longitudinal study of the intensity of memorized labour pain. Int J Nurs Pract 2001;7:46-53. Crossref

2. Rijnders M, Baston H, Schönbeck Y, et al. Perinatal factors related to negative or positive recall of birth experience in women 3 years postpartum in the Netherlands. Birth 2008;35:107-16. Crossref

3. Rachmawati IN. Maternal reflection on labour pain management and influencing factors. Br J Midwifery 2012;20:263-70. Crossref

4. Yerby M. Managing pain in labour. Part 3: pharmacological methods of pain relief. Mod Midwife 1996;6:22-5.

5. Cawthra AM. The use of pethidine in labour. Midwives Chron 1986;99:178-81.

6. Rajan L. The impact of obstetric procedure and analgesia/anaesthesia during labour and delivery on breastfeeding. Midwifery 1994;10:87-103. Crossref

7. Jones L, Othman M, Dowswell T, et al. Pain management for women in labour: an overview of systematic reviews. Cochrane Database Syst Rev 2012;(3):CD009234. Crossref

8. Anim-Somuah M, Smyth RM, Cyna AM, Cuthbert A. Epidural versus non-epidural or no analgesia for pain management in labour. Cochrane Database Syst Rev 2018;(5):CD000331. Crossref

9. Klomp T, van Poppel M, Jones L, Lazet J, Di Nisio M, Lagro-Janssen AL. Inhaled analgesia for pain management in labour. Cochrane Database Syst Rev 2012;(9):CD009351. Crossref

10. Thomson G, Feeley C, Moran VH, Downe S, Oladapo OT. 2019 Women’s experiences of pharmacological and non-pharmacological pain relief methods for labour and childbirth: a qualitative systematic review. Reprod Health 2019;16:71. Crossref

11. Chaillet N, Belaid L, Crochetière C, et al. Nonpharmacologic approaches for pain management during labor compared with usual care: a meta-analysis. Birth 2014;41:122-37. Crossref

12. Field T, Hemandez-Reif M, Taylor S, Quintino O, Burman I. Labor pain is reduced by massage therapy. J Psychosom Obstet Gynecol 1997;18:286-91. Crossref

13. McCaffery M, Beebe A. Pain: clinical manual for nursing practice. J Pain Symptom Manage 1990;5:338-9. Crossref

14. Chang MY, Wang SY, Chen CH. Effects of massage on pain and anxiety during labour: a randomized controlled trial in Taiwan. J Adv Nurs 2002;38:68-73. Crossref

15. Karami NK, Safarzadeh A, Fathizadeh N. Effect of massage therapy on severity of pain and outcome of labour in primipara. Iran J Nurs Midwifery Res 2007;12:6-9.

16. Kimber L, McNabb M, Mc Court C, Haines A, Brocklehurst P. Massage or music for pain relief in labour: a pilot randomised placebo controlled trial. Eur J Pain 2008;12:961-9. Crossref

17. Janssen P, Shroff F, Jaspar P. Massage therapy and labor outcomes: a randomized controlled trial. Int J Ther Massage Bodywork 2012;5:15-20. Crossref

18. Mortazavi SH, Khaki S, Moradi R, Heidari K, Vasegh Rahimparvar SF. Effects of massage therapy and presence of attendant on pain, anxiety and satisfaction during labour. Arch Gynecol Obstet 2012;286:19-23. Crossref

19. Silva Gallo RB, Santana LS, Jorge Ferreira CH, et al. Massage reduced severity of pain during labour: a randomised trial. J Physiother 2013;59:109-16. Crossref

20. Mc Nabb MT, Kimber L, Haines A, McCourt C. Does regular massage from late pregnancy to birth decrease maternal pain perception during labour and birth? A feasibility study to investigate a programme of massage, controlled breathing and visualization, from 36 weeks of pregnancy until birth. Complement Ther Clin Pract 2006;12:222-31. Crossref

21. Levett KM, Smith CA, Bensoussan A, Dahlen HG. Complementary therapies for labour and birth study: a randomised controlled trial of antenatal integrative medicine for pain management in labour. BMJ Open 2016;6:e010691. Crossref

22. Bolbol-Haghighi N, Masoumi SZ, Kazemi F. Effect of massage therapy on duration of labour: a randomized controlled trial. J Clin Diagn Res 2016;10:12-5. Crossref

23. Taghinejad H, Delphisheh A, Suhrabi Z. Comparison between massage and music therapies to relieve the severity of labour pain. Womens Health (Lond) 2010;6:377-81. Crossref

24. Smith CA, Levett KM, Collins CT, Dahlen HG, Ee CC, Suganuma M. Massage, reflexology and other manual methods for pain management in labour. Cochrane Database Syst Rev 2018;(3):CD009290. Crossref

25. Hui AS, Lao TT, Leung TY, Schaaf JM, Sahota DS. Trends in preterm birth in singleton deliveries in a Hong Kong population. Int J Gynaecol Obstet 2014;127:248-53. Crossref

26. The Royal College of Midwives. Childbirth essentials towards natural childbirth and beyond. Available from: https://www.rcm.org.uk/promoting/learning-careers/ accredited-learning/childbirth-essentials-towards-natural-childbirth-and-beyond/. Accessed 19 Feb 2020.

27. Fuchs AR, Behrens O, Liu HC. Correlation of nocturnal increase in plasma oxytocin with a decrease in plasma estradiol/progesterone ratio in late pregnancy. Am J Obstet Gynecol 1992;167:1559-63. Crossref

28. Tsui MH, Ngan Kee WD, Ng FF, Lau TK. A double blinded randomised placebo-controlled study of intramuscular pethidine for pain relief in the first stage of labour. BJOG 2004;111:648-55. Crossref

29. Hong Kong College of Obstetricians and Gynaecologists. Report of the territory-wide audit in obstetrics & gynaecology. 2014. Available from: https://www.hkcog.org.hk/hkcog/pages_4_307.html. Accessed 19 Feb 2020.

There has been a gradual decline in the global tuberculosis (TB) incidence and mortality rates over time, by approximately 2% and 3% per year, respectively.1 Despite this trend, TB remains a leading cause of death, and Mycobacterium tuberculosis drug resistance continues to be a major public health issue. Globally, isoniazid (INH)-resistant, rifampicin (RIF)-susceptible TB (Hr-TB) is the most common drug-resistant form of disease.2 In 2017, the rates of Hr-TB were 7.1% among patients with new TB diagnoses and 7.9% in patients who had previously undergone treatment for TB.1 In Hong Kong, the respective rates were 5.3% and 9.5%, with a combined rate of 5.7% in 2016.3 The Hr-TB is associated with worse clinical outcome and development of multidrug resistance (MDR),4 5 although the findings have been inconsistent.6 7

Isoniazid constitutes a key component in the treatment of drug-susceptible TB, through its inhibition of mycolic acid biosynthesis in the mycobacterial cell wall. Over 85% of INH resistance is conferred by mutations residing in the katG gene and inhA promoter region,8 leading to high and low levels of resistance, respectively. These mutations can readily be detected by commercial molecular line-probe assays. The level of resistance can also vary according to the co-occurrences of less common mutations in other regions.

For the treatment of Hr-TB, current World Health Organization (WHO) guidelines recommend 6 months of RIF, pyrazinamide (PZA), ethambutol, and levofloxacin (LVX). The guidelines also recommend examination of resistances towards fluoroquinolones and PZA, prior to treatment.9

In our locality, routine testing of LVX and PZA susceptibility has not been implemented for the treatment of Hr-TB. To determine the most cost-effective testing strategy, this study reviewed an annual collection of Hr-TB isolates, with the aim of determining the prevalences of LVX resistance and pncA mutations (as a molecular marker of PZA resistance in Hr-TB) and identifying factors associated with these resistances.

Data were reviewed from the TB Reference Laboratory of the Department of Health. This laboratory processes local M tuberculosis strains from specimens collected in out-patient clinics and in-patient hospitals in both public and private sectors. All phenotypic Hr-TB isolates in 2018 were identified and de-duplicated using unique patient identifiers. The following data were included in this analysis: basic patient demographics, phenotypic susceptibility results of first-line drugs (INH at critical concentrations 0.1 μg/mL and 0.4 μg/mL as recommended by the WHO, RIF, streptomycin, and ethambutol), and genotypic susceptibility results (inhA promoter region, katG codon 315, and rpoB hotspot region) from the GenoType MTBDRplus assay, version 2.0 (Hain Lifescience). Any discrepant or unsuccessful test results were resolved by the agar proportion method and/or DNA sequencing; when applicable, these additional data were reported as the final results. All available data from the Laboratory Information System were reviewed for each patient to determine any history of TB, fluorescence microscopy results (according to Global Laboratory Initiative grading10), site of isolation (pulmonary versus extrapulmonary), any documented sputum culture conversion and the duration (ie, time between the date of first positive culture for current infection and the date of consistently negative culture), and microbiological outcome.

