Risk/benefit trade-offs in rheumatology: rofecoxib revisited in the era of JAK inhibitors

In 2004, rofecoxib was withdrawn from the market, after a large trial for polyp prevention (the APPROVE trial) was terminated early due to an increased risk of major adverse cardiovascular events (MACE) in people taking rofecoxib compared with placebo [1]. Shortly after, an ‘expression of concern’ was published by the New England Journal of Medicine [2] over reporting of MACE in the VIGOR trial that compared rofecoxib vs naproxen in people with rheumatoid arthritis [3]. In the VIGOR trial, three myocardial infarctions that occurred in the rofecoxib arm shortly after the study window closed were not included in the original publication. When these additional events were included, the relative risk of myocardial infarction increased from 4.25 (95% CI: 1.39, 17.4) to 5.00 (95% CI: 1.68, 20.1). Additional concerns regarding the presentation of thromboembolic risks were raised [2, 4]. The risks of MACE and concerns over reporting led to thousands of lawsuits. In 2007, a $4.85 billion settlement was reached to end these lawsuits, the largest drug settlement ever at the time. While the decision to withdraw rofecoxib, taken voluntarily by Merck, occurred in the context of pending lawsuits, was it the right decision for patients? Here, comparisons can be made to tofacitinib. Recently, the ORAL Surveillance trial compared tofacitinib to TNF inhibitors (adalimumab or etanercept) in people with rheumatoid arthritis with at least one CV risk factor [5]. ORAL Surveillance failed to find non-inferiority for tofacitinib for both MACE and malignancies, which led to a black box warning for all Janus kinase (JAK) inhibitors, and limitation of its use to people who have not responded or cannot tolerate TNF inhibitors. When viewed as absolute risks, tofacitinib 5 mg BID was associated with 0.2 more MACE events per 100 patient-years than TNF inhibitors, compared with 0.97 more events per 100 patient-years for rofecoxib vs naproxen (Fig. 1). Tofacitinib 5 mg BID was also shown to have a statistically significant increased risk of cancer compared with TNF inhibitors (0.36 more events per 100 patient-years) (Fig. 1). In terms of benefits, both rofecoxib and tofacitinib were similarly effective to their comparator for pain (rofecoxib) and disease control (tofacitinib). Rofecoxib had an added benefit of lowering the risk of major adverse GI events compared with naproxen (2.4 fewer events per 100 patient-years) (Fig. 1). These comparisons, while illustrative, should also consider the difference in trial design between ORAL Surveillance and VIGOR. Oral Surveillance was a non-inferiority safety trial, designed to maximize the detection of MACE and cancers, the co-primary endpoints. It is possible that the observed risks with rofecoxib may have been higher if the trial had a similar design. Comparison of selected risks of tofacitinib vs TNF inhibitors in the ORAL Surveillance trial with rofecoxib vs naproxen in the VIGOR trial. NMSC: non-melanoma skin cancer Following the withdrawal of rofecoxib, people who were taking it to treat their pain had to find alternative approaches. A study from US prescriptions data for people with musculoskeletal disorders found an increase in opioid prescriptions that correlated with a marked drop in non-opioid analgesics in 2005 when rofecoxib was withdrawn from the market [6]. Anecdotally, many people did not achieve the same response with celecoxib or other approaches, and people hoarded supplies of rofecoxib until it ran out. These people had almost certainly heard of the risks of rofecoxib but wanted it anyway. In contrast, JAK inhibitors continue to be a treatment option for people. People may prefer a JAK inhibitor if they have a strong preference for oral therapy, are at low absolute risk of cardiovascular events, or have not responded to other biologic therapy. Before JAK inhibitors, these people may have had to use prednisone and/or NSAIDs, or live with untreated pain, fatigue and other symptoms that come with suboptimally treated RA, which is also associated with an increased risk of MACE. At the time of withdrawal of rofecoxib, an Expert Advisory Committee in Canada recommended that it should be allowed back on the market, citing its efficacy, the low absolute risks of cardiovascular events, and that ‘patients benefit from having a variety of drugs to choose from’ [7]. In contrast, Dr François Bertrand, executive director of medical research for Merck Frosst Canada, commented after the withdrawal of the drug that ‘because of the nature of the events and the availability of other drugs, we decided the right thing was to discontinue [the drug]’ [8]. Although this later statement was perhaps motivated by the pending lawsuits, these conflicting statements highlight the need to have robust and transparent frameworks when considering risk/benefit trade-offs in the drug approval process. When effective treatments have serious but rare risks, who should decide whether a treatment is worthwhile? Patient-centred drug evaluation frameworks would put people living with the condition at the forefront of informing these risk/benefit decisions. Patient preference information refers to data, from people living with the condition, on the relative importance of different outcomes or attributes relevant to the treatment decision. It differs from patient-reported outcomes (PROs). Patient preferences tell us about the trade-offs that people are willing to make; people can trade-off amongst different PROs and other non-patient reported outcomes (e.g. risks of cardiovascular events). Recently, efforts have been made to further the incorporation of patient preference information into regulatory decisions. The Medical Device and Innovation Consortium in the USA, with funding from the Food and Drug Administration (FDA), has released guidance on the use of patient-preference information in the approval process for drugs and other medical devices [9]. The PREFER initiative in Europe, a private–public partnership, released recommendations last year on how, why and when to incorporate patient preferences in medical product decision-making [10]. While this progress is encouraging, the field is still in its early stages, and applied examples are rare. These risk/benefit frameworks acknowledge that the balance of benefits and risks will vary between people, depending on their disease characteristics and personal preferences. The MDIC framework proposes that a drug should be approved if the balance of benefits and risks favours the intervention, even for a subgroup of people [9]. Revisiting rofecoxib, patient-centred drug approval processes would ask whether there are groups of patients who would accept the cardiovascular risks for the benefits it provides. Preference data collected from patients, could help inform these judgements. Clearly though, if we accept the role for JAK inhibitors as providing an effective treatment option for people with rheumatoid arthritis, then we should reconsider the role of rofecoxib for people with uncontrolled pain, where there is a huge unmet need. Informing patients of these risks and benefits is essential and should be managed through best practices in shared decision-making, tailoring treatment discussions to the risk profile and preferences of patients. Rheumatic diseases have enormous impact of people’s lives and treatment choices require balancing risks and benefits. Listening to people with the condition and recognizing the diversity in people’s needs and preferences throughout decision-making processes regarding drug development and approval will help ensure more choices for patients. No new data were analysed in support of this manuscript. No specific funding was received from any bodies in the public, commercial or not-for-profit sectors to carry out the work described in this article. Disclosure statement: G.S.H. declares no financial conflicts of interest. G.S.H. has a non-financial academic relationship with the lead author of the VIGOR trial (co-authored publications, graduate supervision).

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.