Diabetes Care 1

Proprotein Convertase Subtilisin/ Luiz Sergio´ F. de Carvalho,1 Alessandra M. Campos,1,2 and Kexin Type 9 (PCSK9) Inhibitors Andrei C. Sposito1 and Incident Type 2 Diabetes Mellitus: A Systematic Review and Meta-analysis With Over 96,000 Patient-Years https://doi.org/10.2337/dc17-1464

OBJECTIVE Like mutations with loss of function in the proprotein convertase subtilisin/kexin type 9 (PCSK9) , inhibitors of PCSK9 (PCSK9i) may potentially favor the manifes- tation of diabetes.

RESEARCH DESIGN AND METHODS A meta-analysis of phase 2/3 randomized clinical trials (RCTs) assessed PCSK9i versus placebo in primary setting. and were used in 98.4% of these studies and balanced between PCSK9i and placebo. We calculated relative risks (RRs) and 95% CIs using random- and fixed-effect models.

RESULTS DIABETES IN COMMUNICATIONS NOVEL We included 68,123 participants (20 RCTs) with median follow-up of 78 weeks. PCSK9i increased fasting blood glucose (RR 1.88 [95% CI 0.91–2.68] mg/dL; I2 =0%; 2 P < 0.001) and HbA1c (0.032% [0.011–0.050]; I = 15.5%; P < 0.001) when compared with placebo. This effect was not sufficient to increase incidence of diabetes (1.04 [0.96–1.13]; I2 =0%;P = 0.427). Exploratory meta-regression analyses indicated an P association between the increased risk of diabetes and the potency ( =0.029)and 1 P Cardiology Department, State University of duration ( = 0.026) of PCSK9i treatment. Campinas, Campinas, SP, Brazil 2Pharmaceutical Sciences Department, Faculty CONCLUSIONS of Health Sciences, University of Brasilia, Bras´ılia, At short-term, PCSK9i therapy favors a small but significant increase in plasma DF, Brazil glycemia and HbA1c. Corresponding author: Andrei C. Sposito, [email protected]. Received 19 July 2017 and accepted 26 October Proprotein convertase subtilisin/kexin type 9 (PCSK9) is a circulating that binds 2017. to the LDL receptor (LDLR), directing it to lysosome degradation pathways (1). Anti- This article contains Supplementary Data online PCSK9 monoclonal (MAb) therapy reduces LDL (LDL-C) by 50–60% at http://care.diabetesjournals.org/lookup/ in several clinical settings and over the maximal dose of statins plus ezetimibe, with a suppl/doi:10.2337/dc17-1464/-/DC1. satisfactory safety profile (2). © 2017 by the American Diabetes Association. Meanwhile, both Mendelian randomization and randomized clinical trials (RCTs) Readers may use this article as long as the work is properly cited, the use is educational and not suggest an inverse association between plasma PCSK9 availability and the incidence for profit, and the work is not altered. More infor- of type 2 diabetes (3). Gain-of-function mutation on the LDLR gene impairs insulin mation is available at http://www.diabetesjournals secretion capacity by pancreatic b-cells (4). By inference, it is plausible that the .org/content/license. Diabetes Care Publish Ahead of Print, published online November 27, 2017 2 PCSK9i Significantly Increase Blood Glucose Diabetes Care

