Valeria Fadda

Neurological adverse events of new generation sodium blocker antiepileptic drugs. Meta-analysis of randomized, double-blinded studies with eslicarbazepine acetate, lacosamide and oxcarbazepine.

PURPOSE Analysis of overall tolerability and neurological adverse effects (AEs) of eslicarbazepine acetate (ESL), lacosamide (LCM) and oxcarbazepine (OXC) from double-blind, placebo-controlled trials. Indirect comparisons of patients withdrawing because of AEs, and the incidence of some vestibulocerebellar AEs between these three antiepileptic dugs (AEDs). METHODS We searched MEDLINE for all randomized, double-blind, placebo-controlled trials investigating therapeutic effects of fixed oral doses of ESL, LCM and OXC in patients with drug resistant epilepsy. Withdrawal rate due to AEs, percentages of patients with serious AEs, and the proportion of patients experiencing any neurological AE, nausea and vomiting were assessed for their association with the experimental drug. Analyses were performed between recommended daily doses of each AED according to the approved summary of product characteristics (SPC). Risk differences were used to evaluate the association of any AE [99% confidence intervals (CIs)] or study withdrawals because of AEs (95% CIs) with the experimental drug. Indirect comparisons between withdrawal rate and AEs dizziness, coordination abnormal/ataxia and diplopia were estimated according to network meta-analysis (Net-MA). RESULTS Eight randomized, placebo-controlled, double-blind trials (4 with ESL, 3 with LCM, and 1 with OXC) were included in our analysis. At high doses (OXC 1200mg, ESL 1200mg and LCM 400mg) there was an increased risk of AE-related study withdrawals compared to placebo for all drugs. Several AEs were associated with the experimental drug. Both number and frequency of AEs were dose-related. At high recommended doses, patients treated with OXC withdrew from the experimental treatment significantly more frequently than patients treated with ESL and LCM. Furthermore, the AEs coordination abnormal/ataxia and diplopia were significantly more frequently observed in patients treated with OXC compared to patients treated with LCM and ESL. CONCLUSIONS The overall tolerability of AEDs and the incidence of several neurological AEs were clearly dose-dependent. Indirect comparisons between these AEDs, taking into account dose-effect, showed that OXC may be associated with more frequent neurological AEs than LCM and ESL.

Results can be summarised in a simple figure

Studies based on network meta-analysis are increasingly being published,1 but little attention has been paid to how the information generated is best summarised graphically. Most authors favour multiple Forest plots, with some …

Anti-platelet treatments in acute coronary syndrome: Simplified network meta-analysis

Left ventricular systolic and diastolic function improve after acute myocardial infarction treated with acute percutaneous coronary intervention, but are not influenced by intracoronary injection of autologous mononuclear bone marrow cells: a 3 year serial echocardiographic substudy of the randomized-controlled ASTAMI study. Eur J Echocardiogr 2011;12:98–106. [9] Henderson NC, Mackinnon AC, Farnworth SL, Kipari T, Haslett C, Iredale JP, et al. Galectin-3 expression and secretion links macrophages to the promotion of renal fibrosis. Am J Pathol 2008;172:288–98. [10] Henderson NC, Mackinnon AC, Farnworth SL, Poirier F, Russo FP, Iredale JP, et al. Galectin-3 regulates myofibroblast activation and hepatic fibrosis. Proc Natl Acad Sci U S A 2006;103:5060–5. [11] Henderson NC, Sethi T. The regulation of inflammation by galectin-3. Immunol Rev 2009;230:160–71. [12] de Boer RA, Voors AA, Muntendam P, van Gilst WH, van Veldhuisen DJ. Galectin-3: a novel mediator of heart failure development and progression. Eur J Heart Fail 2009;11:811–7. [13] Shewan LG, Coats AJ. Ethics in the authorship and publishing of scientific articles. Int J Cardiol 2010;144:1–2.

