Risk factors for losing hepatitis B virus surface antibody in patients with HBV surface antigen negative/surface antibody positive serostatus receiving biologic disease-modifying anti-rheumatic drugs: a nested case-control study

Hepatitis B virus (HBV) reactivation consequent to immunosuppressive therapy is an increasingly prevalent problem with serious clinical implications. Treatment with biologic agents conduces to the loss of protective antibody to HBV surface antigen (anti-HBs), which significantly increases the risk of HBV reactivation. Hence, we investigated the risk factors for losing anti-HBs in patients with rheumatic diseases and HBV surface antigen negative/anti-HBs positive (HBsAg−/anti-HBs+) serostatus during treatment with biologic disease-modifying anti-rheumatic drugs (DMARDs). Using a nested case-control design, we prospectively enrolled patients with rheumatoid arthritis, ankylosing spondylitis, psoriatic arthritis/psoriasis, or juvenile idiopathic arthritis, who were treated with biologic DMARDs at Changhua Christian Hospital, Taiwan, from January 2013 to June 2019 and had HBsAg−/anti-HBs+ serostatus; the analytic sample excluded all patients with HBsAg+ or anti-HBs− serostatus. Anti-HBs titers were monitored 6-monthly and cases were defined as anti-HBs < 10 mIU/ml during follow-up. Cases were matched one-to-all with controls with anti-HBs ≥ 10 mIU/ml on the same ascertainment date and equivalent durations of biologic DMARDs treatment (control patients could be resampled and could also become cases during follow-up). Between-group characteristics were compared and risk factors for anti-HBs loss were investigated by conditional logistic regression analyses. Among 294 eligible patients, 23 cases were matched with 311 controls. The incidence of anti-HBs loss was ~ 2.7%/person-year during biologic DMARDs treatment. Besides lower baseline anti-HBs titer (risk ratio 0.93, 95% CI 0.89–0.97), cases were significantly more likely than controls to have diabetes mellitus (risk ratio 4.76, 95% CI 1.48–15.30) and chronic kidney disease (risk ratio 14.00, 95% CI 2.22–88.23) in univariate analysis. Risk factors remaining significantly associated with anti-HBs loss in multivariate analysis were lower baseline anti-HBs titer (adjusted risk ratio 0.93, 95% CI 0.88–0.97) and chronic kidney disease (adjusted risk ratio 45.68, 95% CI 2.39–871.5). Besides lower baseline anti-HBs titer, chronic kidney disease also strongly predicts future anti-HBs negativity in patients with HBsAg−/anti-HBs+ serostatus who receive biologic DMARDs to treat rheumatic diseases. Patients with low anti-HBs titer (≤ 100 mIU/ml) and/or chronic kidney disease should be monitored during biologic DMARDs therapy, to enable timely prophylaxis to preempt potential HBV reactivation.

Although serum HBV DNA is a defining characteristic of HBV reactivation [1][2][3][4], it does not inevitably progress to morbid viremia but may manifest transiently without symptoms, especially whilst anti-HBs serostatus remains positive [5,6,[10][11][12]. On the other hand, anti-HBs loss consequent to immunosuppressive treatment of patients with HBsAg−/anti-HBc+ serostatus significantly increases the risk of progression from clinically silent to symptomatic HBV reactivation [9,10,12,13]. Previous studies have shown that anti-HBs can decline to seronegative during immunosuppressive therapies, especially in patients with a low baseline anti-HBs titer [11,14], but none have systematically investigated whether there are other predisposing factors.
In this context, especially given burgeoning use of tumor necrosis factor inhibitors (anti-TNF) and other biologic disease-modifying anti-rheumatic drugs (DMARDs) to treat autoimmune diseases, and elevated HBV reactivation rates in this setting [2,4,5], it is imperative to further elucidate risk factors for anti-HBs loss. To this end, we prospectively studied patients with HBsAg−/anti-HBc+ serostatus during/after biologic DMARDs therapy for rheumatic diseases.

