Comment received about 'Systemic pharmacological treatments for chronic plaque psoriasis: a network meta‐analysis'

Sbidian E, Chaimani A, Garcia-Doval I, Doney L, Dressler C, Hua C, Hughes C, Naldi L, Afach S, Le Cleach L. Systemic pharmacological treatments for chronic plaque psoriasis: a network meta‐analysis. Cochrane Database of Systematic Reviews 2021, Issue 4. Art. No.: CD011535. DOI: 10.1002/14651858.CD011535.pub4. Accessed 08 June 2021. (https://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.CD011535.pub4/full)

Key: Correspondence from LEO Pharma (black text) with response from authors for each section (blue text): 

To whom it may concern:

We would like to thank Doctor Emilie Sbidian and the Cochrane Collaboration for the recent update of the systematic review and network meta-analysis of systemic pharmacological treatments for chronic plaque psoriasis.¹ We acknowledge the sizeable breadth and diversity of the evidence base and commend the move towards making this a “living” systematic review. Precisely because it is to become a “living” systematic review, and because the research is being used to inform clinical guidance such as the forthcoming EADV guidelines on the systemic treatment of psoriasis vulgaris, we consider it important to highlight some concerns regarding the methods used to synthesise the evidence, which directly impacts the applicability of the results to real-world clinical practice.

We thank Leo Pharma for their feedback on our living network meta-analysis (NMA) of systematic treatments for chronic plaque psoriasis and for opening up and interesting discussion. It important to highlight that the methods of our living NMA have been pre-specified in a published protocol (Cochrane review until results section corresponded to the protocol according Cochrane process) and have followed the latest guidance for preparing and reporting NMAs. In this way, we reassure transparency and reproducibility of our findings while avoiding selective reporting of results.

We are extremely concerned that the inclusion of unlicensed doses of products in the main analyses leads to product-ranking results that are misleading, are not relevant to clinical decision making and are at variance with similar published analyses^2-4 and those submitted to and accepted by the National Institute for Health and Care Excellence as part of their technology appraisal guidance.^5-17 In particular, the primary network meta-analysis makes comparisons by pooling together any evaluated dose of included treatments rather than restricting comparisons to licensed doses. This is critical to LEO Pharma as the unlicensed 140 mg dose of brodalumab does not exist and as such cannot be given to patients, nor can other doses evaluated in phase 2 dose finding studies (e.g. 70 mg, 280 mg).¹⁸” 

Indeed, we included all doses in our primary analysis as we had pre-specified in our protocol. This decision was based on the following criteria:

a) Our research question is not restricted on licensed interventions; instead, we are interesting on the best drugs overall

b) License is not a criterion for treatment selection for a NMA; the selection should be based on clinical criteria (e.g. considering whether the dose could differentiate the effect of the intervention)

d) The dose-level network is much sparser compared to the drug-level and a sparse network provides less precise and consequently less informative results.

The pooling of any evaluated dose (licensed and unlicensed) has an impact on the effect size estimate for most therapies included in the analysis, though the proportional effect differs by treatment. Table 1 presents the risk ratios from random effects pairwise meta-analyses comparing different IL-17 and IL-23 treatments included in the NMA with placebo at PASI 75 and PASI 90. The impact of pooling any evaluated dose compared to licensed doses only is to underestimate the effects for IL-17s slightly relative to IL-23s. 

We believe, that our analysis as well as the tables you provide below demonstrate the apparent uncertainty about the impact of dose. Dose may have a small or no impact on the effect of the drugs but this remains a question that deserves further investigation. It is unclear to us why you feel that brodalumab was disfavoured by our analysis. Our main efficacy message was that infliximab, ixekizumab, secukinumab, brodalumab, risankizumab and guselkumab were significantly more effective in reaching PASI 90 than ustekinumab, three anti-TNF alpha agents: adalimumab, certolizumab, and etanercept and non-biological treatments. Neither the drug-level nor the dose-level analysis allowed to draw inference about the differences of these 6 most effective drugs. This can be seen by the respective forest plots. We find the interpretation of very small differences in RR as evidence for difference between drugs without taking into account the respective confidence intervals misleading and not useful for informing daily practice. 

Table 2 shows the risk ratios for IL-17s and risankizumab vs the weight-based doses of ustekinumab (45 mg if <100 kg; 90 mg if >100 kg).  Brodalumab was the only therapy for which two doses were compared against ustekinumab in phase 3 trials.¹⁹ In the AMAGINE-2 and AMAGINE-3 studies, the 140 mg dose was found to be comparable to ustekinumab whilst the 210 mg dose was found to be significantly more efficacious.  Pooling the results for both 140 mg and 210 mg versus ustekinumab leads the reader to believe that brodalumab performs less well against ustekinumab than the other IL-17s and risankizumab.  However, when only the licensed and marketed 210 mg dose is compared, the effect size versus ustekinumab is very similar to the other IL-17s and risankizumab. Together, these analyses highlight a lack of similarity in what is being compared across different comparisons in the evidence network as a whole and highlights the potential for bias when propagated through the full NMA. 

