Leading beyond regulatory approval: Opportunities for statisticians to optimize evidence generation and impact clinical practice.

The role and value of statistical contributions in drug development up to the point of health authority approval are well understood. But health authority approval is only a true 'win' if the evidence enables access and adoption into clinical practice. In today's complex and evolving healthcare environment, there is additional strategic evidence generation, communication, and decision support that can benefit from statistical contributions. In this article, we describe the history of medical affairs in the context of drug development, the factors driving post-approval evidence generation needs, and the opportunities for statisticians to optimize evidence generation for stakeholders beyond health authorities in order to ensure that new medicines reach appropriate patients.

Parahydrogen-Induced Polarization of a Labeled, Cancer-Targeting DNA Aptamer.

Enhancing NMR signals of biomacromolecules by hyperpolarization offers exciting opportunities for diagnostic applications. However, their hyperpolarization via parahydrogen remains challenging as specific catalytic interactions are required, which are difficult to tune due to the large size of the biomolecule and its insolubility in organic solvents. Herein, we show the unprecedented hyperpolarization of the cancer-targeting DNA aptamer AS1411. By screening different molecular motifs for an unsaturated label in nucleosides and in DNA oligomers, we were able to identify structural prerequisites for the hyperpolarization of AS1411. Finally, adjusting the polarity of AS1411 by complexing the DNA backbone with amino polyethylene glycol chains allowed the hydrogenation of the label with parahydrogen while the DNA structure remains stable to maintain its biological function. Our results are expected to advance hyperpolarized molecular imaging technology for disease detection in the future.

Multilevel network meta-regression for population-adjusted treatment comparisons.

Standard network meta-analysis (NMA) and indirect comparisons combine aggregate data from multiple studies on treatments of interest, assuming that any effect modifiers are balanced across populations. Population adjustment methods relax this assumption using individual patient data from one or more studies. However, current matching-adjusted indirect comparison and simulated treatment comparison methods are limited to pairwise indirect comparisons and cannot predict into a specified target population. Existing meta-regression approaches incur aggregation bias. We propose a new method extending the standard NMA framework. An individual level regression model is defined, and aggregate data are fitted by integrating over the covariate distribution to form the likelihood. Motivated by the complexity of the closed form integration, we propose a general numerical approach using quasi-Monte-Carlo integration. Covariate correlation structures are accounted for by using copulas. Crucially for decision making, comparisons may be provided in any target population with a given covariate distribution. We illustrate the method with a network of plaque psoriasis treatments. Estimated population-average treatment effects are similar across study populations, as differences in the distributions of effect modifiers are small. A better fit is achieved than a random effects NMA, uncertainty is substantially reduced by explaining within- and between-study variation, and estimates are more interpretable.

Combination of several matching adjusted indirect comparisons (MAICs) with an application in psoriasis.

In health technology assessment (HTA), beside network meta-analysis (NMA), indirect comparisons (IC) have become an important tool used to provide evidence between two treatments when no head-to-head data are available. Researchers may use the adjusted indirect comparison based on the Bucher method (AIC) or the matching-adjusted indirect comparison (MAIC). While the Bucher method may provide biased results when included trials differ in baseline characteristics that influence the treatment outcome (treatment effect modifier), this issue may be addressed by applying the MAIC method if individual patient data (IPD) for at least one part of the AIC is available. Here, IPD is reweighted to match baseline characteristics and/or treatment effect modifiers of published data. However, the MAIC method does not provide a solution for situations when several common comparators are available. In these situations, assuming that the indirect comparison via the different common comparators is homogeneous, we propose merging these results by using meta-analysis methodology to provide a single, potentially more precise, treatment effect estimate. This paper introduces the method to combine several MAIC networks using classic meta-analysis techniques, it discusses the advantages and limitations of this approach, as well as demonstrates a practical application to combine several (M)AIC networks using data from Phase III psoriasis randomized control trials (RCT).