Selected M tuberculosis strains were retrieved from the laboratory archive and sub-cultured, then subjected to LVX susceptibility testing using the BD BACTEC MGIT 960 system (Becton, Dickson and Company), in accordance with the manufacturer’s instructions. The LVX critical concentration at 1.0 μg/mL was used as recommended by the WHO. DNA extraction was performed using GenoLyse (Hain Lifescience). The pncA gene was amplified with the primers pncA-1F (5′-CGCTCAGCTGGTCATGTTC-3′) and pncA-1R (5′-CCCACCGGGTCTTCGAC-3′) to produce an amplicon of 798 bp, using the GeneAmp PCR System 9700 (Applied Biosystems). Each 25-μL reaction mixture contained 12.5 μL of GoTaq G2 Hot Start Colorless Master Mix (Promega) [1×], 0.25 μL of each primer (1.0 pM/μL), 9.5 μL PCR-grade water, and 2.5 μL of template DNA. Reaction mixtures were amplified using the following protocol: 2 minutes at 95°C; 35 cycles of 1 min at 95°C, 1 minute at 65°C, and 1 minute at 72°C; and 10 minutes at 72°C. Sequencing was performed using the 3730 × l DNA Analyzer (Applied Biosystems) with the same primers. Resulting sequences were compared with the sequence of wild-type M tuberculosis H37Rv for detection of pncA mutations. Strains with detected pncA mutations were subjected to further analysis of PZA susceptibility via the MGIT 960 system in accordance with the manufacturer’s instructions, with a critical concentration of 100 μg/mL.

Univariate analysis comprised odds ratio calculations and Fisher’s exact test. Firth logistic regression was employed for multivariable analysis because of the low outcome frequency. IBM SPSS Statistics Subscription (Windows version, IBM Corp, Armonk [NY], United States) was used for data analysis. A P value of <0.05 was considered statistically significant. This article complies with the STROBE Statement reporting guidelines.

In total, 8865 M tuberculosis isolates from 3411 patients were processed in 2018. Susceptibility profiles were available for all except repeated strains collected from the same patient within 3 months. De-duplication of 393 isolates yielded 160 patients with phenotypic Hr-TB, amounting to a case rate of 4.7%. rpoB hotspot mutations were identified in five cases by GenoType MDRTBplus, three of which were confirmed to confer RIF resistance according to rpoB sequencing results. These three cases were considered to be RIF-resistant and excluded from further analysis. The mean age of the patients in the remaining 157 cases was 61.3 years (range, 15-95 years); the male to female ratio was 1.8. Pulmonary TB was present in 90.4% of affected patients (n=142), while 9.6% of affected patients (n=15) had extrapulmonary involvement. There was a documented history of TB by positive culture in 4.5% of affected patients (n=7). Microscopy results were available for cases in which direct specimens were received by our laboratory at the time of initial diagnosis (n=45). The majority of specimens was acid-fast bacillus smear-negative (n=33), followed by 1-4 acid-fast bacilli per length (n=4), scanty (n=4), 1+ (n=2), and 2+ and 3+ (n=1 each). In terms of microbiological outcome, 76.4% of affected patients (n=120) had documented sputum culture conversion without recurrence after follow-up of at least 9 months, among which 106 attained conversion within 5 months after diagnosis. The median interval required for culture conversion was 72.5 days (range, 9-622 days). No clearance was documented for patients in the remaining 37 cases, for whom no follow-up specimens were received for >1 year.

Table 1 shows the patterns of resistance among first-line drugs and distributions of cases with LVX resistance and pncA mutations. Table 2 shows the patterns of mutations detected. With respect to INH, 45.2% (n=71) of the isolates exhibited low-level resistance (minimum inhibitory concentration 0.1-0.4 μg/mL) and 54.8% (n=86) of the isolates exhibited high-level resistance (minimum inhibitory concentration >0.4 μg/mL).

For LVX, only one strain was resistant, while 155 were sensitive (one isolate failed to be recovered). The number of resistant cases was insufficient for correlation analysis. pncA mutations were detected in 4.5% (7/157 cases), and all detected mutations were previously identified as PZA resistance–related. All isolates with pncA mutations were phenotypically resistant to PZA. Univariate analysis showed that pncA mutations were associated with documented history of TB in the affected patient (odds ratio [OR]=11.60, 95% confidence interval [CI]=1.79-75.00; P=0.03) and INH poly-resistance (OR=19.06, 95% CI=1.07-339.78; P=0.04) but not mono-resistance, as shown in Table 3. In multivariable logistic regression, pncA mutations were associated with documented history of TB in the affected patient (OR=18.03, 95% CI=2.25-153.85; P=0.008), INH triple resistance (OR=409.11, 95% CI=6.52 to >1000; P<0.001), and INH double resistance (OR=13.50, 95% CI=1.43 to >1000; P=0.019).

In this study, the frequency of LVX resistance was very low in Hr-TB. Levofloxacin is an important drug in the management of drug-resistant TB. In a 2009 study in Shanghai, Xu et al11 estimated the rates of LVX resistance to be 1.9% in strains pan-susceptible to first-line drugs, 6.7% in INH mono-resistant strains, and ≤25% in MDR-TB. Independent risk factors associated with LVX resistance included MDR, RIF mono-resistance, poly-resistance (resistance to ≥2 first-line drugs, but not MDR), age ≥46 years, and TB re-treatment, but not INH mono-resistance. These findings were consistent with the results of a 2017 study in Ningbo (a city near Shanghai), whereby Che et al12 revealed no LVX resistance in strains pan-susceptible to first-line drugs, 3% in strains with any resistance to first-line drugs except MDR, and ≤30% in MDR-TB. Most fluoroquinolone resistance was present in the MDR group. Che et al12 also found that prevalence of LVX resistance in MDR-TB increased with duration of inappropriate treatment before LVX; this relationship was stronger than any relationships with previous fluoroquinolone exposures. Further studies are needed to determine whether this finding also applies to Hr-TB. Levofloxacin resistance is uncommon in non-MDR strains, especially in the present study. Since 2015, our laboratory has performed LVX susceptibility testing on 640 Hr-TB strains, identifying only two resistant cases (0.31%; unpublished data), including the one in this study. Clinicians could treat affected patients empirically, in accordance with WHO recommendations; further testing could be arranged based on clinical, radiological, and microbiological responses during early follow-up. In patients with Hr-TB that exhibits LVX resistance or poly-resistance to other primary agents, individualised adjustment of the treatment regimen is recommended in accordance with WHO guidelines9 (eg, exclusion of LVX or inclusion of second-line TB medicines). In our single case with LVX resistance, the patient had no history of TB. His sputum acid-fast bacillus smear result was 2+. The Hr-TB strain isolated was initially LVX-sensitive without mutations in rpoB, gyrA, or gyrB on diagnosis and follow-up at 8 months. The patient continued to exhibit sputum culture-positive results after 15 months of treatment; further analysis revealed that the isolate had acquired gyrA mutation (detected by line-probe assays) and LVX phenotypic resistance. It then developed into an MDR strain; the patient eventually achieved culture conversion at approximately 18 months after diagnosis. Underlying hetero-resistance was a possible contributing factor.

Phenotypic PZA susceptibility testing has often been problematic. pncA mutations constitute the most common and primary determinant of PZA resistance, with reported sensitivity of approximately 80% to 95% and specificity of approximately 85% to 100%.13 14 Resistance-conferring mutations are known to be diverse and scattered over the gene’s full length. Our seven isolates all had unique mutations (Gln10Arg, His51Tyr, Trp68Stop, Thr47Pro, Ala102Thr, Asp63Gly, and Val139Ala), which were associated with PZA resistance at a high probability of 0.985 according to available literature15 (except Val139Ala, probability of resistance=0.783). These seven isolates were also confirmed to be phenotypically resistant to PZA in our study.

Thus far, investigations of phenotypic and genotypic PZA resistance have generally focused on MDR-TB. Concerning non-MDR-resistant strains, there is considerable geographical variation in the reported rates, from 2.92% to 4.8% in the United States,16 17 to 24% in the Western Pacific and 75% in South East Asia.18 For Hr-TB in particular, phenotypic PZA resistance was reported in 2% of INH mono-resistant strains in South Africa,19 no PZA resistance was detected (phenotypic or pncA mutation) in Georgia (Eastern Europe) among Hr-TB strains,20 and ≤14.9% of Hr-TB strains exhibited pncA mutations in Vietnam.21 In our study, we observed that 4.5% of Hr-TB isolates had pncA mutations. Such variation may be related to multiple factors, including differences in inclusion criteria, testing method, treatment, infection control practice, and disease burden. In terms of risk factors, we found that history of TB in the affected patient and poly-resistance (but not mono-resistance) were significantly associated with pncA mutations. These results are consistent with findings from previous studies, which also showed an association with MDR.21 22 23 Clinicians are advised to consider the possibility of PZA resistance in patients with Hr-TB and these risk factors. Routine testing for LVX or PZA susceptibility in patients with INH mono-resistant TB alone (ie, no other risk factors) is not considered cost-effective, based on our findings.

