Association Between Biologic Therapies for Chronic Plaque Psoriasis and Cardiovascular Events: Title and subTitle BreakA Meta-analysis of Randomized Controlled Trials FREE

Quiz Ref IDIn the past decade, important new findings have emerged linking autoimmune diseases including rheumatoid arthritis (RA), psoriasis, and Crohn disease with chronic systemic inflammation and a subsequent increase in occlusive vascular disease and cardiovascular risk.^{1 - 3} It has been proposed that control of inflammation could help reduce cardiovascular morbidity. Indeed, a cardioprotective effect has been suggested with systemic agents such as methotrexate and anti–tumor necrosis factor α (TNF-α) agents in RA and psoriasis populations.^{4 - 5}

Given this context, we were concerned about preliminary reports of a numerical excess of major adverse cardiovascular events (MACEs [a composite end point of myocardial infarction, cerebrovascular accident, or cardiovascular death]) in randomized controlled trials (RCTs) of psoriasis patients treated with ustekinumab (Centocor Ortho Biotech Inc, Horsham, Pennsylvania) and briakinumab (Abbott Laboratories, Abbott Park, Illinois)—highly effective monoclonal antibodies to the p40 subunit common to interleukin 12 (IL-12) and IL-23.^{6 - 16} There were 10 MACEs in anti–IL-12/23–treated patients in the placebo-controlled phases of phase 2 and 3 studies of ustekinumab (n = 5) and briakinumab (n = 5) compared with zero events in placebo-treated patients, and a paucity of events reported from studies of anti–TNF-α–treated psoriasis patients with similar disease severity. A total of 53 MACEs have occurred to date across all phases of these studies; 26 MACEs, including 1 cardiovascular death in studies of ustekinumab (additional cases confirmed by Centocor)^{6 - 11} and 27 MACEs, including 4 cardiovascular deaths, in studies of briakinumab (additional cases confirmed by Abbott).^{12 - 16}

A meta-analysis was conducted to evaluate cardiovascular outcomes in studies comparing biologic agents with placebo for the treatment of psoriasis. We reviewed all RCTs of anti–IL-12/23 agents (ustekinumab and briakinumab) and anti–TNF-α agents (adalimumab, etanercept, and infliximab) for the treatment of chronic plaque psoriasis (CPP) and performed a meta-analysis using the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) 2009 guidelines.¹⁷ Analysis methods were specified in advance and documented in a review protocol. Randomized, placebo-controlled, double-blind, monotherapy studies of IL-12/23 antibodies and anti–TNF-α agents for the treatment of CPP in adults that had publicly available safety outcome data for MACEs were included in the meta-analysis. Studies of psoriatic arthritis were excluded. No language, publication date, or publication status restrictions were imposed.

Studies were identified by systematic electronic literature searches in the Cochrane Central Register of Controlled Trials, ClinicalTrials.gov, and Ovid MEDLINE from inception to June 2010 using combinations of the terms psoriasis, trial, and the names of each of the study drugs (including former names if applicable eg, CNT0-1275, ABT-874). An updating search was performed in May 2011. Two investigators independently searched data while 6 investigators reviewed the abstracted data. Abstractors were not blinded to study drug, authors, institutions, or journals when reviewing the studies. A data abstraction form included information on treatment allocation, methods of randomization, concealment and blinding, inclusion and exclusion criteria, duration and severity of disease, and patient characteristics. This information was used to assess quality and to compare the eligibility criteria of studies. A quantitative quality assessment was performed using a predefined scoring system devised by the Cochrane Collaboration, which allocated categories (present, absent, or not stated) for adequacy of randomization, concealment of allocation, and blinding of patients.¹⁸ All included studies were deemed to be of high quality according to this quantitative assessment (eTable; studies were designed for efficacy). Disagreements were resolved by consensus of the 6 reviewers. To avoid publication bias, the results of all suitable nonpublished completed studies registered with the Cochrane Central Register of Controlled Trials and ClinicalTrials.gov were procured through abstract publications or poster presentations and a funnel plot was generated to assess the possibility of publication bias.¹⁹ The individual sponsor companies were contacted if ambiguity existed regarding the definitions of cardiovascular events reported and in each case this information was provided.

The primary outcome measure was the number of MACEs during the placebo-controlled phase of treatment. Numbers were based on patients who received at least 1 dose of study agent or placebo. All doses of study agents were combined for comparison.

Absolute risk differences were used as an effect measure, measuring the excess probability of MACEs in those receiving active treatment compared with those receiving placebo. All statistical tests were 2 sided and conducted at a significance level of .05. The statistical analysis was performed by independent academic biostatisticians from Baylor University Medical Center. Data were analyzed with Review Manager (RevMan) 5.1 developed by the Cochrane Collaboration, Baltimore, Maryland.