Modeling Heterogeneity in Patients' Preferences for Psoriasis Treatments in a Multicountry Study: A Comparison Between Random-Parameters Logit and Latent Class Approaches.

BACKGROUND: Either a random-parameters logit (RPL) or latent class (LC) model can be used to model or explain preference heterogeneity in discrete-choice experiment (DCE) data. The former assumes continuous distribution of preferences across the sample, while the latter assumes a discrete distribution. This study compared RPL and LC models to explore preference heterogeneity when analyzing patient preferences for psoriasis treatments. METHODS: Using DCE data collected from respondents with moderate-to-severe plaque psoriasis, we calculated and compared preference weights derived from RPL and LC models. We then compared how RPL and LC explain preference heterogeneity by exploring differences across subgroups defined by observed characteristics (i.e., country, age, gender, marital status, and psoriasis severity). RESULTS: While RPL and LC models resulted in the same mean preference weights, different preference-heterogeneity patterns emerged from the two approaches. In both models, country of residence and self-reported disease severity could be linked to systematic differences in preferences. The RPL also identified gender and marital status, but not age, as sources of heterogeneity; the LC membership probability model indicated that age was a significant factor, but not gender or marital status. CONCLUSIONS: Using data from a psoriasis patient survey to compare two widely used methods for exploring heterogeneity identified differences in results between stated-preferences: subgroup analysis in the RPL model and inclusion of subgroup characteristics in the class membership probability function of the LC model. Researchers should model data using the most adaptable approach to address the initial study question.

What matters most? Different stakeholder perspectives on estimands for an invented case study in COPD.

In drug development, we ask ourselves which population, endpoint and treatment comparison should be investigated. In this context, we also debate what matters most to the different stakeholders that are involved in clinical drug development, for example, patients, physicians, regulators and payers. With the publication of draft ICH E9 addendum on estimands in 2017, we now have a common framework and language to discuss such questions in an informed and transparent way. This has led to the estimand discussion being a key element in study development, including design, analysis and interpretation of a treatment effect. At an invited session at the 2018 PSI annual conference, PSI hosted a role-play debate where the aim of the session was to mimic a regulatory and payer scientific advice discussion for a COPD drug. Including role-play views from an industry sponsor, a patient, a regulator and a payer. This paper presents the invented COPD case-study design and considerations relating to appropriate estimands are discussed by each of the stakeholders from their differing viewpoints with the additional inclusion of a technical (academic) perspective. The rationale for each perspective on approaches for handling intercurrent events is presented, with a key emphasis on the application of while-on-treatment and treatment policy estimands in this context. It is increasingly recognised that the treatment effect estimated by the treatment policy approach may not always be of primary clinical interest and may not appropriately communicate to patients the efficacy they can expect if they take the treatment as directed.

A statistical method to convert published response rates into marginal distributions with an example application in psoriasis.

Assessment of severity is essential for the management of chronic diseases. Continuous variables like scores obtained from the Hamilton Rating Scale for Depression or the Psoriasis Area and Severity Index (PASI) are standard measures used in clinical trials of depression and psoriasis. In clinical trials of psoriasis, for example, the reduction of PASI from baseline in response to therapy, in particular the proportion of patients achieving at least 75%, 90%, or 100% improvement of disease (PASI 75, PASI 90, or PASI 100), is typically used to evaluate treatment efficacy. However, evaluation of the proportions of patients reaching absolute PASI values (eg, ≤1, ≤2, ≤3, or ≤5) has recently gained greater clinical interest and is increasingly being reported. When relative versus absolute scores are standard, as is the case with the PASI in psoriasis, it is difficult to compare absolute changes using existing published data. Thus, we developed a method to estimate absolute PASI levels from aggregated relative levels. This conversion method is based on a latent 2-dimensional normal distribution for the absolute score at baseline and at a specific endpoint with a truncation to allow for baseline inclusion criterion. The model was fitted to aggregated results from simulations and from 3 phase III studies that had known absolute PASI proportions. The predictions represented the actual results quite precisely. This model might be applied to other conditions, such as depression, to estimate proportions of patients achieving an absolute low level of disease activity, given absolute values at baseline and proportions of patients achieving relative improvements at a subsequent time point.