In our study, most patients with Hr-TB had a satisfactory microbiological outcome without recurrence. However, the outcome was unknown in 23.6% of cases. Most of these involved patients were diagnosed and managed in hospitals, where only positive isolates were subjected to reference testing at our laboratory. Based on the recommended management regimen for TB in Hong Kong, these patients would have been followed up until sputum culture conversion. Considering the time elapsed since the latest positive laboratory result, the patients in most cases with unknown outcomes likely already had culture conversion.

In general, Hr-TB is presumably associated with higher rates of treatment failure and MDR acquisition, compared with drug-susceptible strains.4 5 Important considerations for disease control include recognition of risk factors associated with Hr-TB (eg, history of TB treatment, incarceration in prisons, and homelessness7 24), early detection of INH resistance, exclusion of RIF resistance by appropriate molecular methods (eg, line-probe assays),2 and compliance with current treatment guidelines. With the maturation of whole-genome sequencing, the technology is becoming integrated into the local routine drug susceptibility testing protocol for all M tuberculosis isolates; multiple and uncommon molecular markers of drug resistance might thus be identified earlier in a single set of assays for proper treatment guidance.

Because our laboratory serves as a local TB reference laboratory, this study was able to use a representative M tuberculosis collection that reflected the status of Hr-TB in Hong Kong. Furthermore, because our laboratory serves as a supranational reference laboratory, we were able to perform an array of confirmatory tests in accordance with WHO recommendations, thus ensuring reliable results. However, this study had some limitations. First, it lacked data regarding microscopy results, drug treatment, and clinical outcome, thus preventing a more comprehensive assessment. Second, the small sample size and low outcome frequencies of LVX and PZA non-susceptibilities led to limited correlation analysis. Third, less common mutations mediating INH resistance were not examined to determine their prevalences; Hr-TB cases with these mutations would only be identified on the basis of phenotypic results. A substantial proportion of cases (22.3%; n=35) had negative GenoType MDRTBplus results. Because this test only detects the most common mutations at the inhA promoter region and katG codon 315, other relevant mutations might have been missed. Furthermore, in 3.2% (n=5) of cases with high-level INH resistance, mutations were detected in the inhA promoter region alone. The availability of preliminary information before receiving the INH phenotypic susceptibility results might lead to mismanagement of patient treatment (eg, INH inclusion in the therapeutic regimen). Fourth, low-level RIF resistance mediated by mutations outside the rpoB hotspot region and undetected by phenotypic testing was not ruled out in this study. Inclusion of such isolates might have led to overestimation of true resistance in Hr-TB. Fifth, PZA resistance could be mediated by other less common mutations, which were not examined by performing phenotypic PZA susceptibility testing on all Hr-TB isolates; thus, we may have underestimated its prevalence. Our laboratory has found that all PZA-resistant M tuberculosis isolates thus far could be confirmed through pncA mutation analysis. Nevertheless, we presume that these limitations do not greatly affect the reliability and overall interpretation of our findings.

In Hong Kong, the rate of Hr-TB was 4.7% among active TB cases in 2018. Levofloxacin and PZA resistances among Hr-TB were uncommon (0.6% and 4.5%, respectively). History of TB in the affected patient and INH poly-resistance, including both double and triple resistances, were independent risk factors for pncA mutations. The findings from this study do not support routine susceptibility testing for these two drugs in patients with INH mono-resistant TB who lack additional risk factors.

Concept or design: KKM Ng.
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: KKM Ng.

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 contribution by staff, especially Mr Steven CW Lui and Ms Angela WL Lau, at the Tuberculosis Laboratory and 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.

Ethics approval for this study was obtained from the Ethics Committee of the Department of Health, Hong Kong SAR Government (Ref: LM 425/2019). All included patients consented to testing for tuberculosis.

1. World Health Organization. Global tuberculosis report 2018. Available from: https://www.who.int/tb/publications/global_report/en/. Accessed 19 Sep 2019.

2. Romanowski K, Campbell JR, Oxlade O, Fregonese F, Menzies D, Johnston JC. The impact of improved detection and treatment of isoniazid resistant tuberculosis on prevalence of multi-drug resistant tuberculosis: a modelling study. PLoS One 2019;14:e0211355. Crossref

3. Tuberculosis & Chest Service of Department of Health, Hong Kong SAR Government. Annual report 2016. Available from: https://www.info.gov.hk/tb_chest/doc/Annual_Report_2016.pdf. Accessed 19 Sep 2019.

4. Gegia M, Winters N, Benedetti A, van Soolingen D, Menzies D. Treatment of isoniazid-resistant tuberculosis with first-line drugs: a systematic review and meta-analysis. Lancet Infect Dis 2017;17:223-34. Crossref

5. Karo B, Kohlenberg A, Hollo V, et al. Isoniazid (INH) mono-resistance and tuberculosis (TB) treatment success: analysis of European surveillance data, 2002 to 2014. Euro Surveill 2019;24:1800392. Crossref

6. Salindri AD, Sales RF, DiMiceli L, Schechter MC, Kempker RR, Magee MJ. Isoniazid monoresistance and rate of culture conversion among patients in the State of Georgia with confirmed tuberculosis, 2009-2014. Ann Am Thorac Soc 2018;15:331-40. Crossref

7. Cattamanchi A, Dantes RB, Metcalfe JZ, et al. Clinical characteristics and treatment outcomes of isoniazid-monoresistant tuberculosis. Clin Infect Dis 2009;48:179-85. Crossref

8. Seifert M, Catanzaro D, Catanzaro A, Rodwell TC. Genetic mutations associated with isoniazid resistance in Mycobacterium tuberculosis: a systematic review. PLoS One 2015;10:e0119628. Crossref

9. World Health Organization. WHO consolidated guidelines on drug-resistant tuberculosis treatment. Available from: https://www.who.int/tb/publications/2019/consolidated-guidelines-drug-resistant-TB-treatment/en/. Accessed 19 Sep 2019. Crossref

10. Global Laboratory Initiative. Mycobacteriology Laboratory Manual. 1st ed. Global Laboratory Initiative; 2014

11. Xu P, Li X, Zhao M, et al. Prevalence of fluoroquinolone resistance among tuberculosis patients in Shanghai, China. Antimicrob Agents Chemother 2009;53:3170-2. Crossref

12. Che Y, Song Q, Yang T, Ping G, Yu M. Fluoroquinolone resistance in multidrug-resistant Mycobacterium tuberculosis independent of fluoroquinolone use. Eur Respir J 2017;50:1701633. Crossref

13. Streicher EM, Maharaj K, York T, et al. Rapid sequencing of the Mycobacterium tuberculosis pncA gene for detection of pyrazinamide susceptibility. J Clin Microbiol 2014;52:4056-7. Crossref

14. Khan MT, Malik SI, Ali S, et al. Pyrazinamide resistance and mutations in pncA among isolates of Mycobacterium tuberculosis from Khyber Pakhtunkhwa, Pakistan. BMC Infect Dis 2019;19:116. Crossref

15. Miotto P, Cabibbe AM, Feuerriegel S, et al. Mycobacterium tuberculosis pyrazinamide resistance determinants: a multicenter study. mBio 2014;5:e01819-4.Crossref

16. Kurbatova EV, Cavanaugh JS, Dalton T, Click ES, Cegielski JP. Epidemiology of pyrazinamide-resistant tuberculosis in the United States, 1999-2009. Clin Infect Dis 2013;57:1081-93. Crossref

17. Budzik JM, Jarlsberg LG, Higashi J, et al. Pyrazinamide resistance, Mycobacterium tuberculosis lineage and treatment outcomes in San Francisco, California. PLoS One 2014;9:e95645. Crossref

18. Whitfield MG, Soeters HM, Warren RM, et al. A global perspective on pyrazinamide resistance: systematic review and meta-analysis. PLoS One 2015;10:e0133869. Crossref

19. Whitfield MG, Streicher EM, Dolby T, et al. Prevalence of pyrazinamide resistance across the spectrum of drug resistant phenotypes of Mycobacterium tuberculosis. Tuberculosis (Edinb) 2016;99:128-30. Crossref

20. Sengstake S, Bergval IL, Schuitema AR, et al. Pyrazinamide resistance-conferring mutations in pncA and the transmission of multidrug resistant TB in Georgia. BMC Infect Dis 2017;17:491. Crossref

21. Huy NQ, Lucie C, Hoa T, et al. Molecular analysis of pyrazinamide resistance in Mycobacterium tuberculosis in Vietnam highlights the rate of pyrazinamide resistance-associated mutations in clinical isolates. Emerg Microbes Infect 2017;6:e86. Crossref

