T. Tzellos and A. Kyrgidis have made equal contribution to this study.
Conflict of interest TT has been reimbursed for travel expenses and hotel accommodation to attend Greek Dermatological Congresses by Janssen-Cilag, which produces ustekinumab (Stelara®) and by MSD, which produces infliximab (Remicade®). CCZ has been reimbursed for travel expenses and hotel accommodation and has received an honorarium for participating and lecturing at the Advisory Board for hidradenitis suppurativa of Abbott, which produces adalimumab (Humira®). He has also participated in clinical studies on the treatment of hidradenitis suppurativa with adalimumab.
Objective To detect a detrimental or beneficial effect of anti-IL-12/23 biological agents (ustekinumab and briakinumab) for the treatment of chronic plaque psoriasis on major adverse cardiovascular events (MACEs).
Design Systematic review and meta-analysis MEDLINE, EMBASE, the Cochrane Skin Group Specialised Register, Cochrane Central Register of Controlled Trials (CENTRAL) in The Cochrane Library, SciVerse Scopus and ongoing trial registries were searched from inception until December 2011. Search strategy, eligibility criteria, data and statistical analysis methods were defined prior to the literature search. Randomized, placebo-controlled, double-blind, monotherapy studies with safety data for MACEs of IL-12/23 antibodies in adults were eligible for inclusion. Studies of psoriatic arthritis were excluded. Information from each study was extracted independently by two reviewers, using a standardized data extraction form. The primary outcome measure was the number of MACEs during the placebo-controlled phase of treatment.
Results MACEs include myocardial infarction, cerebrovascular accident or cardiovascular death. No statistical heterogeneity across the studies using the I2 statistic (I2 = 0) was found. We employed Peto one-step method to determine odds ratios and quantify a possible detrimental or beneficial association of IL-12/23 antibodies treatment with MACEs. We found a possible higher risk of MACEs in those patients treated with IL-12/23 antibodies compared with those at placebo (OR = 4.23, 95% CI: 1.07–16.75, P = 0.04). This study is unaffected by non-reporting of outcomes with no events.
Conclusion Compared with placebo, there was a significant difference in the rate of MACEs observed in patients receiving anti-IL-12/23 biological agents.
Chronic plaque psoriasis (CPP) is a very common, immune mediated, chronic disease occurring worldwide, which affects skin and joints and has a negative impact on both the physical and psychological well-being of patients.1 Interestingly, it has also been suggested that CPP is associated with chronic systemic inflammation and thus with an increased cardiovascular risk.2 Having taken this into account, the hypothesis was formed that controlling this chronic inflammation process would reduce the cardiovascular risk. Indeed, a recent meta-analysis indicated that methotrexate use in patients with rheumatoid arthritis, CPP or polyarthritis is associated with a lower risk for cardiovascular disease and that a direct treatment of inflammation may reduce cardiovascular risk.3 For biological agents, on the other hand, up to now, two meta-analyses failed to detect a detrimental or beneficial effect on major adverse cardiovascular events (MACEs).4,5 Both meta-analyses concluded that more robust evidence is needed. Preliminary reports of MACEs in randomized controlled trials (RCTs) of psoriasis patients treated with anti-IL-12/23 biologic agents, ustekinumab and briakinumab have been documented.6,7 Utility of biological agents and more specifically anti-IL-12/23 in CPP treatment necessitates the use of the most appropriate methods to evaluate their safety. Single trials are underpowered to provide conclusive evidence regarding rare cardiovascular events and meta-analyses conducted with the most appropriate statistical model can improve power.8 To achieve this, we conducted a meta-analysis of randomized, placebo-controlled, double-blind, monotherapy studies of IL-12/23 antibodies for the treatment of CPP in adult patients.