Cost per additional responder for ixekizumab and other FDA-approved biologics in moderate-to-severe plaque psoriasis.

BACKGROUND: Evidence of the cost-efficacy of ixekizumab for the treatment of moderate-to-severe plaque psoriasis (PsO) in the US is limited. OBJECTIVE: To estimate the number needed to treat (NNT) and monthly cost of achieving one additional Psoriasis Area and Severity Index (PASI) 75, 90, and 100 responder for ixekizumab and other Food and Drug Administration (FDA)-approved biologics in PsO. METHODS: A network meta-analysis estimated the probability of achieving PASI 75, 90, or 100 response during induction for each biologic. NNTs were calculated using response difference of each respective biologic vs placebo at the end of induction. Monthly costs per additional PASI responder were based on FDA-approved doses, wholesale acquisition costs, and induction NNTs. RESULTS: Induction NNTs for ixekizumab 80 mg once every 2 weeks (Q2W) relative to placebo were consistently lower across all levels of clearance compared with the other biologics. Monthly cost per additional responder was lowest for ustekinumab 45 mg at PASI 75 and for secukinumab 300 mg and ixekizumab 80 mg Q2W at PASI 90. Ixekizumab 80 mg Q2W had the lowest cost for PASI 100. CONCLUSION: In this analysis, ixekizumab is the most cost-efficient biologic in the US when targeting complete resolution, as measured by PASI 100 in PsO.

Structured Benefit-risk assessment: a review of key publications and initiatives on frameworks and methodologies.

Introduction The conduct of structured benefit-risk assessment (BRA) of pharmaceutical products is a key area of interest for regulatory agencies and the pharmaceutical industry. However, the acceptance of a standardized approach and implementation are slow. Statisticians play major roles in these organizations, and have a great opportunity to be involved and drive the shaping of future BRA. Method We performed a literature search of recent reviews and initiatives assessing BRA methodologies, and grouped them to assist those new to BRA in learning, understanding, and choosing methodologies. We summarized the key points and discussed the impact of this emerging field on various stakeholders, particularly statisticians in the pharmaceutical industry. Results We provide introductory, essential, special interest, and further information and initiatives materials that direct readers to the most relevant materials, which were published between 2000 and 2013. Based on recommendations in these materials we supply a toolkit of advocated BRA methodologies. Discussion Despite initiatives promoting these methodologies, there are still barriers, one of which being the lack of a consensus on the most appropriate methodologies among stakeholders. However, this opens up opportunities, for statisticians in the pharmaceutical industry especially, to champion appropriate BRA methodology use throughout the pharmaceutical product lifecycle. Conclusions This article may serve as a starting point for discussions and to reach a mutual consensus for methodology selection in a particular situation. Regulators and pharmaceutical industry should continue to collaborate to develop and take forward BRA methodologies, and by clear communication develop a mutual understanding of the key issues. Copyright © 2015 John Wiley & Sons, Ltd.

Nocebo effects in the treatment of major depression: results from an individual study participant-level meta-analysis of the placebo arm of duloxetine clinical trials.