22. Li D, Hu Y, Werngren J, et al. Multicenter study of the emergence and genetic characteristics of pyrazinamide-resistant tuberculosis in China. Antimicrob Agents Chemother 2016;60:5159-66. Crossref

23. Chiu YC, Huang SF, Yu KW, Lee YC, Feng JY, Su JY. Characteristics of pncA mutations in multidrug-resistant tuberculosis in Taiwan. BMC Infect Dis 2011;11:240. Crossref

24. Smith CM, Trienekens SC, Anderson C, et al. Twenty years and counting: epidemiology of an outbreak of isoniazid-resistant tuberculosis in England and Wales, 1995 to 2014. Euro Surveill 2017;22:30467. Crossref

Various randomised controlled trials have demonstrated that non–vitamin K oral anticoagulants (NOACs), including factor Xa and direct thrombin inhibitors, are superior to warfarin, a vitamin K antagonist, in terms of efficacy and safety for preventing stroke and systemic thromboembolisms in patients with non-valvular atrial fibrillation (NVAF).1 2 3 4 The superiority of NOACs compared with warfarin appears to be consistent across ethnic groups, including Asian populations.5 Furthermore, data from two sub-studies of the NOAC trials6 7 and a real-world cohort study8 suggested that NOACs may also be superior to warfarin in terms of maintaining and preserving renal function. A US-based cohort study demonstrated a lower risk of decline in renal function among patients receiving NOACs than among patients receiving warfarin.8 Moreover, findings from the ROCKET AF (Rivaroxaban Once-Daily, Oral, Direct Factor Xa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation) and RE-LY (Randomized Evaluation of Long Term Anti-coagulation Therapy) trials revealed more rapid estimated glomerular filtration rate (eGFR) decline in patients receiving warfarin than in patients receiving rivaroxaban and dabigatran, respectively.6 7 8 Further studies have demonstrated that warfarin treatment may be associated with more rapid progression of chronic kidney disease and can cause acute kidney injury.8 9 10 This decline in renal function has been attributed to a phenomenon known as ‘warfarin-related nephropathy’, which is associated with vitamin K antagonism and excessive anticoagulation.8 9 11 In contrast, NOACs may offer renovascular protection through pharmacological mechanisms such as the inhibition of thrombin and factor Xa.8 12 13

Differences in the pharmacological actions of warfarin and NOACs are reflected in the growing research that suggests NOACs are more effective than warfarin for preserving renal function.8 14 Considering that Asian warfarin users tended to have a lower time in therapeutic range (TTR)15 16 of the international normalised ratio (INR), which is associated with decline in renal function,6 14 the renal effects of NOACs compared with warfarin may differ from the effects in non-Asians. Dosage prescription patterns, such as the frequency of low-dose NOAC prescriptions, also vary between Asian and non-Asian populations14 17 18; this may also affect renal outcomes because the renal effects of NOACs appear to be dose-dependent.14 19 Furthermore, because of differences in NOAC-related bleeding risk between Asian and non-Asian populations, the renal protection effects of NOACs may also vary; major bleeding can cause decline in renal function.5 14 20 21 To our knowledge, there remain limited data comparing NOACs to warfarin in terms of decline in renal function among Asian patients. In this study, we sought to assess the renal outcomes of an ethnic Chinese patient population with NVAF who received NOACs (ie, apixaban, dabigatran, and rivaroxaban) compared with patients who received warfarin.

This retrospective cohort study included four study groups: warfarin, apixaban, dabigatran, and rivaroxaban. Each NOAC was compared with warfarin.

Data were extracted from patients’ electronic medical records in the Prince of Wales Hospital of Hong Kong. In total, 2346 consecutive patients with a prescription of warfarin or one of the three NOACs (apixaban, rivaroxaban, and dabigatran) in our hospital were screened for eligibility for this analysis. Inclusion criteria were: first began to receive an oral anticoagulant between 1 January 2012 and 31 December 2016, NVAF, age ≥18 years, minimum on-treatment duration of 3 months, and availability of laboratory data concerning serum creatinine at baseline and during follow-up. We excluded warfarin-experienced or NOAC-experienced patients to minimise confounding bias.8 22 Other exclusion criteria were previous kidney failure, valvular atrial fibrillation,23 and/or other indications for anticoagulation. A pre-study power analysis to detect a 10% difference in the incidence of ≥30% decline in eGFR between NOACs (pooled for consideration as a single unit) and warfarin revealed that the minimum sample size was 200 patients per arm (all NOACs vs warfarin). The sample size in each NOAC group was then matched to the approximate proportion of patients that were prescribed each of the three NOACs in actual clinical practice, in accordance with the preferences of local physicians. In our hospital during the study period, rivaroxaban was available earlier locally and was more frequently prescribed than the other two NOACs; apixaban was the last NOAC to receive local approval and therefore exhibited a lower rate of prescription at the time of the study. This paper adheres to the STROBE reporting guidelines for observational studies.

We studied the three renal outcome endpoints: ≥30% decline in eGFR, doubling of serum creatinine, and kidney failure. Doubling of serum creatinine has been used as a surrogate endpoint when studying the progression of kidney disease in clinical trials.24 Based on the findings of clinical trials and meta-analyses, the National Kidney Foundation and US Food and Drug Administration proposed that with at least 2 to 3 years of follow-up, a 30% to 40% decline in eGFR may also be regarded as a surrogate end point; thus, it has been used as a renal endpoint in previous cohort analyses.8 24 Because doubling of serum creatinine and kidney failure occur late in kidney disease, ≥30% decline in eGFR serves as a more sensitive renal decline endpoint; this change is clinically significant regardless of a low follow-up or event rate.8 24 Kidney failure is defined as eGFR <15 mL/min/1.73 m², long-term kidney dialysis, or kidney transplantation.8 25 Efficacy outcomes were stroke (ischaemic or haemorrhagic) or systemic embolism (SE). Based on the initial dose prescribed, the prevalence of dose reduction for each NOAC (apixaban 2.5 mg twice daily, dabigatran 75 mg twice daily, and rivaroxaban 15 mg once daily) without a renal indication (eGFR <30 mL/min for apixaban and dabigatran; eGFR <50 mL/min for rivaroxaban) was assessed.26

Using the treatment initiation date as our index date, we retrieved the pretreatment creatinine value nearest to the index date as the baseline creatinine; we used this value to calculate the baseline eGFR by means of the Chronic Kidney Disease Epidemiology Collaboration equation.8 27 Hospital electronic records were used to identify co-morbidities and specific drug prescriptions within 3 months prior to the index date. Baseline HAS-BLED and CHA2DS2-VASc scores were also recorded. The TTR of patients in the warfarin cohort was calculated as the number of INRs in therapeutic range (INR=2-3) divided by the total INRs recorded for each patient during the analysed period.28 29 Patients were followed up until the end of treatment, death, or when any efficacy or renal endpoint(s) were reached.

For minimisation of potential confounding, we used inverse probability of treatment weighting (IPTW) to balance identified covariates.8 14 17 30 Generalised boosted models were used to estimate propensity scores and weights for optimal balance across treatment groups.31 32 Weights were obtained to gather estimates representing the mean effects of treatment among treated groups.8 14 Baseline characteristics (eg, patient baseline medications and pre-existing co-morbidities which may affect outcomes) were included in our model (online supplementary Table).8 14 Both CHA2DS2-VASc and HAS-BLED scores were not included in the model because they are composite scores derived from other covariates.14 The absolute standardised mean difference was calculated for each NOAC versus warfarin to ensure that the cohorts were sufficiently balanced before comparison of each NOAC to warfarin. An absolute standardised mean difference of <0.2 is considered balanced for each baseline covariate when comparing each NOAC to warfarin.8 31 Fisher’s exact test was used to compare the frequencies of dose reduction without renal indication among NOACs.

Because of some extremely high or low weights in our weighted population, we truncated weights at the 1st and 99th percentiles before conducting weighted analysis.8 33 We calculated hazard ratios using weighted Cox proportional hazards regression, then generated weighted Kaplan–Meier curves that compared each NOAC to warfarin. Cumulative incidences for the Kaplan–Meier curves were presented as mean percentage incidences with 95% confidence intervals. A P value of <0.05 was considered statistically significant. Predefined subgroup analysis was performed for factors potentially associated with renal outcome, including age (≥75 or <75 years); sex; and baseline diabetes mellitus, heart failure, and eGFR (≥60 mL/min/1.73 m² or <60 mL/min/1.73 m²).