improvement in LDLR turnover and function- diagram can be found in Supplementary (1.50 years). When limiting the analysis to ality after PCSK9 MAb underlies a decline Fig. 1A and our network profile in Supple- studieswithfollow-upof$48 weeks(Sup- in b-cell function favoring type 2 diabe- mentary Fig. 1B. plementary Fig. 4C), there was no signifi- tes manifestation. Hence, we conducted a For our coprimary outcome meta-analysis, cant increase in the risk of diabetes. As systematic review and meta-analysis of we included 20 RCTs (5–19) with 68,123 shown in Supplementary Fig. 2C and D RCTs with PCSK9 inhibitors (PCSK9i) in or- patients. Baseline study characteristics can and Supplementary Table 4, there was der to explore the existence and magni- be found in Supplementary Table 1A. no significant publication bias in funnel plots tude of this metabolic interaction. Baseline patient characteristics can be and no small study bias by the Egger test. found in Supplementary Tables 2 and 3. Across anti-PCSK9 MAbsdalirocumab, d RESEARCH DESIGN AND METHODS , and no differ- PCSK9i and Glycemic Levels ence was found to the propensity of favor- Cochrane guidelines and Preferred Report- The changes in both fasting blood glucose ing incident or worsening type 2 diabetes, ing Items for Systematic Reviews and (FBG) and HbA from baseline to follow- as shown in Supplementary Fig. 3A and Meta-Analyses statements were used to 1c up were compared in patients randomized B. Exploratory meta-regression analyses in- conduct the meta-analysis and report our to PCSK9i or placebo. Only trials with back- dicated the association between the mag- findings. A detailed description of all pro- ground were included in this analy- nitude and duration of LDL-C lowering cedures is included in the Supplementary sis, because no trials with restricted statin and the risk of worsening diabetes or Data (from section Data Source and use or open-label statins reported glyce- new onset (Supplementary Fig. 5). Search to Included and Excluded stud- mic parameters. Patients treated with ies). Briefly, MEDLINE (PubMed), Co- PCSK9i had an absolute increase (weighted chrane Library, and ClinicalTrials.gov mean difference) of 1.88 mg/dL (95% CI CONCLUSIONS databases were searched for original ar- 0.91–2.68), which was significantly dif- Consistently with observational and mech- ticles from inception to 19 March 2017 to ferent from placebo (Fig. 1A) (standard- anistic data, this meta-analysis reveals identify all RCTs using PCSK9i therapy. ized mean difference 0.166% [95% CI the existence of a small albeit significant Original trials were eligible for the present 0.143–0.188; I2 =0%;P , 0.001). Regard- increase in plasma levels of glucose and meta-analysis if they met the following ing HbA levels, compared with baseline, HbA after a short-term (1.5-year) treat- criteria: 1) phase 2 or 3 RCT; 2)partici- 1c 1c patients treated with PCSK9i had a weighted ment with PCSK9i; this effect was propor- pants with familial or nonfamilial hyper- mean difference of 0.032% (0.011–0.050) tional to the decrease in plasma LDL-C. In cholesterolemia; 3) participants in the (Fig. 1B) (standardized mean differ- this cohort of patients and under such a treatment group received PCSK9i versus ence 0.096% [0.074–0.119]; I2 =15.5%; short period of time, this increase was not control group, who received placebo with P , 0.001). As shown in Supplementary sufficient to impact on incident diabetes. or without other lipid-lowering therapy, ir- Fig. 2A and B and Supplementary Table 4, These findings are strengthened by the very respective of balanced use of statin or there was no significant publication bias low heterogeneity among trials in both con- ezetimibe across treatment arms; and 4) in funnel plots and no significant small tinuous and dichotomous end points and treatment duration $12 weeks. Two in- study bias according to the Egger tests. In the scrutiny of robust sensitivity analysis. vestigators independently abstracted data addition, even after excluding the results In line with our findings, a meta-analysis by using prespecified forms and indepen- of the SPIRE-1 and SPIRE-2 trials, HbA including 91,140 patients followed during dently appraised the accuracy of the ab- 1c and FBG significantly increased with ;4 years found a 9% increased risk for stractions and resolved any discrepancies PCSK9i in comparison with placebo. Addi- incident diabetes with statins (20). This by consensus after discussion with the third tional sensitivity analyses are available in incidence was amplified by intensive sta- investigator. Baseline data were obtained the Supplementary Data (Additional re- tin therapy and in the individuals with by weighted calculation. To identify poten- sults in the Exploratory meta-regression greater decrease in LDL-C ($50%). Con- tial effects of PCSK9i therapy on incident analyses on PCSK9i effect on incident currently, observational studies using diabetes, we calculated an overall relative type 2 diabetes section). Mendelian randomization approach ob- risk (RR) with both random- and fixed-effect served an inverse association between model meta-analyses. Odds ratios and risk PCSK9i and Incident Diabetes genetically determined cholesterol levels ratios were universally identical during data In most of the included RCTs, the use of and incident diabetes (21). An inverse as- analysis. Further details regarding data anal- statins and ezetimibe was reported to be sociation was also described between the ysis will be found in the Supplementary balanced between PCSK9i and control severity of LDLR dysfunction in patients Data (Data Synthesis and Analysis). For arms, except in the following trials: GAUSS, with familial hypercholesterolemia and the summary treatment effect estimate, a ODYSSEY OPTIONS I, ODYSSEY COMBO II, propensity to develop diabetes (3). In par- two-tailed P value ,0.05 was considered and ODYSSEY MONO (1.6% of the total allel, in vitro and in vivo models indicate statistically significant. We analyzed data sample size). LDLR and intracellular cholesterol content with Stata 13. As seen in Supplementary Fig. 4A and as key regulators of pancreatic b-cell ho- B, PCSK9i did not significantly increase the meostasis (4). Altogether, our findings and RESULTS incidence of diabetes (RR 1.045 [95% these prior studies support the concept of Using MEDLINE/PubMed, Cochrane Library CI 0.954–1.135]; I2 = 0%; P = 0.345) or an inverse and dose-dependent effect of databases, and ClinicalTrials.gov, we iden- the compound end point of incident or wors- LDL-C levels on plasma glucose levels. tified 133 citations using the previously ening diabetes (RR 1.035 [95% CI 0.949– As seen with statins, heterogeneity of defined search terms. Our search flow 1.128]; I2 =0%;P = 0.429) after 78 weeks this lipid-lowering effect on glucose care.diabetesjournals.org de Carvalho, Campos, and Sposito 3