BACKGROUND: The nocebo effect, when a harmless substance creates harmful effects in a person who takes it, is a clinically salient yet seldom studied phenomenon that may be associated with poorer treatment outcomes, perceived adverse events, and treatment discontinuation. The covert presence of nocebo responders in clinical trials may contribute to outcome variance in both placebo and active treatment arms for important primary and secondary endpoints. Nocebo effects are thought to be driven by expectancy and conditioning. METHOD: This study analyzed pooled clinical trial data in the placebo arms of controlled trials of antidepressant medications to investigate variables associated with the emergence of adverse outcomes in placebo-treated participants (N = 2,457). Specifically, we examined treatment-emergent adverse events (TEAEs) and discontinuation in placebo-treated individuals. Trials were commenced between 1993 and 2010 as studies of duloxetine versus active comparator and/or placebo. RESULTS: TEAEs were reported by 1,569 placebo-treated participants (63.9%), with 115 (4.7%) discontinuing from the studies due to TEAEs and 274 (11.2%) showing worsening of Hamilton Depression Rating Scale total score during placebo treatment. There was specifically no evidence to support the expectancy hypothesis, that reported TEAEs were influenced by adverse effects described in the clinical trials participant information and consent forms, or the conditioning hypothesis, that reported TEAEs would be influenced by adverse effect profiles of previous antidepressant medications used by these study participants. There was some evidence to suggest that people who had previously used complementary medications were more likely to report TEAEs. Variables specific to individual studies were the strongest predictors of TEAEs. DISCUSSION: In this study, TEAEs were very common among placebo-treated clinical trial participants. Unexpectedly, there was no evidence to associate TEAEs with adverse clinical outcomes, nor were the conditioning or expectancy hypotheses supported by these data. CONCLUSIONS: The nocebo effect is a common, covert, and poorly understood driver of clinical outcomes that requires further investigation.

Efficacy comparison of duloxetine and SSRIs at doses approved in Japan.

BACKGROUND: Approved doses of antidepressants in Japan are usually lower than those in the USA and European Union, but to date meta-analyses comparing antidepressants have all used the higher doses approved in the USA and European Union and often have used indirect comparisons. The purpose of this study was to conduct an integrated database analysis of patient level data to compare the effects of duloxetine with those of selective serotonin reuptake inhibitors (SSRIs) at the doses approved in Japan. METHODS: Pooled data were analyzed from four randomized, double-blind, placebo-controlled studies that compared duloxetine at the dose range approved in Japan (40-60 mg/day) with other SSRIs (paroxetine 20 mg/day or escitalopram 10 mg/day) and placebo in patients with major depressive disorder. In total, 1,694 patients were included in the analysis (duloxetine, n=688; selective serotonin reuptake inhibitors, n=690; placebo, n=316). The primary outcome measure was the mean change from baseline at week 8 in 17-item Hamilton Rating Scale for Depression (HAMD17) total and subscale scores. RESULTS: Duloxetine and both selective serotonin reuptake inhibitors were superior to placebo in HAMD17 total score at week 8 in both the all-randomized group and the more severe subgroup (HAMD17 total scores ≥19). Duloxetine was superior to SSRIs in improving the HAMD17 Retardation subscale score (least squares mean difference [95% confidence interval]): all-randomized group, -0.33 [-0.60, -0.07], P=0.015; severe subgroup, -0.45 [-0.83, -0.07], P=0.020). CONCLUSION: Within the dose range approved in Japan for patients with major depressive disorder, duloxetine and selective serotonin reuptake inhibitors demonstrated comparable overall efficacy, with a possible advantage for duloxetine in improving loss of energy and interest. To the best of our knowledge, this analysis is unique not only in evaluating dosages specific to Japan, but also in using individual patient data and the same endpoint across studies to allow for strictly direct head-to-head data comparisons as opposed to pooling direct and indirect comparisons.

Does Comorbid Disruptive Behavior Modify the Effects of Atomoxetine on ADHD Symptoms as Measured by a Continuous Performance Test and a Motion Tracking Device?