We identified 600 patients with NVAF who were receiving oral anticoagulants: 200, 100, 100, and 200 patients were receiving warfarin, apixaban, dabigatran, and rivaroxaban, respectively. After IPTW, all identified baseline characteristics were balanced between warfarin and each NOAC group (Table 1). The mean follow-up duration of the overall study cohort was 1000 ± 436 days. The mean follow-up durations for each NOAC group were as follows: apixaban (790 ± 345 days), dabigatran (1187 ± 322 days), and rivaroxaban (999 ± 430 days); the median follow-up durations were 806, 1416, and 1074 days, respectively. The mean TTR of the warfarin cohort was 44.3%. The frequencies of dose reduction without a renal indication for apixaban, rivaroxaban and dabigatran were 46.9%, 35.7% and 2.0%, respectively (P<0.001 for dabigatran vs both apixaban and rivaroxaban).

When the three NOACs were pooled for consideration as a single unit and compared with warfarin, NOAC users exhibited lower risks of ≥30% decline in eGFR (hazard ratio [HR]=0.339; 95% confidence interval [CI]=0.276-0.417; P<0.001) and doubling of serum creatinine (HR=0.550; 95% CI=0.387-0.782; P<0.001). Individual comparisons of each NOAC to warfarin (Table 2) revealed that dabigatran and rivaroxaban users both had lower risks of ≥30% decline in eGFR (both P<0.001) and doubling of serum creatinine (both P<0.05). However, apixaban users only had a lower risk of ≥30% decline in eGFR (P<0.001). Despite trends suggestive of lower kidney failure risk in patients receiving dabigatran or rivaroxaban, the overall use of NOACs was not significantly associated with lower kidney failure risk, compared with the use of warfarin. Figure 1 shows the weighted Kaplan–Meier curves for the renal endpoints. For the efficacy outcome, dabigatran was associated with a lower incidence of stroke/SE (HR=0.151; 95% CI=0.054-0.423); P<0.001 vs warfarin), whereas the use of apixaban or rivaroxaban was not significantly associated with a lower incidence of stroke/SE, compared with the use of warfarin (Fig 2).

Analysis of the main renal endpoint, ≥30% decline in eGFR, consistently favoured the use of dabigatran or rivaroxaban, compared with warfarin, across all subgroups (Fig 3). However, the use of apixaban was not associated with a reduced risk of ≥30% decline in eGFR in three subgroups: men (P=0.057), patients with heart failure (P=0.835), and patients without diabetes mellitus (P=0.090).

Our cohort study provides important insights into the long-term renal impacts of NOACs versus warfarin in an ethnic Chinese population. Decline in renal function was evident among both warfarin and NOAC users in our cohort. However, the use of NOACs was generally associated with better long-term renal outcomes, compared with the use of warfarin, among Chinese patients. The general superiority of NOACs compared with warfarin was most evident for the ≥30% decline in eGFR surrogate endpoint. The use of dabigatran or rivaroxaban was associated with lower risks of ≥30% decline in eGFR and doubling of creatinine in the overall population and across predefined demographic and clinical subgroups; in contrast, the use of apixaban was not associated with a lower risk of doubling of serum creatinine in the overall population, nor was it associated with a lower risk of ≥30% decline in eGFR among several subgroups (men, patients without diabetes mellitus, and patients with heart failure).

Pharmacological mechanisms may explain our findings concerning NOAC superiority. Because warfarin is a vitamin K antagonist, it has inhibitory effects on matrix gamma-carboxyglutamic acid, a vitamin K–dependent protein which normally protects against vascular calcification; thus, warfarin administration potentially stimulates and accelerates the calcification of renal vascular tissue, which promotes nephropathy.8 34 35 The mechanism of warfarin-related nephropathy has various contributing factors, such as the occurrence of glomerular haemorrhage and subsequent tubular injury because of red blood cell casts and haem-related free radical injury.10 36 37 Alternatively, NOACs may offer renovascular protection through distinct mechanisms such as the inhibition of thrombin and factor Xa.8 12 13

Our data are generally consistent with previous studies concerning the renal outcomes of NOACs versus warfarin. A study by Yao et al8 regarding the renal outcomes of NOACs showed that dabigatran was associated with a lower risk of ≥30% decline in eGFR, while rivaroxaban was associated with lower risks of ≥30% decline in eGFR and doubling of serum creatinine. Analysis of the RE-LY and ROCKET AF trials similarly showed more rapid decline in eGFR among warfarin users, compared with dabigatran and rivaroxaban users.6 7 8 Hernandez et al38 demonstrated rivaroxaban superiority for adverse renal events, compared with warfarin, in patients with NVAF who had diabetes mellitus. However, our results for apixaban were inconsistent with the findings of the ARISTOTLE trial, which did not show significant apixaban superiority in terms of renal function preservation; the analysis showed similar but slightly greater decline in eGFR among apixaban users, compared with warfarin users.8 39 Yao et al8 also showed no clear benefits for apixaban, compared with warfarin, in terms of renal protection. Nonetheless, a study in Taiwan by Chan et al14 showed that, compared with warfarin, all three NOACs were associated with lower risk for acute kidney injury in both chronic kidney disease-free and chronic kidney disease cohorts.

The differences between our findings and the results of previous studies—especially with respect to apixaban in our cohort versus the ARISTOTLE subanalysis39—may have several explanations. As mentioned by Chan et al,14 Asian populations tended to have lower TTR with warfarin usage, compared with non-Asians15 16; our warfarin cohort had a mean TTR of 44.3%, which was considerably lower than findings in non-Asian populations.15 16 Combined with findings that renal deterioration is greater when warfarin is poorly controlled—especially with INR levels above the target range, as demonstrated in the RE-LY trial6—indicates that Asian populations, such as the Chinese, may have an elevated risk of warfarin-related nephropathy.14 Because Asian patients may be more susceptible to renal decline associated with warfarin use, apixaban may appear superior to warfarin in Asian populations, although this superiority may not persist in non-Asian populations.8 Additional apixaban superiority in Asian populations, as discussed by Chan et al,14 may also be explained by the superior efficacy and safety of NOACs in Asians, compared with non-Asians.5 21 Because major bleeding can be associated with renal function deterioration, the greater efficacy and safety of NOACs in Asians may facilitate renal risk reduction in such populations.5 14 21 Notably, there was a high prevalence of non-guideline dose reduction without a renal indication26 in the apixaban group, compared with other NOACs, in our study; this dose reduction has been associated with worse stroke prevention effectiveness and provides no safety benefit.40 Nonetheless, apixaban was not significantly associated with risk reduction of the other two renal outcomes in our study, compared with warfarin. This may suggest uncertainty concerning its renal risk reduction superiority compared with warfarin. Overall, such inconsistencies across studies indicate the need for additional research; they may also reflect insufficient statistical power in our study to generate more robust conclusions.

The aforementioned findings concerning greater risk of warfarin-related nephropathy and possible lower risk of renal decline with NOAC usage in Asian populations are also potentially reflected in the comparatively lower HRs for renal endpoints in our study, compared with the US-based cohort reported by Yao et al.8 When the NOACs were pooled (for consideration as a single unit) and compared with warfarin, HRs for ≥30% decline in eGFR and doubling of serum creatinine were both lower in our population, compared with the pooled results described by Yao et al8 (HR=0.77; 95% CI=0.66-0.89 and HR=0.62; 95% CI=0.40-0.95).

Notable strengths of our study were its long study period and subsequent long mean follow-up duration. The longer follow-up duration, compared with previous cohort studies, indicates that previous findings concerning NOAC superiority for renal outcomes also persist during longer follow-up periods. Furthermore, our database comprised each patient’s complete laboratory data; this allowed accurate recording of each renal outcome through serum creatinine and eGFR values, thus enhancing the consistency and preciseness of renal measurement across all patients. We minimised potential confounding by only including patients who were first-time users of oral anticoagulants; this enabled us to balance numerous important baseline characteristics.

Regarding limitations, although we utilised IPTW to balance baseline covariates, confounding bias may have persisted in the study.8 14 Nonetheless, we achieved balance concerning the most important identified baseline covariates that may impact renal function across treatment groups. Moreover, although the smaller number of events may have limited the statistical power with respect to the less sensitive endpoint of kidney failure, by including ≥30% decline in eGFR as a sensitive renal outcome, we were able to sufficiently assess early renal decline. Smaller declines in renal function (eg, ≥30% decline in eGFR) serve as valuable and sensitive indicators of renal decline8 that has been regarded as a useful surrogate endpoint for progression to kidney failure24; it is also reportedly associated with risks of end-stage renal disease and mortality.41 Frequency of testing, as mentioned in Yao et al,8 may also affect results; the inclusion of patients with more follow-up creatinine tests leads to greater sensitivity concerning outcome incidence, compared with patients who underwent fewer tests. To minimise the potential impact of this sensitivity on the renal endpoints, we only included patients for whom creatinine tests were available throughout the entire follow-up period; this was possible because all patients were treated in a single centre.