Study subjects
The study sample comprised patients at Changhua Christian Hospital, Taiwan, with rheumatoid arthritis, ankylosing spondylitis, psoriasis, psoriatic arthritis, or juvenile idiopathic arthritis, who were treated with biologic DMARDs from January 2013 to June 2019. Only patients with HBsAg−/anti-HBs+ serostatus were included and all with HBsAg+ or HBsAg−/anti-HBs− serostatus were excluded (Fig. 1). All study subjects fulfilled international diagnostic criteria for these diseases and were treated according to Taiwan Rheumatology Association guidelines for screening and management of viral hepatitis [15].

Hepatitis B serologic testing and HBV DNA
HBV serology and DNA assays were done every 6 months according to Taiwan Rheumatology Association recommendations [15]. HBV assays included serum HBsAg, anti-HBs and anti-HBc, measured by Architect i2000SR chemiluminescent microparticle immunoassay (Abbott Laboratories, Abbot Park, Illinois, USA). HBV immunization history of people with anti-HBs+/anti-HBc− serostatus was not ascertained.
Anti-HBs titer < 10 mIU/ml was defined as seronegative. Low anti-HBs titer was defined as 10-100 mIU/ml, based on evidence of significantly increased likelihood of anti-HBs loss and detectable HBV DNA at anti-HBs titers below 100 mIU/ml and protection against HBV reactivation above this threshold [11,14]. Serum HBV DNA viral load was quantified by Abbott RealTime HBV (Abbott Laboratories, Abbott Park, Illinois, USA), with a minimal sensitivity of 10 IU/ml. HBV DNA titer ≥ 10 mIU/ml was defined as detectable viral load [11], while the criteria defining clinical HBV reactivation at any serial 6-monthly follow-up check, were HBV replication ≥ 2 log increase from baseline or a new appearance of HBV DNA to ≥ 100 IU/ml in people with previously stable or undetectable levels [1].

Nested case-control design
We used a nested case-control design, which enables efficient analysis of time-dependent exposures on rare outcomes where only a limited sample from a larger population is practical, without compromising statistical power [17][18][19]. Unlike conventional cohort studies, which compare cases versus controls from a fully enumerated population, a nested case-control design identifies occurrences of events of interest in a defined sub-population and matches these with a specified number of control subjects drawn from the same sub-sample, but who were not yet affected by the same event when it occurred in their corresponding case [18,19]. This design means that controls can become cases later during follow-up and that each patient may serve as a control repeatedly (though at different times); thus, cases are compared with controls from the same patient sample, which lessens the likelihood of selection bias [17][18][19][20].
The first prescription of a biologic DMARD defined the start point. Cases were defined upon occurrences of serum anti-HBs titer < 10 mIU/ml during follow-up, with the anti-HBs, HBV surface antibody; HBsAg, HBV surface-antigen; mIU, million International Units. a One patient could serve as a control repeatedly during follow-up and control subjects could become cases during the study date when anti-HBs loss was ascertained designated the event date. Each case was matched one-to-all with subjects whose serum HBsAb titer was ≥ 10 mIU/ml on the respective case ascertainment date and who had an equivalent duration of biologic DMARDs treatment.

Data analysis and statistics
All analyses were done using nonparametric statistical software (LogXact 11, Cytel Software Corp, Cambridge, MA, USA) with penalized maximum likelihood to remove first-order bias. A p-value < 0.05 for two-sided tests was considered statistically significant. Continuous variables were expressed as means plus/minus standard deviation or mean [range], categorical variables as numbers (percentages). Conditional logistic regression analysis was used to estimate risk ratios and 95% confidence intervals for loss of anti-HBs; putative associated factors included age, sex, type of rheumatic disease, conventional DMARDs, biologic DMARDs (anti-TNF or others), comorbidity, and baseline anti-HBs titer.