Temporal trend of short-term mortality in severely ill patients receiving parenteral glutamine supplementation

The meta-analysis by Bollhalder et al.1 has systematically reviewed all placebo-controlled randomized trials evaluating intravenous glutamine. Glutamine supplementation was shown to reduce short-term mortality in the subgroup of critical patients (pooled relative risk [RR] 1⁄4 0.71; 95% confidence interval [CI]: 0.53 to 0.96; 7 studies), but in the overall analysis this result was at limits of statistical significance (pooled RR 1⁄4 0.89; 95% CI: 0.77 to 1.04; 22 studies). Since Bollhalder et al.1 did not study whether their results on short-term mortality were affected by any temporal trend, we undertook one such analysis based on standard metaregression techniques.2,3 Our metaregression was aimed at examining the same studies included by Bollhalder et al.1; also the main end-point (short-term mortality defined as in-hospital mortality) was the same. Covariates assessed by our metaregression included publication date, glutamine dose, and setting; a stepwise approach was employed that retained in the model all covariates showing p < 0.10. Our metaregression (based on the overall set of 22 randomized studies) identified publication date as the only covariate that was retained in themodel (p1⁄4 0.031). The other 2 covariates (glutamine dose and setting) were left out with p-values of 0.91 and 0.82, respectively. Interestingly enough, the covariate of setting showed no significant impact on short-term mortality. Figure 1 summarizes these results. Panel A shows the temporal trend of short-term mortality for patients receiving glutamine, while Panel B shows the same information for controls. In the

Nephrotoxicity of Different Formulations of Amphotericin B: Summarizing Evidence by Network Meta-analysis

TO THE EDITOR—The nephrotoxicity of amphotericin B (AmB) has long been known, but the risk of this side effect can be reduced if the drug is formulated as either an extemporaneous lipid emulsion (AmB-LE) or a liposomal product (AmBLIPO). Information on the relative nephrotoxicity of these formulations is still scarce, and so assessing whether AmBLIPO and/or AmB-LE are less nephrotoxic than conventional AmB (ie, AmB deoxycholate, AmB-Dx) is still a matter of interest [1]. This assessment, however, is hampered by the lack of head-tohead trials comparing AmB-LIPO versus AmB-LE; hence, any comprehensive analysis in this area must rely also on indirect comparisons between the 2 lipid-based formulations. Network meta-analysis is an evidencebased tool that is specifically aimed at these indirect comparisons [2–4]. On the other hand, the similar pattern of nephrotoxicity observed for AmB-LIPO versus AmB-LE [1] raises another methodological question because the indirect comparison between these 2 formulations should be based on a noninferiority design [5].

Erythropoiesis-stimulating agents in heart failure: no proof of effectiveness or proof of no effectiveness?

The editorial by Kleijn et al. provides an exhaustive update on the controversy surrounding the use of erythropoietin-stimulating agents (ESAs) in heart failure (HF). Several small studies and one meta-analysis had previously suggested a beneficial effect, but the recent large-scale trial found no benefit, given that both the composite endpoint and all-cause mortality were not affected by ESAs. The editorial therefore concludes that ‘anaemia may [. . ..] be a marker of disease severity rather than a therapeutic target in patients with HF, and ESA treatment for the correction of anaemia is therefore not recommended.’ One might argue whether a single large-scale trial, as opposed to numerous previous small studies, can be sufficient to settle this controversy. We have therefore addressed this question by applying trial sequential analysis (TSA). The main advantage of TSA is that this technique helps in interpreting controversial findings by classifying results according to four mutually exclusive categories (superiority, inferiority, futility, or inconclusive). Our TSA had the purpose of re-examining, based on all-cause mortality, the 10 trials studied in the meta-analysis by Kotecha et al. In particular, we added to these trials the results of Swedberg et al., thus obtaining a data set of 11 studies (3042 patients). We assumed two-sided testing, type-1 error1⁄4 5%, power1⁄4 80%, event frequency for controls 1⁄4 14.8%, and anticipated a relative risk reduction of 50%. A cumulative z-curve graph was constructed (with boundaries for superiority, inferiority, or futility according to the O’Brien–Fleming alpha-spending function). The results of our TSA (Figure 1) clearly indicate futility of ESA treatment, i.e. proof of no effectiveness. This is more informative than the mere conclusion of no proof of effectiveness. In summary, our results fully support the conclusion by Klein et al., and the main implication of our TSA is that further trials on this issue are not recommended.