Pooling doses consolidates comparisons and appears to reduce imprecision in the effect sizes observed; however, that precision comes at a cost.  Not only do the results appear to be biased by the pooling, but they are made less applicable to real-world clinical decision making. The approach also results in the exclusion of several relevant studies which could contribute valuable indirect evidence for the doses indicated for clinical practice.  Examples include PRISTINE,²⁰ IXORA-P,²¹ SCULPTURE²² and SIGNATURE.²³

All these trials were included in the sensitivity analysis, which took into account the licensed and unlicensed doses.

There is also a clear dose-response relationship for brodalumab (see Table 3).  When the 210 mg dose is compared to the 140 mg dose, the pooled risk ratio at PASI 75 is 1.26 (1.20 to 1.34) and at PASI 90 is 1.39 (1.25 to 1.54), indicating a statistically significant difference and evidence that these doses should not be pooled.

Table 1. IL-17s and IL-23s vs placebo: comparison of pooling licensed doses only versus any evaluated dose

Table 2. Risk ratios vs ustekinumab: comparison of pooling licensed doses only versus any evaluated dose

As we pointed out earlier, the above tables highlight the uncertainty around the effect of dose. Interpreting point estimates in isolation without considering the confidence intervals is not appropriate. 

Table 3. Risk ratios vs placebo: comparison of pooling licensed doses only versus any evaluated dose

It is hard to justify a dose-response relationship with the estimates as there is large overlap of the confidence intervals. Also, the number of studies that evaluates the different doses is different and this certainly should be taken into account when interpreting these results. However, we agree that there is room for a proper dose-response meta-analysis for drugs with several studies. Probably this will be the best way to resolve this controversy and this is something we plan to do. 

We acknowledge that a dose-level sensitivity analysis was performed and presented, which addresses our above concerns.  However, it is unclear why this dose-level analysis is a sensitivity analysis instead of the main analysis.  The dose-level comparison includes the most relevant data to decision makers and is likely subject to less statistical heterogeneity and prone to less bias.  In addition, the risk ratios and relative rank of therapies changes among the most efficacious therapies (see Table 4), though this does not get much discussion.  The change in the risk ratio for brodalumab is particularly noteworthy.  We would urge the authors that this become the primary analysis as the systematic review is continuously updated.

Interestingly, almost no statistical heterogeneity was observed for all outcomes in our NMA. Regarding ranking, it is not the main output of our NMA but it is provided only as a supplementary output. The reason for this, is because it has been recommended to avoid too much emphasis on ranking when there is a lot of uncertainty in the relative effects among the drugs. This is the case here. Below you rank the drugs using the RRs, which completely ignores the confidence intervals. In our review, we used the SUCRA, which to some degree takes into account also the confidence intervals. In both cases, though, over-interpretation of ranking should be avoided. Here it seems that you consider a difference between RR=29.52 and RR=28.12 a difference in ranking which is misleading.  

Table 4. Comparison of risk ratios and relative rank from drug-level and dose-level analyses on PASI 90 outcome

It also appears that the NMA did not account for variation in placebo response rates in the included studies, something that has been highlighted in the literature as a potential source of heterogeneity and bias, which should be accounted for quantitatively.²⁴  

We just have demonstrated previously that no real variation in placebo response rates exist. Specifically, using all the studies in our NMA, the PBO response rate for PASI 90 was 1% with 95% CI: 1 to 2%). (data submitted for publication)

Furthermore, in order to take account of the evolving scientific landscape and rising patient expectations when updating guidelines, the substantial body of data that now exists for evaluating PASI 100 as an outcome should be included.  Different biologics have different probabilities of achieving PASI 100, even with similar levels of near clearance (PASI 90). Several studies have shown that the difference for a patient between achieving 90% improvement in PASI and achieving full clearance is dramatic, and very clear in terms of PRO measures of health-related quality of life.

We agree that PASI100 is an interesting outcome. However, it is a new outcome used mainly in studies evaluating the most recent drugs. Thus, a lot of missing data is expected. We will consider this outcome in future updates of the review if it is reported in several studies.

We would welcome the opportunity to discuss our feedback with the authors of the review as we share the Cochrane Collaboration’s vision of improving the quality of healthcare through decision making based on high-quality, relevant and up-to-date synthesised research evidence. 