Based on a background rate of 0.0012 events per patient-year (aggregate rate in the placebo group for all 22 studies), a sample size of 4284 patient-years would be required to demonstrate a 0.5% absolute increase in the frequency of MACEs with 80% power during the placebo-controlled phase using the typical ratio of 2:1 (biologic treatment:placebo study design; eFigure 1) demonstrates the increasing number of patients required to achieve higher degrees of statistical power, SiZ statistical software, Cytel Inc, Cambridge, Massachusetts). For 3 of the studies, which included exclusively Asian patients, an additional sensitivity analysis excluding these 3 studies was also performed.

The initial literature search was completed in June 2010. An updating search was performed in May 2011. Of the 532 citations screened, a total of 22 RCTs comprising 10 183 patients met the predefined inclusion criteria (PRISMA flow diagram, Figure 1).^{6 - 8 ,10 - 15 ,20 - 32} Minimal individual study bias was observed with regard to the methodological quality of the studies (quantitative quality assessment, eTable) and eligibility criteria for all studies were very similar. Baseline patient characteristics were highly comparable between studies including patient age, duration of psoriasis, severity of baseline psoriasis as defined by the Psoriasis Area and Severity Index (PASI), and percentage of body surface area (BSA) affected (Table 1). Cardiovascular risk factors were listed as exclusion criteria for a phase 2 briakinumab study and a phase 4 etanercept study (which excluded patients with diabetes, unstable ischemic heart disease, and congestive heart failure).^{12 ,29} There was no evidence of statistical heterogeneity across the studies using the I² test (I² = 0),³³ allowing for combination of trial results using the Mantel-Haenszel fixed-effects method.

Table 2 lists the number of MACEs during the placebo-controlled phase of these studies. The risk difference of MACE between patients treated with study drug and those receiving placebo is shown in Figure 2 and Figure 3. During the placebo-controlled phases of the anti–IL-12/23 studies, 10 of the 3179 patients treated with anti–IL-12/23 therapies had a MACE compared with no events in the 1474 patients treated with placebo (Mantel-Haenszel risk difference, 0.012 events/person-year; 95% confidence interval [CI], −0.001 to 0.026; P =.12; Figure 2). In studies of anti–TNF-α agents, 1 of the 3858 patients receiving anti–TNF-α treatments had a MACE compared with 1 of the 1812 treated with placebo (Mantel-Haenszel risk difference, −0.0005 events/person-year; 95% CI, −0.010 to 0.009; P =.94; Figure 3). A sensitivity analysis excluding 3 studies of Asian patients treated with ustekinumab and adalimumab did not appreciably alter these results (Mantel-Haenszel risk difference in the anti–IL-12/23 group compared with placebo, 0.013; 95% CI, −0.001 to 0.027; P =.09; eFigure 2, and Mantel-Haenszel risk difference in the anti–TNF-α group compared with placebo, −0.0005; 95% CI, −0.010 to 0.009; P =.94; eFigure 3). There was no evidence of publication bias (eFigure 4), although many of the studies had zero results, making this measure of bias less reliable.

Quiz Ref IDThis meta-analysis did not show a significant increase in the risk of MACEs associated with the use of anti–IL-12/23 agents. Limitations of this study, however, prevent us from determining whether these drugs expose psoriasis patients to increased cardiovascular risk. Access to patient-level data for these studies was not granted by any of the study sponsors, which precluded the use of a more statistically robust time-to-event analysis. The small number of MACEs that occurred in placebo-controlled phases of these studies and the limited duration of the placebo-controlled phases reduce the power of this meta-analysis to detect a change in risk. Moreover, the difficulty in obtaining accurate P values for rare and no-event data makes this analysis less reliable.¹⁸ The abstractors were not blinded to authors, institutions, or journals when reviewing studies to be included in the meta-analysis, leading to another potential source of bias. The results of all suitable nonpublished completed studies registered with the Cochrane Central Register of Controlled Trials and ClinicalTrials.gov were procured through abstract publications or poster presentations in the public domain. Although publications of all studies included reported adverse events, including serious adverse events, failure to report serious adverse events in these publications may constitute another source of publication bias. A myocardial infarction in an infliximab-treated patient and a myocardial infarction in a placebo-treated patient were not reported in the serious adverse event sections of publications of phase 3 infliximab studies (data on file, Centocor).³⁴ Quiz Ref IDThere also may be a temporal effect biasing the results of our meta-analysis, as the majority of the anti–TNF-α studies were conducted in a time when there was less vigilance for adverse cardiac events in clinical studies. For example, all potential cardiovascular events that occurred in phase 2 and 3 studies of anti–IL-12/23 agents were adjudicated by a team of cardiology experts.