OBJECTIVE: To compare the reduction of ADHD symptoms under atomoxetine in patients with and without comorbid oppositional defiant disorder (ODD) or conduct disorder (CD) using a computer-based continuous performance test (cb-CPT) combined with an infrared motion tracking (MT) device. METHOD: Secondary analysis of a placebo-controlled study in ADHD patients (6-12 years old) treated with atomoxetine (target dose: 1.2 mg/kg per day). Cb-CPT/MT scores were analyzed using ANCOVA (last observation carried forward [LOCF]). RESULTS: The data (N = 125) suggested a more pronounced atomoxetine effect in the group with comorbid ODD/CD as measured by all cb-CPT/MT parameters except for "normalized variation of reaction time" (nVRT). CONCLUSION: The results showed that atomoxetine reduced ADHD severity as measured by cb-CPT and MT parameters regardless of whether comorbid ODD/CD was present. The treatment effect of atomoxetine on hyperactivity appears to be more pronounced in the subgroup of patients with comorbid ODD/CD than in the subgroup without this comorbidity.

Neuropathic pain phenotyping as a predictor of treatment response in painful diabetic neuropathy: data from the randomized, double-blind, COMBO-DN study.

Sensory profiles are heterogeneous in neuropathic pain disorders, and subgroups of patients respond differently to treatment. To further explore this, patients in the COMBO-DN study were prospectively assessed by the Neuropathic Pain Symptom Inventory (NPSI) at baseline, after initial 8-week therapy with either duloxetine or pregabalin, and after subsequent 8-week combination/high-dose therapy. Exploratory post hoc cluster analyses were performed to identify and characterize potential subgroups through their scores in the NPSI items. In patients not responding to initial 60 mg/d duloxetine, adding 300 mg/d pregabalin for combination treatment was particularly effective regarding the dimensions pressing pain and evoked pain, whereas maximizing the duloxetine dose to 120 mg/d appeared more beneficial regarding paresthesia/dysesthesia. In contrast, adding 60 mg/d duloxetine to 300 mg/d pregabalin in case of nonresponse to initial pregabalin led to numerically higher decreases in all NPSI dimensions/items compared to maximizing the pregabalin dose to 600 mg/d. Cluster analysis revealed 3 patient clusters (defined by baseline scores for the 10 NPSI sensory items) with different pain profiles, not only in terms of overall pain severity, but also across NPSI items. Mean Brief Pain Inventory average pain improved in all clusters during combination/high-dose therapy. However, in patients with severe pain, the treatment effect showed a trend in favor of high-dose monotherapy, whereas combination therapy appeared to be more beneficial in patients with moderate and mild pain (not significant). These complementary exploratory analyses further endorse the idea that sensory phenotyping might lead to a more stratified treatment and potentially to personalized pain therapy.

Findings from the observational COMPLY study in children and adolescents with ADHD: core symptoms, ADHD-related difficulties, and patients' emotional expression during psychostimulant or nonstimulant ADHD treatment.

The aim of this study was to explore the course of attention-deficit/hyperactivity disorder (ADHD) core symptoms, ADHD-related difficulties, and emotional expression during ADHD pharmacotherapy and associations between them. This prospective, observational study examines pediatric patients with ADHD who newly initiated stimulant, atomoxetine or a combination of both treatments. Data were collected at baseline; weeks 1, 2, and 4; and months 3, 6, 9, and 12. Physicians rated ADHD core symptoms using the ADHD Rating Scale (ADHD-RS); patients, parents, and physicians rated ADHD-related difficulties using the Global Impression of Perceived Difficulties (GIPD) Scale; and patients and parents rated emotional expression using the Expression of Emotion Scale for Children (EESC). Results were analyzed using mixed model repeated measures. Associations are presented by Spearman's correlations. Overall, 504 patients, mean age 9.6 years, 72.6% males, were analyzed. Fifty percent of patients started atomoxetine, 49.0% stimulant and 1% a combination of both. ADHD-RS, GIPD, and EESC scores decreased significantly in both monotherapy groups. Correlations between ADHD-RS and parent- or physician-rated GIPD scores were at-best moderate and increased over time but remained low to moderate for patient-rated GIPD [patient, r=0.43 (95% CI 0.34, 0.51); parent, r=0.58 (0.50, 0.64); physician, r=0.55 (0.48, 0.62)]. Correlations between ADHD-RS and patient- or parent-rated EESC scores were low at baseline (r<0.2) and increased over time mostly for parent ratings [patient, r=0.35 (0.26, 0.44); parent, r=0.41 (0.32, 0.50)]. These data support the effectiveness of ADHD pharmacotherapy. The at-best moderate correlations between ADHD core symptoms and ADHD-related difficulties or emotional expression assessed by different raters indicate potentially important patient outcomes beyond core symptoms.

