The impact of continuity correction methods in Cochrane reviews with single-zero trials with rare events: A meta-epidemiological study.

Meta-analyses examining dichotomous outcomes often include single-zero studies, where no events occur in intervention or control groups. These pose challenges, and several methods have been proposed to address them. A fixed continuity correction method has been shown to bias estimates, but it is frequently used because sometimes software (e.g., RevMan software in Cochrane reviews) uses it as a default. We aimed to empirically compare results using the continuity correction with those using alternative models that do not require correction. To this aim, we reanalyzed the original data from 885 meta-analyses in Cochrane reviews using the following methods: (i) Mantel-Haenszel model with a fixed continuity correction, (ii) random effects inverse variance model with a fixed continuity correction, (iii) Peto method (the three models available in RevMan), (iv) random effects inverse variance model with the treatment arm continuity correction, (v) Mantel-Haenszel model without correction, (vi) logistic regression, and (vii) a Bayesian random effects model with binominal likelihood. For each meta-analysis we calculated ratios of odds ratios between all methods, to assess how the choice of method may impact results. Ratios of odds ratios <0.8 or <1.25 were seen in ~30% of the existing meta-analyses when comparing results between Mantel-Haenszel model with a fixed continuity correction and either Mantel-Haenszel model without correction or logistic regression. We concluded that injudicious use of the fixed continuity correction in existing Cochrane reviews may have substantially influenced effect estimates in some cases. Future updates of RevMan should incorporate less biased statistical methods.

Targeted next-generation sequencing to diagnose drug-resistant tuberculosis: a systematic review and meta-analysis.

BACKGROUND: Targeted next-generation sequencing (NGS) can rapidly and simultaneously detect mutations associated with resistance to tuberculosis drugs across multiple gene targets. The use of targeted NGS to diagnose drug-resistant tuberculosis, as described in publicly available data, has not been comprehensively reviewed. We aimed to identify targeted NGS assays that diagnose drug-resistant tuberculosis, determine how widely this technology has been used, and assess the diagnostic accuracy of these assays. METHODS: In this systematic review and meta-analysis, we searched MEDLINE, Embase, Cochrane Library, Web of Science Core Collection, Global Index Medicus, Google Scholar, ClinicalTrials.gov, and the WHO International Clinical Trials Registry Platform for published and unpublished reports on targeted NGS for drug-resistant tuberculosis from Jan 1, 2005, to Oct 14, 2022, with updates to our search in Embase and Google Scholar until Feb 13, 2024. Studies eligible for the systematic review described targeted NGS approaches to predict drug resistance in Mycobacterium tuberculosis infections using primary samples, reference strain collections, or cultured isolates from individuals with presumed or confirmed tuberculosis. Our search had no limitations on study type or language, although only reports in English, German, and French were screened for eligibility. For the meta-analysis, we included test accuracy studies that used any reference standard, and we assessed risk of bias using the Quality Assessment of Diagnostic Accuracy Studies-2 tool. The primary outcomes for the meta-analysis were sensitivity and specificity of targeted NGS to diagnose drug-resistant tuberculosis compared to phenotypic and genotypic drug susceptibility testing. We used a Bayesian bivariate model to generate summary receiver operating characteristic plots and diagnostic accuracy measures, overall and stratified by drug and sample type. This study is registered with PROSPERO, CRD42022368707. FINDINGS: We identified and screened 2920 reports, of which 124 were eligible for our systematic review, including 37 review articles and 87 reports of studies collecting samples for targeted NGS. Sequencing was mainly done in the USA (14 [16%] of 87), western Europe (ten [11%]), India (ten [11%]), and China (nine [10%]). We included 24 test accuracy studies in the meta-analysis, in which 23 different tuberculosis drugs or drug groups were assessed, covering first-line drugs, injectable drugs, and fluoroquinolones and predominantly comparing targeted NGS with phenotypic drug susceptibility testing. The combined sensitivity of targeted NGS across all drugs was 94·1% (95% credible interval [CrI] 90·9-96·3) and specificity was 98·1% (97·0-98·9). Sensitivity for individual drugs ranged from 76·5% (52·5-92·3) for capreomycin to 99·1% (98·3-99·7) for rifampicin; specificity ranged from 93·1% (88·0-96·3) for ethambutol to 99·4% (98·3-99·8) for amikacin. Diagnostic accuracy was similar for primary clinical samples and culture isolates overall and for rifampicin, isoniazid, ethambutol, streptomycin, and fluoroquinolones, and similar after excluding studies at high risk of bias (overall sensitivity 95·2% [95% CrI 91·7-97·1] and specificity 98·6% [97·4-99·3]). INTERPRETATION: Targeted NGS is highly sensitive and specific for detecting drug resistance across panels of tuberculosis drugs and can be performed directly on clinical samples. There is a paucity of data on performance for some currently recommended drugs. The barriers preventing the use of targeted NGS to diagnose drug-resistant tuberculosis in high-burden countries need to be addressed. FUNDING: National Institutes of Allergy and Infectious Diseases and Swiss National Science Foundation.