Figure 1—Meta-analysis of studies assessing the effect of PCSK9i vs. placebo on FBG change (A) and HbA1c change (B). The odds ratio for each study is indicated by a gray box, and the horizontal line shows the corresponding 95% CI. The combined estimate is indicated by the diamond-shaped box. The dashed vertical line points to the center of estimates with the combined trials. Yellow boxes are the center of the estimate. Maroon bars summarize CIs of each individual study. The blue images are the summaries for each drug and for the three drugs in combination. There is no comparison between them. The i2 (heterogeneity) for bococizumab can be calculated only if SPIRE-1 and SPIRE-2 trials are separated. However, these data are not published in separate, 2 but only in combination. The new picture excluded the i for bococizumab and on fasting blood glucose and for bococizumab on HbA1c.D+L Subtotal/Overall, random-effect meta-analysis; I-V Subtotal/Overall, fixed-effects meta-analysis; SMD, standardized mean difference. (A high-quality color representation of this figure is available in the online issue.) 4 PCSK9i Significantly Increase Blood Glucose Diabetes Care

is clearly influenced by dura- manuscript. A.C.S. is the guarantor of this work SAR236553, to reduce low-density tion of treatment. Indeed, the meta- and, as such, had full access to all the data in the cholesterol in patients with heterozygous famil- regression analyzes obtained in our study study and takes responsibility for the integrity of ial hypercholesterolaemia on stable statin dose the data and the accuracy of the data analysis. with or without ezetimibe therapy: a phase 2 indicated a progressively greater imbalance randomised controlled trial. Lancet 2012;380: in glucose homeostasis as the duration of References 29–36 treatment increases. As the follow-up period 1. Lo Surdo P, Bottomley MJ, Calzetta A, et al. 13. Roth EM, McKenney JM, Hanotin C, Asset G, is restricted to 114 weeks, it was not possible Mechanistic implications for LDL receptor degra- Stein EA. with or without an antibody to foresee the long-term nature of this effect dation from the PCSK9/LDLR structure at neutral to PCSK9 in primary hypercholesterolemia. N Engl pH. EMBO Rep 2011;12:1300–1305 J Med 2012;367:1891–1900 curve shape. In this meta-analysis, although 2. Stein EA, Giugliano RP, Koren MJ, et al.; 14. Raal F, ScottR, Somaratne R, et al. Low-density the incidence of diabetes may have been PROFICIO Investigators. Efficacy and safety of lipoprotein cholesterol–lowering effects of minimized by the short follow-up period, evolocumab (AMG 145), a fully human monoclo- AMG 145, a to proprotein the achieved statistical power (1-b)was nal antibody to PCSK9, in hyperlipidaemic patients convertase subtilisin/kexin type 9 serine protease 97% for both FBG and HbA , which rein- on various background lipid therapies: pooled in patients with heterozygous familial hypercho- 1c analysis of 1359 patients in four phase 2 trials. lesterolemia: the Reduction of LDL-C with PCSK9 forces that our estimates might be close Eur Heart J 2014;35:2249–2259 Inhibition in Heterozygous Familial Hypercholes- to the actual impact of blood cholesterol 3. Besseling J, Kastelein JJ, Defesche JC, Hutten terolemia Disorder (RUTHERFORD) randomized reduction on blood glucose homeostasis. BA, Hovingh GK. Association between familial trial. Circulation 2012;126:2408–2417 Main limitations that must be borne in hypercholesterolemia and prevalence of type 2 15. Bays H, Gaudet D, Weiss R, et al. Alirocumab – as add-on to atorvastatin versus other lipid treat- mind in this meta-analysis are study-level diabetes mellitus. JAMA 2015;313:1029 1036 4. Roehrich ME, Mooser V, Lenain V, et al Insulin- ment strategies: ODYSSEY OPTIONS I randomized nature, the rarity of the clinical outcome, secreting beta-cell dysfunction induced by human trial. J Clin Endocrinol Metab 2015;100:3140– the inclusion of unbalanced trials with . J Biol Chem 2003;278:18368–18375 3148 ezetimibe versus PCSK9i (1.6% of sample 5. Sullivan D, Olsson AG, Scott R, et al. Effect of a 16. Koren MJ, Scott R, Kim JB, et al. Efficacy, size), and