Demographic characteristics and clinical status
The analytic samples drawn from 294 patients with HBsAg−/anti-HBs+ serostatus at baseline, comprised 23 cases and 311 matched controls (Fig. 1); Table 1 shows their demographic and clinical characteristics. Mean age and rheumatic disease types were similar between case and control groups. No patients with HBsAg−/anti-HBs+ serostatus had detectable HBV DNA at enrolment. Compared with controls, cases had lower baseline serum anti-HBs titers, more prevalent comorbidities (including hepatitis C virus infection, chronic liver disease, diabetes mellitus, chronic kidney disease), and relatively higher accumulated doses of sulfasalazine, leflunomide, and prednisolone. Most people in both groups used anti-TNF agents (etanercept, adalimumab, golimumab). No study subjects were kidney transplant recipients.

Incidence of anti-HBs loss and associated risk factors
The incidence rate of anti-HBs loss in 294 patients with HBsAg−/anti-HBs+ serostatus during biologic DMARDs treatment was 23/852 person-years:~2.7%/person-year. Table 2 shows risk factors associated with loss of anti-HBs in univariate and multivariate conditional logistic regression analyses. Besides lower baseline anti-HBs titer (risk ratio 0.93, 95% CI 0.89-0.97), cases were significantly more likely than controls to have diabetes mellitus (risk ratio 4.76, 95% CI 1.48-15.30) and chronic kidney disease (risk ratio 14.00, 95% CI 2.22-88.23) in the univariate analysis. However, the only factors remaining significant in the multivariate model, were lower baseline serum anti-HBs titer (adjusted risk ratio 0.93, 95% CI 0.88-0.97) and chronic kidney disease (adjusted risk ratio 45.68, 95% CI 2.39-871.5).

Clinical features and outcomes of subjects with anti-HBs loss
Thirteen cases had rheumatoid arthritis (Table 3). All cases' baseline anti-HBs titers were low (≤ 100 mIU/ml), mean 22.6. Fourteen were prescribed anti-TNF agents: four etanercept, six adalimumab, and four golimumab. Three were prescribed tofacitinib. Two cases each were prescribed ustekinumab or tocilizumab, while one each received abatacept or rituximab.
No cases (nor anti-HBs+ controls) had clinical HBV reactivation during follow-up (852 person-years), and no cases developed alanine transaminase elevation, or received any anti-viral treatment during median follow-up of 30 months (range 0-77) after anti-HBs loss. Only one of the 16/23 cases whose serum HBV DNA was monitored after anti-HBs loss ever had a detectable viral load (Table 3), which was observed only once, with no recurrence as of August 2020.