Antipsychotic drugs for relapse prevention in schizophrenia

www.thelancet.com Vol 380 September 22, 2012 1055 trial requires it to be viewed in the context of the totality of evidence; the estimates of eff ect seen in IST-3 were comparable with, and reinforced the results of, previous trials. Fatovich and colleagues state that we had decided “ordinal analysis was not appropriate”; this was the view of the steering committee in 2009. However, as stated in the statistical analysis plan, that decision was altered in 2011, and the ordinal analysis was added, before the investigators were unmasked to the data, as a secondary outcome, because empirical evidence had emerged to show that the ordinal method was not only statistically more effi cient (crucial given that the sample size had been reduced relative to the original target) but also robust against even substantial deviations from the proportional odds assumption. Smith believes that we changed the primary measure of outcome from the modifi ed Rankin Scale to the Oxford Handicap Scale (OHS). We did not; we merely clarifi ed that we would use the 1990 modifi cation of the Rankin Scale (now referred to as the OHS) throughout. We agree with David Barer, David Curtis, and Newman and Shreves that discussions with patients and their families about the immediate risks of death and long-term benefi ts of thrombolysis should be informed by good quality data. IST-3 on its own lacks statistical power to provide reliable answers on whether rt-PA is more benefi cial in specifi c categories—eg, severe strokes—or perhaps harmful at 3·0– 4·5 h; such answers could emerge from the individual patient data metaanalysis by the Stroke Thrombolysis Trialists’ Collaboration, which will also provide the independent scrutiny suggested by Smith.

Risk of intracranial haemorrhage in patients with atrial fibrillation treated with novel oral anticoagulants: testing the equivalence margins between dabigratran, rivaroxaban and apixaban

To the Editor—In evaluating novel oral anticoagulants (NOACs) in atrial fibrillation, the efforts of many researchers worldwide have permitted the advantages of these new drugs over warfarin, in terms of effectiveness and safety, to be reliably determined by direct head-to-head comparisons [1]. As regards safety, incidence rates of intracranial haemorrhage (ICH) among patients treated with NOACs or warfarin in different populations have been explored by several studies, and a significant reduction of ICH has been found with NOACs [1, 2]. On the other hand, the issue of whether the presently approved NOACs (dabigratran, rivaroxaban, and apixaban) differ in their respective risk of ICH has been investigated to a lesser extent. The most comprehensive study in this field is the meta-analysis by Chatterjee and co-workers [2], in which five trials comparing a NOAC with warfarin were included, as well as one trial comparing apixaban with aspirin; each of the NOACs significantly reduced the risk of ICH compared with warfarin or aspirin, while no significant difference was found between specific NOACs by indirect comparison [2]. On the other hand, many publications try to state equivalence, or even superiority, in the risk of ICH between the approved NOACs, but these statements are not supported by specific data. An increasing amount of research [3] has recently been aimed at differentiating between “no proof of difference” (i.e. failed demonstration of superiority) and “proof of no difference” (demonstration of non-inferiority/equivalence or futility, where futility is essentially represented by equivalence with no treatment). Clearly, proof of no difference is a much more informative result than no proof of difference; however, the former requires that an equivalence (or non-inferiority margin) is incorporated in the analysis. To study the differences between individual NOACs in ICH incidence, we re-analyzed the results by Chatterjee and co-workers [2], and in particular, we carried out a formal equivalence testing based on confidence intervals (CIs). Our results were expressed as risk differences (RDs), the values of which were incorporated in a standard equivalence graph [4–6]. In handling these equivalence tests, we did not declare any predefined margin of equivalence, but we carried out a post-hoc analysis in which we determined which equivalence margins were compatible with the conclusion of equivalence (see our Supplementary document for methodological details). Figure 1 summarises, in Panel A, the results of our network meta-analysis; Panel B shows our equivalence testing. In light of these findings, no statement of superiority of a single NOAC over the others is justified based on indirect comparisons in different populations; this conclusion (no proof of difference) is in keeping with HTA reports from Canada [7] and Norway [8]. More interestingly, our results indicate that the equivalence between the four treatments can be considered demonstrated only if the margin is extended up to an unreasonable RD of ±10.2 %. Therefore, stating that these agents are equivalent in their risk of ICH is not currently justified, because this Electronic supplementary material The online version of this article (doi:10.1007/s00228-014-1644-7) contains supplementary material, which is available to authorized users. A. Messori :D. Maratea :V. Fadda : S. Trippoli Health Technology Assessment Unit, ESTAV Toscana Centro, Regional Health Service, 50100 Firenze, Italy