We remain concerned about the MACE rate of 1.33 per 100 patient-years (95% CI, 0.43-3.10) in the placebo-controlled phase of the phase 3 briakinumab study, with an overall rate of 0.60 events per 100 patient-years (95% CI, 0.35-0.94) across all treatment periods.¹⁶ Indeed, after identification of these cases, a statistical analysis was performed by the manufacturers of briakinumab to determine whether a specific subset of treated patients may be at particularly high risk of a MACE. This analysis led to an amendment to the study protocol for the open-label continuation phase of this study in May 2010 to adjust the exclusion criteria, visit procedures, and discontinuation criteria for the enrolled patients. Patients with 2 or more predefined cardiovascular risk factors who had not previously experienced failure or intolerance to anti–TNF-α therapies or other systemic therapies were withdrawn from the study. Subsequently, in July 2011, all clinical trials of briakinumab were discontinued by Abbott, pending further investigations particularly relating to possible mechanistic links to MACEs. Until more definitive data become available, we believe that dermatologists should exercise heightened vigilance for cardiovascular risk factors when initiating anti–IL-12/23 agents in psoriasis patients.

Quiz Ref IDAlthough the roles of IL-12/23 in the development and progression of atherosclerosis have yet to be clearly elucidated, preliminary evidence suggests that IL-12 is proatherogenic and that its inhibition should confer cardioprotection.^{35 - 37} In the phase 2 study of ustekinumab, however, serum levels of the p40 subunit of IL-12 were shown to paradoxically increase 13-fold in the first 12 weeks of treatment with a gradual decrease to above-baseline levels at week 32.³⁸ Contrary to traditional understanding of antibody-cytokine interactions, the binding of antibody may produce agonistic rather than antagonistic activity.^{39 - 43}

There is conflicting evidence regarding the role of IL-17, an IL-23–induced cytokine, in atherosclerotic plaque inflammation.^{37 ,44 - 51} Recent evidence has suggested that IL-17 and signal transducer and activator of transcription (STAT) 3, a transcription factor downstream of IL-23, may have counterregulatory roles in atherosclerosis.^{50 - 51} With numerous therapeutics agents inhibiting IL-17 and IL-23 currently in development, it is essential that the effect of these cytokines on vascular inflammation be fully explored.

Although meta-analysis is a strategy to increase power for the detection of rare events, this cannot correct for limitations in quality of the original clinical trials, most notably the absence of a systematic strategy to screen for, capture, and adjudicate cardiovascular events. This has been recently highlighted by the international controversy surrounding rosiglitazone, a drug used to treat type 2 diabetes.⁵² Although individual RCTs detected no increase in cardiovascular events, observational studies and meta-analyses suggested an increased risk of MACEs associated with the drug.⁵³ More concerning, however, is the time delay between the detection of potential safety signals and their reporting to regulatory agencies or to the medical community at large. This was also highlighted in the case of rofecoxib, in which safety concerns regarding the risk of thrombotic events existed approximately 4 years before the drug was ultimately withdrawn from the market. Although a manufacturer-driven pooled analysis of more than 28 000 patients from 23 RCTS did not show evidence of an excess of cardiovascular events for rofecoxib relative to placebo,⁵⁴ a large RCT of 2586 patients finally confirmed a 2-fold increase in thrombotic events and was stopped prematurely.⁵⁵ The overwhelming question was why it took a relatively small trial to identify this risk several years after the drug had been prescribed to more than 80 million patients, with continued aggressive marketing by the pharmaceutical company despite warnings of a safety signal.^{56 - 58} A new statement from the US Food and Drug Administration in September 2010, however, issued a final rule clarifying that companies must report safety issues occurring during clinical trials within 15 days, including data that suggest the occurrence of serious adverse reactions at higher-than-expected rates, to expedite the review of critical safety information and protect patients enrolled in clinical trials.⁵⁹

Quiz Ref IDThis analysis highlights the inherent limitations of placebo-controlled clinical trials to reliably interpret the significance of rare events given their current design. Although RCTs are currently the criterion standard for measuring clinical efficacy in psoriasis therapies, these studies are designed to detect differences in the severity of psoriasis in response to therapy over short periods of treatment and are often underpowered and of too short duration to detect rare or long-term adverse events. Careful consideration of these issues is warranted to best serve patients in these studies and those who are treated once drugs are approved.