Overall, the general consistency of our results with the findings of previous cohort studies, as well as the findings of the RE-LY and ROCKET AF trials, enhances the reliability and robustness of our results.6 7 8 14 Nonetheless, further studies are needed to identify consistencies among the existing discrepancies, especially concerning apixaban. Greater certainty regarding renal outcomes of all NOACs is also important because one previous meta-analysis of various randomised controlled trials concluded that the risk of kidney failure associated with NOACs was similar to the risk associated with other anticoagulants.42 Finally, although this study only involved Hong Kong Chinese patients, whose responses to NOACs and warfarin may differ from the responses of their non-Asian counterparts, the consistency of the results with findings from studies in other regions suggests widespread applicability of the findings.

In patients with NVAF who are receiving oral anticoagulants, gradual renal impairment is associated with worse clinical outcomes.39 43 Our results suggested that patients receiving oral anticoagulant therapy, particularly warfarin, should undergo close renal function monitoring. Decline in renal function during anticoagulant therapy may be less likely to occur when receiving NOACs than when receiving warfarin. The NOAC efficacy findings in this study were generally consistent with previously reported data in terms of stroke/SE prevention non-inferiority or superiority, compared with warfarin.44 45 46 47 In particular, the superior efficacy of dabigatran compared with warfarin, in this local study population is reassuring. The inconsistencies of NOAC prescribing patterns with drug labelling in routine clinical practice, particularly regarding apixaban, should receive greater attention, because dose reduction in the absence of a renal indication has been associated with worse effectiveness and no safety benefit in apixaban-treated patients with normal or mildly impaired renal function.40

Compared with warfarin, NOAC treatment may be associated with a lower risk of renal decline in Chinese populations; this should be considered by clinicians during the selection of anticoagulant treatment. Further studies are needed in Asian populations (eg, Chinese) to better understand the renal superiority or inferiority of NOACs compared with warfarin. Besides, NOAC-to-NOAC comparisons are needed to inform treatment selection. Additional research is needed in specific populations, such as patients with diabetes mellitus or heart failure, to better understand the impacts of baseline co-morbidities on renal risk reduction related to the use of NOACs, compared with the use of warfarin. Large-scale studies should also investigate how dosage patterns may influence renal outcomes.

Concept or design: APW Lee.
Acquisition of data: All authors.
Analysis or interpretation of data: All authors.
Drafting of the manuscript: All authors.
Critical revision of the manuscript for important intellectual content: APW Lee.

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.

APW Lee has received research grants from Bayer, Pfizer, and Boehringer Ingelheim.

This work was funded by the Hong Kong SAR Government Health and Medical Research Fund (05160976). The funder had no role in study design, data collection/analysis/interpretation, or manuscript preparation.

The study was approved by The Joint Chinese University of Hong Kong–New Territories East Cluster Clinical Research Ethics Committee (Ref CREC 2019.405). Informed consent was waived because of the retrospective nature of this study.

1. Hart RG, Pearce LA, Aguilar MI. Meta-analysis: antithrombotic therapy to prevent stroke in patients who have nonvalvular atrial fibrillation. Ann Intern Med 2007;146:857-67. Crossref

2. Connolly SJ, Ezekowitz MD, Yusuf S, et al. Dabigatran versus warfarin in patients with atrial fibrillation. N Engl J Med 2009;361:1139-51. Crossref

3. Patel MR, Mahaffey KW, Garg J, et al. Rivaroxaban versus warfarin in nonvalvular atrial fibrillation. N Engl J Med 2011;365:883-91. Crossref

4. Granger CB, Alexander JH, McMurray JJ, et al. Apixaban versus warfarin in patients with atrial fibrillation. N Engl J Med 2011;365:981-92. Crossref

5. Lip GY, Wang KL, Chiang CE. Non-vitamin K antagonist oral anticoagulants (NOACs) for stroke prevention in Asian patients with atrial fibrillation: time for a reappraisal. Int J Cardiol 2015;180:246-54. Crossref

6. Böhm M, Ezekowitz MD, Connolly SJ, et al. Changes in renal function in patients with atrial fibrillation: an analysis from the RE-LY trial. J Am Coll Cardiol 2015;65:2481-93. Crossref

7. Fordyce CB, Hellkamp AS, Lokhnygina Y, et al. On-treatment outcomes in patients with worsening renal function with rivaroxaban compared with warfarin: Insights from ROCKET AF. Circulation 2016;134:37-47. Crossref

8. Yao X, Tangri N, Gersh BJ, et al. Renal outcomes in anticoagulated patients with atrial fibrillation. J Am Coll Cardiol 2017;70:2621-32. Crossref

9. Brodsky SV, Nadasdy T, Rovin BH, et al. Warfarin-related nephropathy occurs in patients with and without chronic kidney disease and is associated with an increased mortality rate. Kidney Int 2011;80:181-9. Crossref

10. Brodsky SV, Satoskar A, Chen J, et al. Acute kidney injury during warfarin therapy associated with obstructive tubular red blood cell casts: a report of 9 cases. Am J Kidney Dis 2009;54:1121-6. Crossref

11. Brodsky SV, Collins M, Park E, et al. Warfarin therapy that results in an international normalization ratio above the therapeutic range is associated with accelerated progression of chronic kidney disease. Nephron Clin Pract 2010;115:142-6. Crossref

12. Sparkenbaugh EM, Chantrathammachart P, Mickelson J, et al. Differential contribution of FXa and thrombin to vascular inflammation in a mouse model of sickle cell disease. Blood 2014;123:1747-56. Crossref

13. Lee IO, Kratz MT, Schirmer SH, Baumhäkel M, Böhm M. The effects of direct thrombin inhibition with dabigatran on plaque formation and endothelial function in apolipoprotein E-deficient mice. J Pharmacol Exp Ther 2012;343:253-7. Crossref

14. Chan YH, Yeh YH, Hsieh MY, et al. The risk of acute kidney injury in Asians treated with apixaban, rivaroxaban, dabigatran, or warfarin for non-valvular atrial fibrillation: a nationwide cohort study in Taiwan. Int J Cardiol 2018;265:83-9. Crossref

15. Wallentin L, Yusuf S, Ezekowitz MD, et al. Efficacy and safety of dabigatran compared with warfarin at different levels of international normalised ratio control for stroke prevention in atrial fibrillation: an analysis of the RE-LY trial. Lancet 2010;376:975-83. Crossref

16. Singer DE, Hellkamp AS, Piccini JP, et al. Impact of global geographic region on time in therapeutic range on warfarin anticoagulant therapy: data from the ROCKET AF clinical trial. J Am Heart Assoc 2013;2:e000067. Crossref

17. Nielsen PB, Skjøth F, Søgaard M, Kjældgaard JN, Lip GY, Larsen TB. Effectiveness and safety of reduced dose non- Vitamin K antagonist oral anticoagulants and warfarin in patients with atrial fibrillation: propensity weighted nationwide cohort study. BMJ 2017;356:j510. Crossref

18. Hori M, Matsumoto M, Tanahashi N, et al. Rivaroxaban vs. warfarin in Japanese patients with atrial fibrillation–the J-ROCKET AF study. Circ J 2012;76:2104-11. Crossref

19. Ryan M, Ware K, Qamri Z, et al. Warfarin-related nephropathy is the tip of the iceberg: direct thrombin inhibitor dabigatran induces glomerular hemorrhage with acute kidney injury in rats. Nephrol Dial Transplant 2014;29:2228-34. Crossref

20. Makris K, Spanou L. Acute kidney injury: definition, pathophysiology and clinical phenotypes. Clin Biochem Rev 2016;37:85-98. Crossref

21. Wang KL, Lip GY, Lin SJ, Chiang CE. Non-vitamin K antagonist oral anticoagulants for stroke prevention in Asian patients with nonvalvular atrial fibrillation: meta-analysis. Stroke 2015;46:2555-61. Crossref

22. Ray WA. Evaluating medication effects outside of clinical trials: new-user designs. Am J Epidemiol 2003;158:915-20. Crossref

23. January CT, Wann LS, Calkins H, et al. 2019 AHA/ACC/HRS focused update of the 2014 AHA/ACC/HRS guideline for the management of patients with atrial fibrillation: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines and the Heart Rhythm Society. J Am Coll Cardiol 2019;74:104-32. Crossref

24. Levey AS, Inker LA, Matsushita K, et al. GFR decline as an end point for clinical trials in CKD: a scientific workshop sponsored by the national kidney foundation and the US Food and Drug Administration. Am J Kidney Dis 2014;64:821-35. Crossref

25. Levin A, Stevens PE, Bilous RW, et al. Kidney disease: improving global outcomes (KDIGO) CKD work group. KDIGO 2012 clinical practice guideline for the evaluation and management of chronic kidney disease. Kidney Int Suppl 2013;3:1-150.

