Fracture in clinical studies of tofacitinib in rheumatoid arthritis

Abstract

Background: Preclinical data suggest that tofacitinib would protect bone health in patients with rheumatoid arthritis (RA).

Objective: To assess fracture risk in tofacitinib RA clinical trials.

Design: Post hoc analysis.

Methods: We analyzed pooled data of phase I/II/III and long-term extension studies ('P123LTE cohort'), pooled data of placebo-controlled portions of phase III studies (phase III placebo-controlled cohort), and data from ORAL Surveillance [phase IIIb/IV randomized, open-label trial evaluating tofacitinib 5/10 mg twice daily (BID) vs tumor necrosis factor inhibitor (TNFi) in patients ⩾ 50 years with ⩾ 1 additional cardiovascular risk factor].

Results: In the phase III placebo-controlled cohort, incidence rates (IRs) [95% confidence interval (CI)] of fracture were 2.11 (1.09-3.68), 2.56 (1.23-4.71), and 4.43 (1.78-9.12) per 100 patient-years (PYs) for tofacitinib 5 mg BID, tofacitinib 10 mg BID, and placebo, respectively [tofacitinib 5 mg BID vs placebo: hazard ratio (HR) (95% CI) = 0.55(0.18-1.65); tofacitinib 10 mg BID vs placebo: HR (95% CI) = 0.72 (0.26-2.01)]. In P123LTE, IRs (95% CI) were 2.62 (2.29-2.99) and 2.26 (2.02-2.52) per 100 PY for average tofacitinib 5 and 10 mg BID, respectively. In ORAL Surveillance, IRs (95% CI) were 2.79 (2.34-3.30), 2.87 (2.40-3.40), and 2.27 (1.87-2.74) per 100 PY for tofacitinib 5 mg BID, tofacitinib 10 mg BID, and TNFi, respectively. In ORAL Surveillance, the risk of fracture was numerically higher than TNFi for tofacitinib 5 mg BID [HR (95% CI) = 1.23 (0.96-1.58)] and tofacitinib 10 mg BID [HR (95% CI) = 1.26 (0.97-1.62)]. In ORAL Surveillance, independent predictors of all and osteoporotic fractures with tofacitinib or TNFi included age ⩾ 65, female sex, history of fracture/osteoporosis, and baseline oral corticosteroid use.

Conclusion: This post hoc analysis showed numerically lower fracture risk with tofacitinib versus placebo and numerically greater risk versus TNFi. We did not identify any tofacitinib-specific predictors of fractures, and predictors of fracture were generally aligned with prior literature in the general population and patients with RA. Patients with fracture risk factors should be adequately monitored and treated.

Clinical trial registration: NCT00960440, NCT00847613, NCT00814307, NCT00856544, NCT00853385, NCT01039688, NCT02187055, NCT02831855, NCT00413699, NCT00147498, NCT00413660, NCT00550446, NCT00603512, NCT00687193, NCT00661661, NCT01164579, NCT00976599, NCT01059864, NCT01359150, NCT01262118, NCT01484561, NCT02281552, NCT02147587, NCT02092467.

Keywords: JAK inhibitor; fracture; osteoporosis; rheumatoid arthritistofacitinib.

Figure 1.

IRs (patients with events/100 PY) for all fractures and osteoporotic fractures in (a) phase III placebo-controlled cohort^a,b. (b) P123LTE cohort^b. (c) ORAL Surveillance (⩾ 50 years of age with ⩾ 1 additional CV risk factor at baseline)^c. Exact Poisson (adjusted for PY) 95% CIs are provided for the crude IR. Arrow in Panel A indicates that the 95% CI extends beyond the axis. BID, twice daily; CI, confidence interval; CV, cardiovascular; HR, hazard ratio; IR, incidence rate; n, number of patients with first event within the risk period; N, number of patients within treatment group; PY, patient-year; TNFi, tumor necrosis factor inhibitor. ^aHRs and associated 95% CI were estimated from a Cox regression model including fixed effects of treatment (only the treatment groups involved for the comparison) and study. ^bPY defined as total follow-up time (risk period) calculated up to the day of the first event, subject to an additional period of up to 28 days beyond the last dose. ^cHR (95% CI) based on a simple Cox proportional hazard model with treatment as covariate. PY defined as total follow-up time calculated up to the day of the first event, subject to the defined risk period. The risk period was minimum of (last contact date, last study treatment dose date + 28 days). First events were counted within the risk period. If a patient did not have an event or had an event but outside the risk period, the patient was censored at the end of risk period.

Figure 2.

Predictors for all fractures and osteoporotic fractures from multivariable Cox models via backward selection in (a) P123LTE cohort and (b) ORAL Surveillance. BMI, body mass index; anti-CCP, anti-cyclic citrullinated peptide; CI: confidence interval; COPD, chronic obstructive pulmonary disease; HAQ-DI, Health Assessment Questionnaire–Disability Index; HR, hazard ratio; ILD, interstitial lung disease. *Predictors with p value < 0.05. HRs are shown on a logarithmic scale. Events were counted up to 28 days beyond the last dose. For both cohorts, a backward model selection algorithm was used on a multivariable Cox model which included effects for treatment groups and a set of potential predictors as a starting point, selected based on clinical knowledge and screened via simple Cox analyses (see Tables S2–S5; the p value threshold was < 0.1 for a potential predictor to be included in the multivariable model). The p value cut-off for a predictor to stay in the multivariable model was 0.05. Treatment group was retained regardless of the p value.

Figure 3.

IRs (patients with events/100 PY) for osteoporotic fractures in ORAL Surveillance stratified by independent predictors of fractures. PY defined as total follow-up time calculated up to the day of the first event, subject to the risk period. Within each level of a predictor, HR (95% CI) based on simple Cox proportional hazard model with treatment as covariate. The risk period was minimum of (last contact date, last study treatment dose date + 28 days). First events were counted within the risk period. If a patient did not have an event or had an event but outside the risk period, the patient was censored at the end of risk period. BID, twice daily; anti-CCP, anti-cyclic citrullinated peptide; CI, confidence interval; HR, hazard ratio; IR, incidence rate; PY, patient-year; Ref, reference group; TNFi, tumor necrosis factor inhibitor.