Combining endpoint and change data did not affect the summary standardised mean difference in pairwise and network meta-analyses: An empirical study in depression.

When studies use different scales to measure continuous outcomes, standardised mean differences (SMD) are required to meta-analyse the data. However, outcomes are often reported as endpoint or change from baseline scores. Combining corresponding SMDs can be problematic and available guidance advises against this practice. We aimed to examine the impact of combining the two types of SMD in meta-analyses of depression severity. We used individual participant data on pharmacological interventions (89 studies, 27,409 participants) and internet-delivered cognitive behavioural therapy (iCBT; 61 studies, 13,687 participants) for depression to compare endpoint and change from baseline SMDs at the study level. Next, we performed pairwise (PWMA) and network meta-analyses (NMA) using endpoint SMDs, change from baseline SMDs, or a mixture of the two. Study-specific SMDs calculated from endpoint and change from baseline data were largely similar, although for iCBT interventions 25% of the studies at 3 months were associated with important differences between study-specific SMDs (median 0.01, IQR -0.10, 0.13) especially in smaller trials with baseline imbalances. However, when pooled, the differences between endpoint and change SMDs were negligible. Pooling only the more favourable of the two SMDs did not materially affect meta-analyses, resulting in differences of pooled SMDs up to 0.05 and 0.13 in the pharmacological and iCBT datasets, respectively. Our findings have implications for meta-analyses in depression, where we showed that the choice between endpoint and change scores for estimating SMDs had immaterial impact on summary meta-analytic estimates. Future studies should replicate and extend our analyses to fields other than depression.

Treatment methods for cervical intraepithelial neoplasia in England: A cost-effectiveness analysis.

OBJECTIVE: To compare the cost-effectiveness of different treatments for cervical intraepithelial neoplasia (CIN). DESIGN: A cost-effectiveness analysis based on data available in the literature and expert opinion. SETTING: England. POPULATION: Women treated for CIN. METHODS: We developed a decision-analytic model to simulate the clinical course of 1000 women who received local treatment for CIN and were followed up for 10 years after treatment. In the model we considered surgical complications as well as oncological and reproductive outcomes over the 10-year period. The costs calculated were those incurred by the National Health Service (NHS) of England. MAIN OUTCOME MEASURES: Cost per one CIN2+ recurrence averted (oncological outcome); cost per one preterm birth averted (reproductive outcome); overall cost per one adverse oncological or reproductive outcome averted. RESULTS: For young women of reproductive age, large loop excision of the transformation zone (LLETZ) was the most cost-effective treatment overall at all willingness-to-pay thresholds. For postmenopausal women, LLETZ remained the most cost-effective treatment up to a threshold of £31,500, but laser conisation became the most cost-effective treatment above that threshold. CONCLUSIONS: LLETZ is the most cost-effective treatment for both younger and older women. However, for older women, more radical excision with laser conisation could also be considered if the NHS is willing to spend more than £31,500 to avert one CIN2+ recurrence.

Effect of dairy consumption on cognition in older adults: A population-based cohort study.

OBJECTIVE: We aimed to assess the effect on cognitive function of adding dairy (total, fermented, non-fermented, full fat, low fat, and sugary) to the diet and of substituting some food groups for dairy. DESIGN: Secondary analysis of a prospective population-based cohort study. PARTICIPANTS: We analyzed data from 1334 cognitively healthy participants (median age 67 years at baseline) with a mean follow-up of 5.6 years from the CoLaus|PsyColaus cohort in Lausanne, Switzerland. MEASUREMENTS: The participants completed a food frequency questionnaire at baseline and cognitive tests at baseline and at follow-up. Clinical dementia rating was the primary outcome. Subjective cognitive decline, memory, verbal fluency, executive and motor functions were secondary outcomes. METHODS: Our exposure was the consumption of total and 5 sub-types of dairy products (g/d). We used marginal structural models to compute average causal effects of 1) increasing dairy consumption by 100 g/d and 2) substituting 100 g/d of meat, fish, eggs, fruits and vegetables with dairy on the outcomes. We used inverse probability of the treatment and lost to follow-up weighting to account for measured confounding and non-random loss to follow-up. RESULTS: Overall, the effects of adding dairy products to the diet on cognition were negligible and imprecise. No substitution had a substantial and consistent effect on clinical dementia rating. The substitution of fish [11.7% (-3% to 26.5%)] and eggs [18% (2.3%-33.7%)] for dairy products could negatively impact verbal memory and neurolinguistic processes. CONCLUSION: We found no effect of adding dairy to the diet or substituting meat, vegetables or fruit for dairy on cognitive function in this cohort of older adults. The substitution of fish and eggs for dairy could have a negative effect on some secondary outcomes, but more studies modeling food substitutions are needed to confirm these results.