26. Lau YC, Proietti M, Guiducci E, Blann AD, Lip GY. Atrial fibrillation and thromboembolism in patients with chronic kidney disease. J Am Coll Cardiol 2016;68:1452-64. Crossref

27. Levey AS, Inker LA, Coresh J. GFR estimation: from physiology to public health. Am J Kidney Dis 2014;63:820-34. Crossref

28. Lam AS, Lee IM, Mak SK, Yan BP, Lee VW. Warfarin control in Hong Kong clinical practice: a single-centre observational study. Hong Kong Med J 2020;26:294-303. Crossref

29. 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

30. Larsen TB, Skjøth F, Nielsen PB, Kjældgaard JN, Lip GY. Comparative effectiveness and safety of non-vitamin K antagonist oral anticoagulants and warfarin in patients with atrial fibrillation: propensity weighted nationwide cohort study. BMJ 2016;353:i3189. Crossref

31. McCaffrey DF, Griffin BA, Almirall D, Slaughter ME, Ramchand R, Burgette LF. A tutorial on propensity score estimation for multiple treatments using generalized boosted models. Stat Med 2013;32:3388-414. Crossref

32. Austin PC. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav Res 2011;46:399-424. Crossref

33. Austin PC, Stuart EA. Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies. Stat Med 2015;34:3661-79. Crossref

34. Chatrou ML, Winckers K, Hackeng TM, Reutelingsperger CP, Schurgers LJ. Vascular calcification: the price to pay for anticoagulation therapy with vitamin K-antagonists. Blood Rev 2012;26:155-66. Crossref

35. Schurgers LJ, Joosen IA, Laufer EM, et al. Vitamin K-antagonists accelerate atherosclerotic calcification and induce a vulnerable plaque phenotype. PLoS One 2012;7:e43229. Crossref

36. Ozcan A, Ware K, Calomeni E, et al. 5/6 nephrectomy as a validated rat model mimicking human warfarin-related nephropathy. Am J Nephrol 2012;35:356-64. Crossref

37. Ware K, Brodsky P, Satoskar AA, et al. Warfarin-related nephropathy modeled by nephron reduction and excessive anticoagulation. J Am Soc Nephrol 2011;22:1856-62. Crossref

38. Hernandez AV, Bradley G, Khan M, et al. Rivaroxaban vs. warfarin and renal outcomes in non-valvular atrial fibrillation patients with diabetes. Eur Hear J Qual Care Clin Outcomes 2020;6:301-7. Crossref

39. Hijazi Z, Hohnloser SH, Andersson U, et al. Efficacy and safety of apixaban compared with warfarin in patients with atrial fibrillation in relation to renal function over time: insights from the ARISTOTLE randomized clinical trial. JAMA Cardiol 2016;1:451-60. Crossref

40. Yao X, Shah ND, Sangaralingham LR, Gersh BJ, Noseworthy PA. Non–vitamin K antagonist oral anticoagulant dosing in patients with atrial fibrillation and renal dysfunction. J Am Coll Cardiol 2017;69:2779-90. Crossref

41. Coresh J, Turin TC, Matsushita K, et al. Decline in estimated glomerular filtration rate and subsequent risk of end-stage renal disease and mortality. JAMA 2014;311:2518-31. Crossref

42. Caldeira D, Gonçalves N, Pinto FJ, Costa J, Ferreira JJ. Risk of renal failure with the non-vitamin K antagonist oral anticoagulants: systematic review and meta-analysis. Pharmacoepidemiol Drug Saf 2015;24:757-64. Crossref

43. Roldán V, Marín F, Fernández H, et al. Renal impairment in a ‘real-life’ cohort of anticoagulated patients with atrial fibrillation (implications for thromboembolism and bleeding). Am J Cardiol 2013;111:1159-64. Crossref

44. 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

45. Chan YH, See LC, Tu HT, et al. Efficacy and safety of apixaban, dabigatran, rivaroxaban, and warfarin in Asians with nonvalvular atrial fibrillation. J Am Heart Assoc 2018;7:e008150. Crossref

46. Zhang J, Tang J, Cui X, et al. Indirect comparison of novel oral anticoagulants among Asians with non-valvular atrial fibrillation in the real world setting: a network meta-analysis. BMC Cardiovasc Disord 2019;19:182. Crossref

47. Liu X, Huang M, Ye C, Zeng J, Zeng C, Ma J. The role of non-vitamin K antagonist oral anticoagulants in Asian patients with atrial fibrillation: a PRISMA-compliant article. Medicine 2020;99:e21025. Crossref

It has been projected that, by year 2036, one in three people in Hong Kong will be aged ≥65 years.1 Because the life expectancy of the Hong Kong population consistently ranks among the highest worldwide, the medical and social needs of older individuals are expected to increase rapidly in the next few decades. The waiting time for joint arthroplasty in Hong Kong’s Hospital Authority (HA) system increased from 33 months in 20102 to 50 months in 20193, reflecting increasing demand for such procedures in our population.

Total knee arthroplasty (TKA) is a highly successful operation that provides substantial pain relief and functional improvement for patients with end-stage knee arthritis.4 However, in Hong Kong, there is the prevalent belief among many members of the community that older patients, particularly those aged ≥80 years, have a substantial risk of mortality after TKA. Therefore, despite substantial pain and debilitation, older patients have often avoided TKA as treatment for their knee arthritis. However, this avoidance may no longer be a reasonable approach in the era of modern arthroplasty.

This study examined the age of patients who underwent TKA in the past 10 years in Hong Kong. The aim of the study was to determine the mortality safety profile and clinical outcomes after TKA in patients aged ≥80 years at the time of operation.

All patients who underwent primary TKA between 2010 and 2019 in public hospitals operated by the HA were included in the study. Patient data were extracted from the HA Clinical Data Analyses and Reporting System. Baseline demographic characteristics (eg, age, sex, and diagnosis at the time of operation) were recorded. Patients were stratified according to age (<80 or ≥80 years) at the time of the operation. Incidences of 30-day, 90-day, and 1-year mortality after TKA, as well as the incidence of emergency readmission within 28 days after TKA, were calculated.

Furthermore, all patients who had undergone TKA between 2010 and 2019 in the HA’s Hong Kong West Cluster of hospitals, who were aged ≥80 years at the time of operation, were identified for inclusion in the study. The Hong Kong West Cluster comprises seven hospitals, providing a total of 3142 beds.3

Baseline preoperative parameters, including the Charlson Comorbidity Index (CCI),5 as well as the Knee Society Knee Score (KSKS) and Knee Society Knee Functional Assessment (KSFA) scores,6 7 were recorded. Rehabilitation outcomes, including KSKS and KSFA scores, were recorded at 1 year after surgery and at the latest follow-up.

For data collection and statistical analyses, SPSS (Windows version 26) was used. Age, CCI score, KSKS, and KSFA score were all non-normally distributed, according to the Kolmogorov–Smirnov test. Therefore, the Mann-Whitney U test was used to compare the median ages of patients during 2010-2014 and 2015-2019; to compare the median CCI score between older patients who died within 1 year of the operation and patients who survived; and to compare KSKS and KSFA scores before surgery and at 1 year after surgery.

Between 2010 and 2019, 25 040 TKA procedures were conducted in all HA hospitals. During the same period, 3835 TKA procedures were conducted in the authors’ hospital cluster. An increasing trend was observed in the number of TKA procedures each year; the yearly caseload in 2019 was more than double the yearly caseload in 2010 (Table, Fig).

For all HA hospitals, the mean age (± standard deviation) at the time of operation throughout the study period was 69.4 ± 7.7 years (range, 18-94). Median age at the time of operation significantly increased from 69 years (interquartile range [IQR], 58-80) in 2010-2014 to 70 years (IQR, 59-81) in 2015-2019 (P<0.001).

An increase in the proportion of patients aged ≥80 years at the time of operation was observed throughout the study period (Table, Fig).

Between 2010 and 2019, 2491 TKA procedures were conducted in all HA hospitals in patients aged ≥80 years. The median age at operation was 82 years (IQR, 80-84). Mortality rates within 30 and 90 days after surgery were 0.156% and 0.35%, respectively. The mortality rate within 1 year after surgery was 1.09%. In total, 5.3% of patients required emergency readmission within 28 days of TKA.

During the study period, 22 549 TKA procedures were conducted in all HA hospitals in patients aged <80 years. The mean age (± standard deviation) at operation was 67.9 ± 6.8 years (range, 14-79). Mortality rates at 30 days, 90 days, and 1 year after surgery were 0.047%, 0.128%, and 0.441%, respectively. In total, 4.02% of patients required emergency readmission within 28 days after surgery.

Between 2010 and 2019, 574 TKA procedures were conducted in patients aged ≥80 years at the time of operation in the authors’ HA hospital cluster. The mean age (± standard deviation) at operation was 82.9 ± 2.6 years (range, 80-93).

The mean CCI score (± standard deviation) at the time of operation was 4 ± 1.1 (range, 4-11). The CCI score was not significantly different between older patients who died within 1 year after surgery and older patients who survived beyond this time period (median: 5 vs 4; P=0.565).