An ultra-low energy method for rapidly pre-concentrating microalgae.

This study demonstrates that Nannochloropsis sp. can be effectively separated from its growth medium (0.2-0.3g/L) using electro-coagulation-flocculation in a 100mL batch reactor with nickel electrodes and a treatment time of only 4s. Minimum energy density input for effective separation is 0.03 kWh/m(3). Both energy input and treatment time are much smaller than reported elsewhere. The process results in rapid separation of microalgae (over 90% in 120 min) with minimal damage to algal cells (>90% still alive after processing). At around 4V input, algae can be effectively separated even in very low concentrations. Pulsing is equally effective in separating microalgae as continuous direct current of same magnitude and total exposure time. Algae can separate from their growth medium even if the suspension itself is not treated, but is mixed with treated saltwater with same conductivity. The described method has significant advantages including applicability to continuous processing and water reuse.

Patient expectations in the treatment of painful diabetic polyneuropathy: results from a non-interventional study.

OBJECTIVE: Pain control is the main objective when treating patients with painful diabetic peripheral neuropathic pain (DPNP). However, DPNP is associated with further substantial patient burden that often is not appropriately addressed. Our study identified patients' needs and asked patients what they expected from DPNP treatment. METHODS: Baseline data were collected in a German prospective, non-interventional study in patients with DPNP starting or switching pain medication at the discretion of the investigator. DPNP severity was evaluated using Brief Pain Inventory (BPI) and Clinician/Patient Global Impression-Severity (CGI-S/PGI-S). Primary objective of this study was to evaluate for which interference item of the BPI DPNP patients expected most to improve due to DPNP therapy. RESULTS: We enrolled 2,576 patients with DPNP from 307 outpatient centers (mean [standard deviation {SD}] age: 65.8 years [11.5], 51.2% female). Mean (SD) CGI-S and PGI-S at baseline were 4.4 (0.91) and 4.5 (1.05), respectively. BPI average pain score was 5.1 (2.04). The BPI interference score was 4.8 (2.18); items most impaired at baseline were walking ability 5.5 (2.60) and general activity 5.4 (2.37). The most frequently chosen BPI interference items expected to improve as a result of the pain treatment were: General activity (29.3%; 95% confidence interval [CI] 27.5-31.0%) and walking ability (24.4%; 95% CI 22.8-26.1%), followed by sleep (14.7%), enjoyment of life (13.6%), mood (8.3%), normal work (7.7%), and relations with other people (1.9%). CONCLUSIONS: The majority of patients identified "general activity" and "walking ability" as most relevant BPI interference items for which they expect improvement from DPNP treatment.

Depression symptom clusters and their predictive value for treatment outcomes: results from an individual patient data meta-analysis of duloxetine trials.