Discussion
We believe this to be the first report of risk factors associated with loss of anti-HBs in rheumatic patients during biologic DMARDs therapy, after controlling for putative risk factors. We discovered that besides lower baseline anti-HBs titer, chronic kidney disease independently predicts anti-HBs loss.
Our finding that lower pretreatment anti-HBs titer (≤ 100 mIU/ml) is a risk factor for loss of anti-HBs, is consistent with a study of rituximab-based therapy for lymphoma [14]. Moreover, baseline anti-HBs positivity was protective against HBV reactivation among patients with HBsAg−/anti-HBc+ serostatus who received immunosuppressive or biologic agents [9][10][11][12][13][14]. Therefore, our results suggest that anti-HBs may be lost during/after immunosuppressive/biologic therapy, especially in people with a low baseline titer, with consequently elevated risk of HBV reactivation.
Although we detected no cases of morbid viremia and just one occurrence of minimal HBV DNA during follow-up, this was not necessarily inconsistent with evidence that anti-HBs negativity increases the risk of HBV reactivation in patients with HBsAg−/anti-HBc+ serostatus. Anti-HBs may remain persistently low or negative during treatment, with HBV DNA detected periodically but without progression to morbid reactivation [5,6,11,12]. Reported rates of HBV DNA manifestation during/ after immunosuppressive therapy in patients with HBsAg−/anti-HBs−/anti-HBc+ serostatus are in the order of 1-10%, with symptomatic reactivation in a smaller fraction of cases [9][10][11][12]; therefore, 23 cases may have been too few to detect occasional reactivation events. The follow-up duration (median 30 months after anti-HBs loss) may also have been insufficient. HBV DNA in such patients usually appears late, after several cycles of therapy have diminished anti-HBs to undetectable levels [6], and clinical reactivation may not occur until several years since commencing immunosuppressive More than five-fold upper reference of 40 IU/L therapy [10][11][12]; median time from starting immunosuppressive therapy to HBV reactivation in a cohort of 1042 rheumatic disease patients with resolved HBV infections was 66 months [12]. However, 18/23 cases in our study had total follow-up of < 66 months. This is the first indication of which we know that chronic kidney disease might be a risk factor for loss of anti-HBs in patients treated with biologic DMARDs. This is an important contemporary issue because chronic kidney disease is prevalent among patients with rheumatic diseases, consequent to older age, diabetes-related nephropathy, and widespread use of nephrotoxic medications such as non-steroidal anti-inflammatory drugs or cyclosporine. Studies have shown that patients with chronic kidney disease lose anti-HBs faster than healthy subjects do [21,22]; anti-HBs loss in chronic kidney disease or dialysis patients has been attributed to diminished interleukin-2 secretion, impaired macrophage function, decreasing memory B cell counts, and a weak amnestic response [23][24][25].
Consistent with reports of increased likelihood of anti-HBs loss in patients with diabetes mellitus [26,27], we   found a significant association in univariate analysis; however, statistical significance was lost in multivariate analysis, probably because only 4/23 cases had diabetes. Previous guidelines or reviews have propounded baseline anti-HBs screening prior to using biologic DMARDs, because patients with baseline anti-HBs− serostatus have higher risk of HBV reactivation [1,5,7,8]. However, current guidelines, particularly those focused on biologic DMARDs users, neither describe nor elucidate the potential risk of anti-HBs loss during biologic DMARDs therapy [2,5,8]. Our results imply that there is a window of opportunity to prevent morbid HBV reactivation in patients at increased risk. We contend that clinicians should closely monitor patients with low baseline anti-HBs titer (≤ 100 mIU/ml) and/or chronic kidney disease during subsequent biologic DMARDs therapy, including followup of anti-HBs and HBV DNA titers upon anti-HBs loss, to enable timely intervention with appropriate prophylaxis to preempt potential HBV reactivation. Expert opinion supports this approach; for example, advocating HBV follow-up and immunization to reduce the risk of reactivation during anti-TNF treatment [4].
This study had limitations, foremost the small sample size. With few cases or control group patients with chronic kidney disease, the result of multivariate analysis could reflect over-fitting, as the wide confidence interval suggests; larger-scale studies are warranted to corroborate this novel but tentative finding. We acknowledge that direct serum HBV DNA assays are the ideal way to monitor patients at risk of HBV reactivation; nevertheless, based on our findings, monitoring patients with risk factors for anti-HBs loss may be a convenient and costeffective way of targeting hepatitis B prevention, especially in HBV-endemic regions. Admittedly, only direct monitoring can detect HBV reactivation due to immuneescape HBsAg mutations in anti-HBs+ biologic DMARDs recipients; however, this is a very rare phenomenon [28,29] and, excepting such cases, detectable HBV DNA loads in anti-HBs+ patients otherwise occur only sporadically, and are self-limiting and clinically benign [11]. Despite considerable research into whether or not the risks of anti-HBs loss or HBV reactivation differ between biologic DMARDs, results to date have been inconclusive [30,31]; with only 23 cases, we were unable to ascertain whether individual biologic DMARDs carried similar risks of anti-HBs loss.

Conclusions
This prospective single-center study found that lower baseline anti-HBs titer (≤ 100 mIU/ml) and chronic kidney disease strongly predicted loss of anti-HBs in patients having biologic DMARDs therapy for rheumatic diseases. These insights can be applied to identify patients at increased risk of becoming anti-HBs− and monitor them for potential HBV reactivation from the onset of biologic DMARDs therapy. More research is needed to elucidate other risk factors for loss of anti-HBs and so refine the monitoring strategy to prevent HBV reactivation in patients receiving biologic DMARDs to treat rheumatic diseases.