New and old anti-thrombotic treatments for patients with atrial fibrillation

• The effectiveness of new oral anticoagulants in atrial fibrillation is extensively being studied, but also innovative medical devices aimed at this disease condition need to be evaluated (e.g. Watchman). • Regardless of the type of treatment, the same end-points (stroke or systemic embolism, death from any cause) have been assessed in clinical trials so that meta-analysis techniques can be applied. • Our study employed network meta-analysis to synthetise the effectiveness data of warfarin, new oral anticoagulants, and Watchman and showed the potential advantages of the new treatments as well as the economic implications in terms of budget impact.

Effectiveness and Cost Effectiveness of Bevacizumab in Metastatic Colorectal Cancer

TO THE EDITOR: In the comment by Saltz concerning the cost-effectiveness study on bevacizumab in metastatic colorectal cancer (MCRC) by Goldstein et al, two points deserve additional scrutiny. First, Goldstein et al employed the pivotal trial on bevacizumab in MCRC as the source of effectiveness data for the estimation of the incremental benefit of the drug, and correctly posed the question of whether the data from a single trial could be suitable. In 2014, we published a meta-regression analysis on treatments for MCRC in which all randomized trials conducted between 2000 and 2012 were included (data were from 52 randomized trials with 130 study arms). The improvement in overall survival over this time interval was estimated to be 3.9 months per patient. By using meta-regression, we tried to parcel out the magnitude of the benefit directly related to bevacizumab (as opposed to the benefit related to other factors, such as favorable time trends or type of chemotherapy), and we estimated a value of 1.66 months per patient attributable to bevacizumab (data were from 17 patient arms treated with bevacizumab and 113 patient arms treated without bevacizumab). Interestingly, our estimate of incremental benefit, which was based on a series of numerous randomized trials, was nearly identical to the value determined by Goldstein et al from the pivotal trial (1.68 months per patient). Second, the need to identify practical procedures that can lead to drug price reductions is, of course, a priority in this field. From this point of view, the Italian experience carried out by our National Medicines Agency (ie, AIFA) is interesting, because the agency has chosen to keep the drug prices of anticancer agents close to the prices of these agents in other European countries to avoid international parallel exports driven by price differences. At the same time, the agency has settled with the manufacturers a series of payment-by-results agreements that refund drug costs for all occurrences of clinical failure (ie, paybacks). These paybacks, which are quantified according to the nationwide success rates observed with these drugs in the Italian hospitals, have led to the creation of national registries that include all treated patients. For bevacizumab, the national registry was set up in June 2008 and has since enrolled a total of 9,678 patients treated for MCRC plus 8,720 patients treated for other clinical indications of the drug (information up to date as of March 8, 2015). When the same drug is approved for two or more clinical indications, one advantage of this approach is that the cost-effectiveness ratio is tailored to the benefit expected for each individual indication; this allows us to keep the same nominal price for the different indications of a given drug and, at the same time, to increase the amount of refunded money for the indications when a small benefit is expected. Despite some practical problems that hampered the management of these registries, paybacks are now an important component of our pharmaceutical sustainability. In an Italian study on the cost effectiveness of newly marketed anticancer agents (from 2010 to 2013), the incremental cost-effectiveness ratios (ICERs) were more favorable in Italy (including adjustments for paybacks when applicable) than those reported in the United States for the same agents. With bevacizumab, the drug cost estimate employed by Goldstein et al was