To identify eligible studies for inclusion, the main search was conducted in the electronic databases MEDLINE, EMBASE, the Cochrane Skin Group Specialised Register, Cochrane Central Register of Controlled Trials (CENTRAL) in The Cochrane Library and SciVerse Scopus from inception until October 2011 using a draft search strategy for RCTs for MEDLINE (OVID), as suggested in Cochrane Handbook (Table 1).9 This search strategy was used as basis for search strategies for the other databases listed, after modifications accordingly. For the identification of ongoing trials we searched the US National Institutes of Health Ongoing Trials Register (http://www.clinicaltrials.gov) and The World Health Organization International Clinical Trials Registry platform (http://www.who.int/trialsearch) using the following search string: [psoriasis AND (ustekinumab OR CNTO-1275 OR CNTO 1275 OR monoclonal antibody CNTO 1275 OR monoclonal antibody CNTO-1275 OR briakinumab OR ABT-874 OR ABT 874 OR monoclonal antibody ABT 874 OR monoclonal antibody ABT-874)]. The manual search was concluded by the perusal of the reference sections of all studies (included and excluded) to identify relevant unpublished data. To identify additional eligible studies from conference proceedings we checked the ISI Web of Science, Science Citation Index. We have also scanned the abstracts of the major dermatology conference proceedings listed and recorded in the Cochrane Skin Group Specialised Register. An updated search was conducted in December 2011 using the same search strategy. The main search as well as screening of titles and abstracts was completed independently by two reviewers (AK and TT) with expertise in conducting systematic reviews. Chance-adjusted inter-rater agreement was calculated using Cohen’s kappa statistic and was found to be satisfactory. Any discrepancy was solved by consultation of a third reviewer, not involved in the initial procedure (KT).
Table 1. Draft search strategy for MEDLINE (OVID)
1. randomized controlled trial.pt.
2. controlled clinical trial.pt.
5. clinical trials as topic.sh.
8. 1 or 2 or 3 or 4 or 5 or 6 or 7
10. 8 and 9
11. exp psoriasis/
13. 11 or 12
14. 10 and 13
15. Exp ustekinumab/
17. Exp CNTO 1275/
18. CNTO 1275.mp.
19. Exp monoclonal antibody CNTO-1275/
20. monoclonal antibody CNTO-1275.mp.
21. Exp CNTO-1275/
23. Exp monoclonal antibody CNTO 1275/
24. Monoclonal antibody CNTO 1275.mp.
25. Exp briakinumab/
27. Exp ABT-874/
29. Exp monoclonal antibody ABT-874/
30. monoclonal antibody ABT-874.mp.
31. Exp ABT 874/
32. ABT 874.mp.
33. Exp monoclonal antibody ABT 874/
34. monoclonal antibody ABT 874.mp.
35. 15 or 17 or 19 or 21 or 23 or 25 or 27 or 29 or 31 or 33 or 16 or 18 or 20 or 22 or 24 or 26 or 28 or 30 or 32 or 34
36. 14 and 35
Eligibility of relevant studies/selection of studies
In order that a study be eligible, the following criteria were established: (i) the study should include chronic plaque psoriasis adult patients treated with at least one IL-12/23 antibody; (ii) the study should report safety outcome data for MACEs.
Eligible studies for the meta-analysis were randomized, placebo-controlled, double-blind, monotherapy studies. RCT study quality was assessed independently by two reviewers using the six-item instrument developed by Jadad et al.10
Inter-rater agreement was evaluated by means of kappa test, and disagreement was resolved by consensus. Studies of patients with psoriatic arthritis were excluded. Cohort studies, case control studies, case series, reviews, abstracts, letters to the editor, cross sectional studies were not eligible. Any discrepancy was solved by consultation of a third reviewer, not involved in the procedure (KT). All criteria for inclusion/exclusion of studies in the present systematic review were specified prior to the literature search.
Information from each study was extracted independently by two reviewers (AK and TT), using a standardized data extraction form. Study of general characteristics for both intervention and control groups (author group, journal, year of publication, design, intervention and control group sample size, methodology, and outcomes) was recorded, where available, and double-checked. Overall incidence of MACEs in the treatment and control groups of individual studies was calculated on the intention-to-treat basis.
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 one dose of study agent or placebo. All doses of study agents were combined for comparison.
To quantify the probability of MACEs in those receiving active treatment compared with those receiving placebo, confidence intervals (CI) were calculated from individual studies and combined using Peto Odds Ratio method as the main weighting scheme. This was decided because the Peto method has been reported to perform better among various meta-analytical methods for event rates lower than 1%.11 The Peto one-step method computes an approximation of the log-odds from the ratio of the efficient score to the Fisher information, both evaluated under the null hypothesis. These quantities are estimated, respectively, by the sum of the differences between the observed and expected numbers of events in the treatment arm and by the sum of the conditional hypergeometric variances.11 The Peto OR method does not require zero-cell corrections to produce valid estimates in all circumstances. Trials with no events in both arms are automatically given zero weight and effectively excluded from the analysis.11 The Peto OR method for pooling trials demonstrated the best statistical properties where events were rare, treatment effects small or moderate and the numbers of treated and control participants close to equal in the majority of trials. The method produces estimates except when no events were recorded in any trial.11 Mantel-Haenszel risk difference, although reportedly suboptimal and less powerful for the detection of rare adverse events,11 was also used in this meta-analysis to determine new events under each treatment arm, within each study.
