Alberto Donzelli

Clinical Efficacy and Safety of Ezetimibe on Major Cardiovascular Endpoints: Systematic Review and Meta-Analysis of Randomized Controlled Trials

Background Randomized clinical trials (RCTs) about Ezetimibes efficacy on patient-oriented outcomes have given discordant results. The aim of this study was to determine the net effect of Ezetimibe and of the widely marketed combination, Ezetimibe+simvastatin, on mortality and morbidity outcomes. Methods and Findings We searched for RCT on Ezetimibe using MEDLINE, CCTR, EMBASE, ClinicalTrials.gov databases up to December 2013, Merck and Novartis online registers, and personal communications. Two authors independently selected trials fulfilling these criteria: RCTs comparing Ezetimibe±statin or another lipid-lowering drug against placebo, or against the same lipid-lowering drug at the same dosage, with a follow-up at least 24 weeks and one or more of these outcomes: all-cause mortality, cardiovascular (CV) mortality, stroke, myocardial infarction (MI), cancer, serious adverse events (SAEs); we assessed the risk of bias using the Cochrane checklist. We extracted the data for major clinical events as a dichotomous measure, with the patient the unit of analysis. Pooled analysis was done with random and fixed effect based models. Trials comparing Ezetimibe plus a lipid-lowering drug against the same lipidlowering drug representing the net effect of Ezetimibe, showed a nonsignificant tendency toward damage for cancer, MI, stroke and SAEs. Ezetimibe+simvastatin vs. simvastatin alone showed a stronger tendency towards a higher risk for all-cause death (2.52; 0.65-9.74), CV death (3.04; 0.48-19.21), non-CV death (3.03; 0.12-73.50), MI (1.91; 0.42-8.70), stroke (2.38; 0.46-12.35), cancer (RR 11.11; 0.62-198.29), and SAEs (1.45; 0.95-2.23). Limitations include small numbers of events and inadequate power of the pooling. Trials comparing Ezetimibe+simvastatin vs placebo showed non-significant effects: MI (0.81; 0.66-1.00 p = 0.051), all-cause death (1.02; 0.95-1.09), CV death (0.91; 0.80-1.04), non-CV death (108; 0.99-1.18), stroke (0.86; 0.72-1.04), cancer (1.18; 0.80-1.74), SAEs (1.01; 0.96-1.06). Conclusions Ezetimibe±simvastatin had inconsistent effects on important outcomes. No firm conclusions are possible, but findings indicative of damage suggest much more selective use of Ezetimibe±simvastatin.

Influenza Vaccination in the First Trimester of Pregnancy and Risk of Autism Spectrum Disorder

Influenza Vaccination in the First Trimester of Pregnancy and Risk of Autism Spectrum Disorder To the Editor In the cohort study by Zerbo et al “Association Between Influenza Infection and Vaccination During Pregnancy and Risk of Autism Spectrum Disorder,”1 the authors found that “influenza vaccination anytime during pregnancy was not associated with ASD [autism spectrum disorder] risk,” although the adjusted hazard ratio was of borderline significance (1.10; 95% CI, 1.00-1.21). They conclude that “there was a suggestion of increased ASD risk among children whose mothers received an influenza vaccination in their first trimester” (adjusted hazard ratio, 1.20; 95% CI, 1.04-1.39), “but the association was not statistically significant after adjusting for multiple comparisons.”1 We doubt these reassuring conclusions. First, the multiplicity of the hypotheses tested (for which nominal P values for all comparisons made should be at disposal) corresponds, in the authors’ view, to 4 times of each exposure: overall pregnancy time and 3 trimesters for flu and the same for flu vaccination. Nevertheless, the authors used the Bonferroni correction, which only adjusts for independent comparisons: we doubt that overall pregnancy time can be independent from the 3 trimesters in the same individual. Second, Bonferroni correction is more stringent than other corrections, and other equally legitimate tests might maintain the hazard ratio significance. Third, it is biologically plausible that eventual effects on the nervous system occur in the first trimester during its embryogenesis. Therefore, it could be logical to perform the main analysis in the first trimester. However, the main problem is that the influenza vaccination induces an inflammatory response during pregnancy,2,3 which the authors point out. Magnitude and duration of this response is lower and shorter than that induced by the viral infection, but their rates are very different: a Cochrane review4 demonstrated that only 1 to 3 influenza illnesses for every 100 healthy adults vaccinated are prevented (depending by the annual matching between vaccine and circulating strains): number needed to vaccinate, 33 to 100. Therefore, at a population level, 100 vaccinated women will probably experience more inflammatory effects vs those caused by only 1 to 3 influenzas (focused only on 1 to 3 of these 100 women). Table 31 shows that the inflammatory effect of flu during the first trimester interested only 2 of 1000 women (443 of 196 929), while the vaccination flogosis was distributed among 297 of 1000 vaccinated women (13 477 of 45 231). The logical conclusion, pending properly designed trials for further clarification, should not be “do not change vaccine policy or practice” for pregnant women (without proven important benefits, except for patients with coronary heart disease5). Instead, given the possible risk of 4 additional autism spectrum disorder cases for every 1000 women vaccinated, it seems reasonable to apply the precautionary principle and refrain from vaccinating pregnant women or at least to avoid vaccination in the first trimester of pregnancy.