The median KSKS improved from 45 before surgery to 94 at 1 year after surgery (P<0.001). The median KSFA scores improved from 45 before surgery to 55 at 1 year after surgery (P<0.001).

In the past decade in Hong Kong, the proportion of patients aged ≥80 years at the time of operation increased from 8.5% in 2010 to 11.1% in 2019. Moreover, the proportions of patients aged ≥80 years at the time of operation were 9.3% and 10.3% during 2010-2014 and 2015-2019, respectively. These proportions are higher than the values reported by Yan et al,8 who examined the demographics of TKA usage in the preceding decade (ie, 2000-2009) in Hong Kong.

In 2019 in Hong Kong, the mean age at the time of operation was 69.9 years; this was similar to the mean ages from studies in the United Kingdom (69 years)9 and Australia (68.5 years)10. Furthermore, the mean age at the time of the operation increased throughout the study period among all patients undergoing TKA in the HA system. This pattern has also been observed in Taiwan,11 but it contrasts with the findings in the United States and Canada, where overall decreases in mean age have been observed.12

Overall, the mortality rates among older patients in the present study were similar to those reported in other countries.13 14 However, the mortality rates among older patients undergoing TKA in the present study are higher than the mortality rates reported among patients of all ages undergoing TKA in the HA15 (0.1%, 0.2% and 0.7% at 30 days, 90 days and 1 year, respectively) and in other localities (0.18% within 30 days after surgery).16 Although the mean age of the overall patient cohort was not reported in the study by Lee et al,15 the mean ages of the mortality and non-mortality groups were 78 and 64 years, respectively. These are both substantially lower than the mean age at operation for the older patient group in the present study (82.8 years). Although mortality rates at 30 days, 90 days, 1 year and 5 years after surgery differed between the present study and the study by Lee et al,15 the values were generally comparable.

The differences in mortality risk between older and younger patients in the present study are consistent with the findings in a meta-analysis by Kuperman et al,17 who reported increased mortality in older patients after TKA. This is likely related to underlying differences in the inherent mortality risks between older and younger patients caused by age-related increases in the number of medical co-morbidities. In Hong Kong in 2018, the age-specific mortality rates for the general population aged 80 to 84 years were 5.5% and 3.2% for men and women, respectively.18 In the present study, the overall mortality rate within 1 year after surgery for the older patient cohort was 1.09%, substantially lower than the overall mortality rate of the older general population in Hong Kong. This likely reflects the stringent preoperative screening protocol used for older TKA candidates, which is intended to minimise the risk of mortality.13 15 Therefore, with careful perioperative assessment and management, mortality risk after TKA can be minimised, even in older patients.

Finally, the present study revealed that both KSKS and KSFA scores were significantly improved at 1 year after TKA, compared with scores before surgery. Therefore, pain, objective physical examination findings, and function in terms of walking and ability to climb stairs can be significantly improved after TKA, even in older patients. Our results support the findings in previous literature.19 20

An important limitation of the retrospective study design was that it did not allow us to control for confounding variables, such as differences in surgical expertise and standards of perioperative care among the centres included in this study. However, to our knowledge, this is the only study regarding the incidence of mortality after TKA among older patients in Hong Kong. The results of this study are important for our locality because they describe TKA outcomes from all HA hospitals in the past 10 years in a large cohort of patients.

In conclusion, this study showed that the mean age at the time of TKA has steadily risen in the past 10 years, consistent with population ageing trends in Hong Kong. Furthermore, the findings indicate that TKA is a safe and efficacious procedure, even in older patients. Therefore, provided that proper perioperative assessment and management are conducted, advanced age should not be a deterrent for TKA in the management of end-stage knee arthritis among older patients who can substantially benefit from this procedure.

Concept or design: A Cheung, CH Yan, KY Chiu.
Acquisition of data: A Cheung, PK Chan, H Fu, VWK Chan, MH Cheung.
Analysis or interpretation of data: A Cheung, PK Chan, H Fu, VWK Chan, MH Cheung.
Drafting of the manuscript: A Cheung, PK Chan, KY Chiu.
Critical revision of the manuscript for important intellectual content: A Cheung, CH Yan, KY Chiu.

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.

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

This research has received approval from the Institutional Review Board of The University of Hong Kong/Hospital Authority Hong Kong West Cluster (Ref UW 20-161).

1. Census and Statistics Department, Hong Kong SAR Government. Hong Kong population projections for 2017 to 2066. Available from: https://www.censtatd.gov.hk/ media_workers_corner/pc_rm/hkpp2017_2066/index.jsp. Accessed 19 Mar 2020.

2. Bureau Food and Health Bureau. Legislative Council Panel on Health Services Improvement on Joint Replacement Surgeries in the Hospital Authority. 2011. Available from: https://www.legco.gov.hk/yr10-11/english/panels/hs/papers/hs0613cb2-1992-3-e.pdf. Accessed 20 Mar 2020.

3. Hospital Authority website. Available from: https://www.ha.org.hk/visitor/ha_visitor_text_index.asp?Content_ID=221223&Lang=ENG&Dimension=100. Accessed on 24 Sep 2021.

4. Pavone V, Boettner F, Fickert S, Sculco TP. Total condylar knee arthroplasty: a long-term followup. Clin Orthop Relat Res 2001(388):18-25. Crossref

5. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40:373-83. Crossref

6. Insall JN, Dorr LD, Scott RD, Scott WN. Rationale of the Knee Society clinical rating system. Clin Orthop Relat Res 1989;(248):13-4. Crossref

7. Lingard EA, Katz JN, Wright RJ, Wright EA, Sledge CB, Kinemax Outcomes Group. Validity and responsiveness of the Knee Society Clinical Rating System in comparison with the SF-36 and WOMAC. J Bone Joint Surg Am 2001;83:1856-64. Crossref

8. Yan CH, Chiu KY, Ng FY. Total knee arthroplasty for primary knee osteoarthritis: changing pattern over the past 10 years. Hong Kong Med J 2011;17:20-5.

9. National Joint Registry. National Joint Registry 16th Annual Report 2019. Available from: https://reports.njrcentre.org.uk/portals/0/pdfdownloads/njr%2016th%20annual%20report%202019.pdf. Accessed 4 Mar 2020.

10. Australian Orthopaedic Association National Joint Replacement Registry (AOANJRR). Hip, Knee & Shoulder Arthroplasty: 2020 Annual Report. Adelaide: AOA; 2020: 1-474. Available from: https://aoanjrr.sahmri.com/annual-reports-2020. Accessed 21 Sep 2021.

11. Lin FH, Chen HC, Lin C, et al. The increase in total knee replacement surgery in Taiwan: a 15-year retrospective study. Medicine (Baltimore) 2018;97:e11749. Crossref

12. Ravi B, Croxford R, Reichmann WM, Losina E, Katz JN, Hawker GA. The changing demographics of total joint arthroplasty recipients in the United States and Ontario from 2001 to 2007. Best Pract Res Clin Rheumatol 2012;26:637-47. Crossref

13. Hunt LP, Ben-Shlomo Y, Clark EM, et al. 45-Day mortality after 467,779 knee replacements for osteoarthritis from the National Joint Registry for England and Wales: an observational study. Lancet 2014;384:1429-36. Crossref

14. Clement ND, MacDonald D, Howie CR, Biant LC. The outcome of primary total hip and knee arthroplasty in patients aged 80 years or more. J Bone Joint Surg Br 2011;93:1265-70. Crossref

15. Lee QJ, Mak WP, Wong YC. Mortality following primary total knee replacement in public hospitals in Hong Kong. Hong Kong Med J 2016;22:237-41. Crossref

16. Belmont PJ Jr, Goodman GP, Waterman BR, Bader JO, Schoenfeld AJ. Thirty-day postoperative complications and mortality following total knee arthroplasty: incidence and risk factors among a national sample of 15,321 patients. J Bone Joint Surg Am 2014;96:20-6. Crossref

17. Kuperman EF, Schweizer M, Joy P, Gu X, Fang MM. The effects of advanced age on primary total knee arthroplasty: a meta-analysis and systematic review. BMC Geriatr 2016;16:41. Crossref

18. Census and Statistics Department, Hong Kong SAR. Hong Kong Monthly Digest of Statistics: The mortality trend in Hong Kong, 1986 to 2018. 2019. Available from: https://www.statistics.gov.hk/pub/B71911FB2019XXXXB0100.pdf. Accessed 20 Sep 2021.

19. Kennedy JW, Johnston L, Cochrane L, Boscainos PJ. Total knee arthroplasty in the elderly: does age affect pain, function or complications? Clin Orthop Relat Res 2013;471:1964-9. Crossref

20. Williams DP, Price AJ, Beard DJ, et al. The effects of age on patient-reported outcome measures in total knee replacements. Bone Joint J 2013;95-B:38-44. Crossref