Heterogeneity between the results of different studies was examined by Chi-square tests for significance (P-value < 0.1 was considered statistically significant) and presented by I2 test (I2 > 50%: significant heterogeneity; I2 < 25%: insignificant heterogeneity), which can be interpreted as the percentage of total variation across several studies due to heterogeneity.12 To assess the extent of publication bias, Egger’s test was performed. As an alternative method to present summary statistics and MACEs safety profile, Number Needed to Harm (NNH) was calculated, wherever appropriate, with the use of relevant software (http://www.nntonline.net).12 Meta-analysis was conducted using Review Manager 5 software [Review Manager (RevMan), Oxford, England: The Cochrane Collaboration, 2011] and stata/SE 9.0 for Windows (StataCorp LP, 4905 College Station, TX, USA).
Search results are presented in the PRISMA flow diagram (Fig. 1). Of the 154 titles screened, nine RTCs 6,7,13–19 met inclusion criteria and were included in the meta-analysis (Table 2). For those RCTs we calculated new MACEs per patient-year of biological agent treatment (or risk difference), is also presented in Table 2. No significant heterogeneity among studies was reported (Χ2 = 1.91; DF = 8 (P = 0.98); I2 = 0%). Baseline patient characteristics among studies included 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 and did not show any statistical significant difference (Pearson’s Χ2P = 0.27, Table 2). Figure 2 analyses the number of MACEs during the placebo-controlled phase of these studies. MACEs event rates were 0.28% (95% CI: −0.03 to 0.59), 0.35% (95% CI: −0.03 to 0.74) and 0.31% (95% CI: 0.09–0.53) for ustekinumab, briakinumab and both agents respectively. Analysed separately, the odds ratio for patients treated with briakinumab (OR 4.47, 95% CI: 0.69–28.89, P = 0.12) and ustekinumab (OR 3.96, 95% CI: 0.51–30.41, P = 0.19) was not found statistically significant. When combined, the odds ratio for MACEs between patients treated with biological agent and those receiving placebo was found to be statistically significant (OR = 4.23, 95% CI: 1.07–16.75, P = 0.04, Fig. 2). Egger’s test was non-significant (P = 0.148), possible indication of no publication bias. Number needed to harm was calculated on intention-to-treat basis; treatment with both IL-12/23 antibodies in 318 patients (95% CI: 187.5–1066.9) would result in one MACE event.
Table 2. Summary of randomized controlled trials included in the meta-analysis with baseline patient characteristics
Biological agent group*
Patient age (mean ± SD)
Time since CPP diagnosis (mean ± SD)
Percentage of affected body surface area (mean ± SD)
PASI score (mean ± SD)
Length of placebo-controlled-phase, wk
No. of MACEs in biologic agent group
Events per person-year (95% CI)
*Presented as person-years/absolute number of patients.
†Denotes median values instead of mean.
CPP, Chronic plaque psoriasis; MACEs major adverse cardiovascular events; PASI, Psoriasis Area and Severity Index.
The increasing use of biological agents may allow for even the slightest statistical miscalculations to be amplified and for adverse effects to be reported outside clinical trial settings in higher rates. We employed a more powerful statistical methods in this meta-analysis of randomized, placebo-controlled, double-blind, monotherapy studies of IL-12/23 antibodies for the treatment of CPP in adult patients to exclude a possible higher risk for MACEs in the treatment arm compared to placebo. Notably, the results demonstrate that a higher risk for MACEs in treatment arm patients is plausible.