Efficacy of ezetimibe: A real effect?

We do not agree with the methods and the conclusions of the systematic review of Savarese et al. [1]. Their meta-analysis of seven randomized controlled trials tests the safety and the efficacy of ezetimibe (E), added to a statin, in comparison indifferently versus placebo or active treatment and shows a favorable effect of E on risks of major non-fatal endpoints (myocardial infarction RR: 0.865, 95% CI: 0.801–0.934; stroke RR: 0.840, 95% CI: 0.744–0.949), with a neutral effect on mortality outcomes (all-cause death RR: 1.003, 95% CI: 0.954– 1.055; CV death RR: 0.958, 95% CI: 0.879–1.044) and new cancer risk (RR: 1.040, 95% CI: 0.965–1.120). The authors emphasize the need to assess the safety and efficacy of E using “robust available information”; nevertheless their pooling was a mix of comparisons versus active drug and versus placebo. This method does not allow any real quantification of the net effect of E. In our metaanalysis [2] we conducted these analyses in separate poolings, finding a not significant effect for all the outcomes considered, both versus active comparators (E/statin combination versus the same statin at the same dosage) and versus placebo (E/statin combination versus placebo). When ourmeta-analysis was carried out, the data of IMPROVE-IT [3] (E/simvastatin versus simvastatin at the same dosage) were not yet available; adding these data to our calculations of E/statin versus active comparators (results not yet published), we have found a significant though modest reduction of the myocardial infarction risk (RR: 0.88,

Helicobacter pylori Eradication

3 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 Dear Editors: The net effect of Helicobacter pylori eradication in asymptomatic individuals can be unfavorable. The systematic review associates eradication of H pylori infection with a reduced incidence of gastric cancer. The benefits of eradication vary with baseline gastric cancer incidence (which ranged from 34.3 to 10,256.4 per 100,000 personsyears), but the authors state that they apply to all levels of baseline risk. They claim that

Questioning the Benefits of Statin Therapy in Older People without Established Cardiovascular Disease

Noaman and colleagues [1] found that current available research information about whether statins contribute to a reduction in morbidity and mortality is unclear in the elderly, particularly in the primary prevention setting. While agreeing with this statement, we would further criticise the results of a recent, well publicised meta-analysis [2], not mentioned by Noaman and colleagues, which suggested a favourable effect of statins on major cardiovascular (CV) events in older people without established cardiovascular disease. In fact, this meta-analysis [2] suffers from various methodological pitfalls. The authors declared absence of publications bias, but details of this statement were not provided. In absence of a report of single outcomes’ frequencies, we have calculated the log RRs and their Standard Errors from the confidence intervals. Indeed, the Egger test [3] shows a significant presence of publication bias for the outcome ‘myocardial infarction’ (p = 0.013) and a very suspicious result for the outcome ‘stroke’ (p = 0.077). The authors reported (and we confirm) an excess of heterogeneity in the analysis of the first outcome (Q test p = 0.028), without further investigations about its causes. Furthermore, the authors [2] did not consider important information: for example they excluded the PROSPER [4] primary prevention data from their analysis of all causes mortality, although this information has been available since 2010 [5]. In fact, in the PROSPER trial, pravastatin therapy in primary prevention has been associated with an increase in all-cause mortality [5], although statistically non significant (RR = 1.08 p = 0.26). After redesigning the pooling including these latter data, we have found a slight decrease of the pooled effect (RR = 0.96 p = 0.39, instead of RR = 0.94). Moreover, the only two randomised trials designed to recruit only elderly people [4,6] have shown a trend of an increased mortality from all causes in primary prevention (RR 1.08 p = 0.26 with pravastatin [4], RR = 1.02 p = 0.98 with fluvastatin [6]). Of note, the cited meta-analysis [2] strictly confirmed previous findings [5], namely the absence of any efficacy of statin drugs in the reduction of all-cause mortality in the primary prevention setting. The absence of efficacy of statins in prolonging survival has been related to shortness of follow-up [2]. However, the HPS trial, although not reporting separate primary and secondary results, was performed in individuals with high CV baseline risk (median predicted five-year risk of major vascular events of 18.6%), with the same median basal mortality rate, and demonstrated a survival beneficial effect of statin therapy after just one year [7]. Furthermore, it has been calculated that, notwithstanding the limitations [8] related to the different lengths of follow up of the trials included in the meta-analysis [2], the number needed to treat (NNT) for the entire mean follow-up of 3.5 years was as high as 83 to prevent one myocardial infarction, and 142 to prevent one stroke [9]. In the light of present uncertainty of the risk/benefit ratio in primary prevention, we concur with Noaman and colleagues that in older people without CV disease, statin therapy should be considered with caution, also considering the high prevalence of side-effects of these drugs in the real world, and the well-known inverse relationship between mortality and LDL cholesterol in the elderly [10].