We evaluated individual patient data from phase II to IV clinical trials of duloxetine in major depressive disorder (MDD) (34 studies, 13,887 patients). Our goal was to identify clusters of patients with similar depressive symptom patterns at baseline, as measured by the 17-item Hamilton Depression Rating Scale (HAMD-17), and to investigate their respective predictive value of outcomes as measured by the HAMD-17 total score. Five clusters were identified at baseline: 1) "Lack of insight"; 2) "Sleep/sexual/somatic"; 3) "Typical MDD"; 4) "Gastrointestinal/weight loss"; and 5) "Mild MDD". However, it should be noted that cluster descriptors are not mutually exclusive. Analyses of the HAMD-17 total score results over time were performed using the 18 randomized placebo and/or actively controlled studies representing 6723 patients. At the end of acute treatment (ranging from 4 to 36 weeks), different levels of effect sizes for active therapy (64.5% duloxetine) vs. placebo were detected by cluster. In 3 out of 5 clusters (representing about 80% of the patients), the effect size was significantly different from 0, in favor of active therapy. The effect size was largest in those clusters with severe somatic symptoms ("Sleep/sexual/somatic" cluster [-0.4170], and "Gastrointestinal/weight loss" cluster [-0.338]). In conclusion, our cluster analysis identified 5 clinically relevant MDD patient clusters with specific mean treatment outcomes. Identification of MDD clusters may help to improve outcomes by adapting MDD treatment to particular clinical profiles.

Antidepressants in the treatment for chronic low back pain: questioning the validity of meta-analyses.

OBJECTIVES: To contrast the analgesic effect of duloxetine with antidepressants reported in other published randomized clinical trials (RCTs) and review articles in patients with chronic low back pain (CLBP). METHODS: In this narrative review, the results of 13 RCTs and 5 systematic reviews examining the analgesic effect of various antidepressants in CLBP were contrasted with those of 3 placebo-controlled duloxetine RCTs. Treatment effects based on the Brief Pain Inventory (BPI) average score in the duloxetine RCTs were assessed in all completers (by study and overall) and in last-observation-carried-forward (LOCF) analyses (extracted from study reports). 30%- and 50%-reduction response rates were compared between duloxetine and placebo. RESULTS: Eleven different antidepressants were examined in 13 individual RCTs. Sample sizes, treatment durations, and analysis methods varied across studies. Reviews each included 5 to 9 of the RCTs and came to different conclusions regarding the analgesic effect of antidepressants: 2 found no evidence while 3 reported some evidence. The completer analysis showed greater improvements in BPI average scores with duloxetine vs. placebo (significant in 2 studies). Overall, the least square mean (standard error) difference between treatments was - 0.7 (0.15) (P < 0.0001). Overall response rates were significantly larger with duloxetine than with placebo. CONCLUSIONS: Due to the diversity of previous studies and the pooling methods used, the conclusions regarding the analgesic effect of antidepressants in CLBP drawn from systematic reviews must be interpreted with caution. Appropriately designed and powered studies similar to recently published duloxetine studies are recommended to demonstrate the analgesic effect of antidepressants.

Impact of patient selection and study characteristics on signal detection in placebo-controlled trials with antidepressants.

An increasing rate of antidepressant trials fail due to large placebo responses. This analysis aimed to identify variables influencing signal detection in clinical trials of major depressive disorder. Patient-level data of randomized patients with a duloxetine dose ≥ 60 mg/day were obtained from Lilly. Total scores of the Hamilton Depression Rating scale (HAM-D) were used as efficacy endpoints. In total, 4661 patients from 14 studies were included in the analysis. The overall effect size (ES), based on the HAM-D total score at endpoint, between duloxetine and placebo was -0.272. Although no statistically significant interactions were found, the following results for factors influencing ES were seen: a very low ES (-0.157) in patients in the lowest baseline HAM-D category and in patients recruited in the last category of the recruitment period (-0.122). A higher ES in patients recruited in centers with a site-size at but not more than 2.5 times the average site-size for the study (-0.345). Study characteristics that resulted in low signal detection in our database were: <80% study completers, a HAM-D placebo response >5 points, a high variability of placebo response (SD > 7 points HAM-D), >6 post baseline visits per study, and use of an active control drug. Simpler trial designs, more homogeneous and mid-sized study sites, a primary analysis based on a higher cutoff blinded to investigators to avoid the influence of score inflation in mild patients and, if possible, studies without an active control group could lead to a better signal detection of antidepressive efficacy.