Ryan et al.5 in a recent meta-analysis were not able to detect this association, not due to inclusion of fewer MACEs but because of the use of less powered risk-difference methods. Ryan et al. used Mantel-Haenszel fixed-effects model with absolute risk differences as effect measure.5 Risk difference methods produced estimates for all studies because of using zero-cell corrections. The inability of most meta-analytical methods to produce finite odds ratios in cases where events occur in only one arm limits their applicability and endorses the use of risk difference methods. But risk difference methods yield very conservative confidence interval coverage when events are rare, and have associated poor statistical power, which make them unsuitable for meta-analysis of rare events.11,20 This is especially relevant when outcomes that focus on treatment safety are being studied, as the ability to correctly identify (or attempt to refute) serious adverse event rates is a key issue in drug development. It is likely that outcomes for which no events occur in either arm may not be mentioned in reports of many randomized trials, precluding their inclusion in a meta-analysis.9 It remains unclear, although, when working with published results, whether failure to mention a particular adverse event means there were no such events, or simply that such events were not included as a measured endpoint.9 Seven of the nine studies included in this meta-analysis failed to clearly report in the methodology section the method to screen for MACEs.7,13,14,16–19 Although the results of risk difference meta-analyses will be affected by non-reporting of outcomes with no events, odds-ratio-based methods naturally exclude these data whether or not they are published, and are therefore unaffected.11 The latter argument may partly explain the differences between the present meta-analysis and the one by Ryan et al.5 Bradburn reported that at a baseline event rate of 1%, as is the case in the present meta-analysis, the Peto method had the best performance among all meta-analytical methods considered, although risk difference methods have been reported to exhibit low statistical power at event rates of 1% or lower.11 Included trials in this meta-analysis have more patients in the intervention arm than those in the control (aggregate 3179 vs. 1474, Table 2). Bradburn et al. demonstrated that applying the Peto method in these situations, imbalance in the arm sizes leads to underestimation of the number of events in the larger group (which in our case is the biological treatment group) whereas the bias in estimates of even large treatment effects are negligible.11 Therefore, even using Peto OR our study may have still underestimated the true MACEs rate.
Greenland and Salvan21 argue that the Peto one-step method is badly biased in some situations. They demonstrated considerable bias using example study data with imbalances between the numbers in the two groups of 8:1 or higher and odds ratios of above 2.5.21 However, such scenarios are rarely encountered in the meta-analysis of RCTs.11 Greenland and Salvan report that such biases in the Peto method are negligible in meta-analyses of randomized trials involving small effects and a reasonable number of outcome events.21 In all cases, the Peto OR is a more powerful meta-analytical statistical method and therefore more suitable to exclude a possible association of IL-12/23 antibodies with MACEs.
Briakinumab and ustekinumab are both human monoclonal antibodies targeting the same shared sub-unit (p40) of IL-12 and IL-23, belong to the same drug class and in the RCTs included were used for the same indication. In this context, a meta-analysis of drug class is valid, since a class effect cannot be excluded with the current evidence, and such meta-analyses can lead to important findings.22 Although they are prescribed with a different dosage schema, there is no evidence up to now indicating that the risk for MACEs is dose-related.
Recent evidence suggests that inflammatory joint disease may play a role in cardiovascular morbidity in psoriatic arthritis.23 The role of the presence of psoriatic arthritis in patients with CPP treated with these biological agents in the rate of MACEs remains to be elucidated. Furthermore, 9 of 10 MACEs were reported in patients with at least one baseline pre-existing cardiovascular risk factor, whereas Leonardi et al.18 fail to clearly report any pre-existing cardiovascular risk factor.
Limitation of this meta-analysis remains the fact that all these RCTs were designed to detect mostly efficacy, with short periods of treatment (12–20 weeks) and included patients with a mean age less than 50 years. The presence of uncontrolled diabetes, unstable ischemic heart disease and congestive heart failure was listed as exclusion criterion for a briakinumab study.19 Furthermore, MACEs are long term and relatively rare outcomes that cannot be adequately studied and detected in such RCTs. Despite all these, this meta-analysis of RCTs with the use of a more powerful statistical approach was able to detect a statistically significant increase in MACEs associated with anti-IL-12/23 use. In the general population, patients with risk factors for cardiovascular morbidity may use these agents for a long period of time, and therefore the real rate of MACEs could be higher. This evidence clearly indicates that increased vigilance and post marketing surveillance for cardiovascular risk factors must be applied when anti-IL-12/23 are used. Recommendations for future studies most definitely include the conduction of a well organized meta-analysis of observational studies to detect high risk subgroups and acquire a better scheme of the true effect and most importantly an individual patient data meta-analysis of RCTs.
The authors would like to thank Dr Konstantinos Toulis for his contribution as the consultant third reviewer not involved in the initial process in ‘search strategy’ and ‘selection of studies’ procedures.