Yours sincerely,

Emma Borg
Head of Patient Access
Global Bio-Dermatology

Alexandre Abramavicus, MD
Principle Professional
Medical Deparment

(When asked about any affiliation with or involvement in any organisation with a financial interest in the subject matter of the comment. The comment contributor answered: I am an employee of LEO Pharma A/S).

We hope our responses explained well our approach and mitigated your concerns about our conclusions in which we have included brodalumab as one of the best choices.

1. Sbidian E, Chaimani A, Afach S, et al. Systemic pharmacological treatments for chronic plaque psoriasis: a network meta-analysis. Cochrane Database Syst Rev 2020;1:CD011535.

2. Armstrong AW, Puig L, Joshi A, et al. Comparison of Biologics and Oral Treatments for Plaque Psoriasis: A Meta-analysis. JAMA Dermatol 2020.

3. Cameron C, Druchok C, Hutton B, et al. Guselkumab for the Treatment of Moderate-to-Severe Plaque Psoriasis During Induction Phase: A Systematic Review and Network Meta-Analysis. Journal of Psoriasis and Psoriatic Arthritis 2018;4:81-92.

4. Sawyer LM, Malottki K, Sabry-Grant C, et al. Assessing the relative efficacy of interleukin-17 and interleukin-23 targeted treatments for moderate-to-severe plaque psoriasis: A systematic review and network meta-analysis of PASI response. PLoS One 2019;14:e0220868.

5. National Institute for Health and Care Excellence. Etanercept and efalizumab for the treatment of psoriasis - Technology Assessment Report, 2006.

6. National Institute for Health and Care Excellence. Infliximab for the treatment of adults with psoriasis - manufacturer's submission, 2008.

7. National Institute for Health and Care Excellence. Adalimumab for the treatment of adults with psoriasis - manufacturer's submission, 2008.

8. National Institute for Health and Care Excellence. Ustekinumab for the treatment of adults with moderate to severe psoriasis - manufacturer's submission, 2009.

9. National Institute for Health and Care Excellence. Secukinumab for treating moderate to severe psoriasis - manufacturer's submission, 2015.

10. National Institute for Health and Care Excellence. Apremilast for treating moderate to severe psoriasis - manufacturer's submission, 2016.

11. National Institute for Health and Care Excellence. Dimethyl fumarate for treating moderate to severe psoriasis - manufacturer's submission, 2017.

12. National Institute for Health and Care Excellence. Ixekizumab for treating moderate to severe psoriasis - manufacturer's submission, 2017.

13. National Institute for Health and Care Excellence. Brodalumab for treating moderate to severe psoriasis - manufacturer's submission, 2018.

14. National Institute for Health and Care Excellence. Guselkumab for treating moderate to severe psoriasis - manufacturer's submission, 2018.

15. National Institute for Health and Care Excellence. Certolizumab for treating moderate to severe psoriasis - manufacturer's submission, 2019.

16. National Institute for Health and Care Excellence. Tildrakizumab for treating moderate to severe psoriasis - manufacturer's submission, 2019.

17. National Institute for Health and Care Excellence. Risankizumab for treating moderate to severe psoriasis - manufacturer's submission, 2019.

18. Papp KA, Leonardi C, Menter A, et al. Brodalumab, an anti-interleukin-17-receptor antibody for psoriasis. N Engl J Med 2012;366:1181-9.

19. Lebwohl M, Strober B, Menter A, et al. Phase 3 Studies Comparing Brodalumab with Ustekinumab in Psoriasis. N Engl J Med 2015;373:1318-28.

20. Strohal R, Puig L, Chouela E, et al. The efficacy and safety of etanercept when used with as-needed adjunctive topical therapy in a randomised, double-blind study in subjects with moderate-to-severe psoriasis (the PRISTINE trial). J Dermatolog Treat 2013;24:169-78.

21. Langley RG, Papp K, Gooderham M, et al. Efficacy and safety of continuous every-2-week dosing of ixekizumab over 52 weeks in patients with moderate-to-severe plaque psoriasis in a randomized phase III trial (IXORA-P). Br J Dermatol 2018;178:1315-1323.

22. Mrowietz U, Leonardi CL, Girolomoni G, et al. Secukinumab retreatment-as-needed versus fixed-interval maintenance regimen for moderate to severe plaque psoriasis: A randomized, double-blind, noninferiority trial (SCULPTURE). J Am Acad Dermatol 2015;73:27-36 e1.

23. Warren RB, Barker J, Finlay AY, et al. Secukinumab for patients failing previous tumour necrosis factor-alpha inhibitor therapy: results of a randomized open-label study (SIGNATURE). Br J Dermatol 2019.

24. Cameron C, Hutton B, Druchok C, et al. Importance of assessing and adjusting for cross-study heterogeneity in network meta-analysis: a case study of psoriasis. J Comp Eff Res 2018;7:1037-1051.