Non-specific effects of vaccinations in high-income settings: How to address the issue?

ABSTRACT “Non-specific effects” of vaccines go beyond the specific protective effects against the targeted diseases. They, if real, could theoretically be beneficial, neutral or negative. This article intends to answer the following questions: Do the non-specific effects of vaccines exist? Almost certainly yes, and they can be important in low-income countries Are non-specific effects also present in high-income countries? At least to some extent, it seems quite logical Can non-specific effects be systematically identified by the current systems of side effects/unintended reactions monitoring? Most likely not Could the Institute of Medicine proposals and some ongoing attempts solve the issue? It seems unlikely Could there be better, feasible and ethically acceptable ways to achieve the aforementioned objective? A proposal is presented about this issue, with the potential both to solve the problem with the most valid methods, and to overcome the ethical problems that have so far precluded the adoption of RCTs to study possible vaccine non-specific effects, monitored by long follow-up.

Comment on Itoh et al. Intensive Treat-to-Target Statin Therapy in High-Risk Japanese Patients With Hypercholesterolemia and Diabetic Retinopathy: Report of a Randomized Study. Diabetes Care 2018;41:1275–1284

The commercially funded randomized study by Itoh et al. (1), comparing treat-to-target approaches in patients with diabetic retinopathy with hypercholesterolemia (baseline LDL cholesterol [LDL-C] ∼106 mg/dL), “found no significant decrease in CV [cardiovascular] events or CV-associated deaths with intensive therapy.” The authors speculated about a nonsignificant decrease “possibly because the between-group difference of LDL-C was lower than expected (27.7 mg/dL at 36 months of treatment).” The intensive statin therapy targeted LDL-C <70 mg/dL, while the standard one targeted 100–120 mg/dL. The primary outcome, a composite of CV events, tended to occur less often in the intensive group: in 129 patients versus 153 in the standard group (hazard ratio [HR] 0.84, 95% CI 0.67–1.07). An exploratory finding also showed fewer cerebral events (HR 0.52, …

The systematic review of randomized controlled trials of PCSK9 antibodies challenges their “efficacy breakthrough” and the “lower, the better” theory

Abstract Background: A Cochrane review with meta-analysis showed controversial results about the efficacy of PCSK9 antibodies in the prevention of cardiovascular diseases. This review gives the opportunity to test the relationship between LDL-C levels and cardiovascular events. Methods: The authors analyzed the relationship between the calculated LDL-C lowering and the risk of all-cause mortality, fatal and non-fatal myocardial infarction, fatal and non-fatal stroke, any adverse event, cardiovascular events and cardiovascular disease mortality. Results: No beneficial relationship was found between LDL-C lowering and cardiovascular events explored by meta-regression; instead, there was a trend toward harm. For any of the other outcomes there was no significant association between LDL-C lowering and risk. Furthermore, the authors calculated the efficacy that would be expected through the LDL-C lowering showed in the meta-analysis, considering widely accepted predictions. These were respected only for stroke, while the observed efficacy on other cardiovascular events was significantly lower than the expected, and no significant result was observed at all for fatal outcomes. A separate meta-analysis of trials recruiting familial hypercholesterolemia patients have showed a tendency to harm for almost all outcomes. Conclusions: The relationship between LDL-C lowering and cardiovascular events has not showed any significant association (and even a tendency toward harm), challenging the “lower the better” theory. A separate meta-analysis of trials recruiting familial hypercholesterolemia patients has showed a tendency to harm for all outcomes with PCSK9 antibodies. Therefore, at the moment, the data available from randomized trials does not clearly support the use of these antibodies.

Healthy-vaccinated effect

In the study by Casado and colleagues,[1][1] older inpatients with laboratory-confirmed severe and nonsevere influenza were each matched with three controls who had been hospitalized for causes other than acute respiratory diseases. The adjusted effectiveness of influenza vaccination in patients

PCSK9-AB and risk of neurocognitive events – comments to Robinson’s meta-analysis

Second, limiting the evaluation to alirocumab therapies and focusing to very-low LDL-Cholesterol values subgroups reduces the power of comparisons and increases the risk of false negatives. The apparent lack of association between the LDL-Cholesterol <25 and <15 mg/dl and the risk of neurocognitive disorders can be due to the scarce power of these analyses. Indeed, the sample size of the groups exposed to those targets [1] can demonstrate with sufficient statistical power (≥80%), only very large hypothetical increases in risk of neurocognitive events (RRs ≥2.6 and ≥3.4 respectively).