the lack of information on the monoclonal antibody to PCSK9 on low-density safety, and tolerability of a monoclonal antibody to proprotein convertase subtilisin/kexin type 9 as concomitant use and up- or down-titration lipoprotein cholesterol levels in statin-intolerant patients: the GAUSS randomized trial. JAMA 2012; monotherapy in patients with hypercholestero- of antidiabetic medications. Pooling raw 308:2497–2506 laemia (MENDEL): a randomised, double-blind, data from PCSK9i RCTs would be the best 6. Kastelein JJP, Ginsberg HN, Langslet G, et al. placebo-controlled, phase 2 study. Lancet 2012; approach to unveil the particularities of this ODYSSEY FH I and FH II: 78 week results with 380:1995–2006 metabolic interaction under a great robust- alirocumab treatment in 735 patients with het- 17. Kereiakes DJ, Robinson JG, Cannon CP, et al. Efficacy and safety of the proprotein convertase ness. Besides that, the data presented in erozygous familial hypercholesterolaemia. Eur Heart J 2015;36:2996–3003 subtilisin/kexin type 9 inhibitor alirocumab this study are the best approximation 7. Robinson JGG, Farnier M, Krempf M, et al.; among high cardiovascular risk patients on maxi- available for glimpsing the relationship ODYSSEY LONG TERM Investigators. Efficacy and mally tolerated statin therapy: The ODYSSEY between blood cholesterol and glucose. safety of alirocumab in reducing lipids and car- COMBO I study. Am Heart J 2015;169:906–915. In conclusion, the current study indi- diovascular events. N Engl J Med 2015;372:1489– e13 18. Cannon CP, Cariou B, Blom D, et al.; ODYSSEY cates that treatment with PCSK9i is associ- 1499 8. Kiyosue A, Honarpour N, Kurtz C, Xue A, COMBO II Investigators. Efficacy and safety of ated with increased blood glucose at mean Wasserman SM, Hirayama A. A phase 3 study of alirocumab in high cardiovascular risk patients follow-up of 1.50 years. The effect on evolocumab (AMG 145) in statin-treated Japa- with inadequately controlled hypercholesterolae- diabetes risk is only apparent among in- nese patients at high cardiovascular risk. Am J mia on maximally tolerated doses of statins: the dividuals who achieved very low levels of Cardiol 2016;117:40–47 ODYSSEY COMBO II randomized controlled trial. 9. Sabatine MS, Giugliano RP, Wiviott SD, et al.; Eur Heart J 2015;36:1186–1194 LDL-C after treatment. Open-Label Study of Long-Term Evaluation 19. Farnier M, Jones P, Severance R, et al. Efficacy against LDL Cholesterol (OSLER) Investigators. Ef- and safety of adding alirocumab to ficacy and safety of evolocumab in reducing lipids versus adding ezetimibe or doubling the rosuvas- Funding. A.C.S. was supported by a fellowship and cardiovascular events. N Engl J Med 2015; tatin dose in high cardiovascular-risk patients: The grant of research productivity from the Brazilian 372:1500–1509 ODYSSEY OPTIONS II randomized trial. Atheroscle- National Research Council. 10. Roth EM, McKenney JM. ODYSSEY MONO: rosis 2016;244:138–146 Duality of Interest. No potential conflicts of in- effect of alirocumab 75 mg subcutaneously every 20. Sattar N, Preiss D, Murray HM, et al. Statins terest relevant to this article were reported. 2 weeks as monotherapy versus ezetimibe over and risk of incident diabetes: a collaborative Author Contributions. L.S.F.d.C. was responsi- 24 weeks. Future Cardiol 2015;11:27–37 meta-analysis of randomised statin trials. Lancet ble for study design, data collection, data anal- 11.BlomDJ,HalaT,BologneseM,etal.; 2010;375:735–742 ysis, data interpretation, figures, and writing. DESCARTES Investigators. A 52-week placebo- 21. Lotta LA, Sharp SJ, Burgess S, et al. Association A.M.C. was responsible for the literature search, controlled trial of evolocumab in hyperlipidemia. between low-density lipoprotein cholesterol- data collection, figures, and manuscript review. N Engl J Med 2014;370:1809–1819 lowering genetic variants and risk of type 2 A.C.S. was responsible for study design, data 12. Stein EA, Gipe D, Bergeron J, et al. Effect of a diabetes: a meta-analysis. JAMA 2016;316:1383– interpretation, writing, and final review of the monoclonal antibody to PCSK9, REGN727/ 1391