Lipid-Lowering Therapy and Risk of Hemorrhagic Stroke: A Systematic Review and Meta-Analysis of Randomized Controlled Trials.

BACKGROUND: There is debate over whether statins increase risk of hemorrhagic stroke, so we assessed current evidence, including data from new statin trials and trials of nonstatin low-density lipoprotein-cholesterol (LDL-C)- and triglyceride-lowering therapies. METHODS AND RESULTS: We performed a systematic review of large randomized clinical trials (≥1000 patients with ≥2 years follow-up) of LDL-C-lowering therapy (statin, ezetimibe, and PCSK-9 [proprotein convertase subtilisin/kexin type 9] inhibitor) and triglyceride-lowering therapy (omega-3 supplements and fibrate) that reported hemorrhagic stroke as an outcome. We searched MEDLINE, Embase, and Cochrane Library up to July 2, 2021 and updated a meta-analysis of cardiovascular statin trials published in 2012. Among our several subgroup analyses, we looked at difference depending on stroke status and also depending on age. We identified 37 trials for LDL-C lowering (284 301 participants) and 11 for triglyceride lowering (120 984 participants). Overall, we found a higher risk of hemorrhagic stroke for LDL-C lowering, risk ratio (RR) 1.16 (95% CI, 1.01-1.32, P=0.03). For statins (33 trials, 216 258 participants), RR=1.17 (95% CI, 1.01-1.36); for PCSK-9 inhibitors (2 trials, 46 488 participants), RR=0.86 (95% CI, 0.43-1.74); and for ezetimibe (2 trials, 21 555 participants), RR=1.14 (95% CI, 0.64-2.03). In statin trials of patients with previous stroke/transient ischemic attack, RR was 1.46 (95% CI, 1.05-2.04), and in trials with mean age ≥65 years old, RR=1.34 (95% CI, 1.04-1.73) (Pint=0.14 and Pint=0.23 respectively); for triglyceride lowering (11 trials, 120 984 participants), RR=1.05 (95% CI, 0.86-1.30). CONCLUSIONS: We found evidence for a small increased risk of hemorrhagic stroke events with LDL-C-lowering therapies but no clear evidence for triglyceride-lowering therapies. REGISTRATION: URL: https://www.crd.york.ac.uk/prospero; Unique identifier: CRD42021275363.

Efficacy and effectiveness of antipsychotics in schizophrenia: network meta-analyses combining evidence from randomised controlled trials and real-world data.

BACKGROUND: There is debate about the generalisability of results from randomised clinical trials (RCTs) to real-world settings. Studying outcomes of treatments for schizophrenia can shed light on this issue and inform treatment guidelines. We therefore compared the efficacy and effectiveness of antipsychotics for relapse prevention in schizophrenia and estimated overall treatment effects using all available RCT and real-world evidence. METHODS: We conducted network meta-analyses using individual participant data from Swedish and Finnish national registries and aggregate data from RCTs. The target population was adults (age >18 and <65 years) with schizophrenia and schizoaffective disorder with stabilised symptoms. We analysed each registry separately to obtain hazard ratios (HRs) and 95% CIs for relapse within 6 months post-antipsychotic initiation as our main outcome. Interventions studied were antipsychotics, no antipsychotic use, and placebo. We compared HRs versus a reference drug (oral haloperidol) between registries, and between registry individuals who would be eligible and ineligible for RCTs, using the ratio of HRs. We synthesised evidence using network meta-analysis and compared results from our network meta-analysis of real-world data with our network meta-analysis of RCT data, including oral versus long-acting injectable (LAI) formulations. Finally, we conducted a joint real-world and RCT network meta-analysis. FINDINGS: We included 90 469 individuals from the Swedish and Finnish registries (mean age 45·9 [SD 14·6] years; 43 025 [47·5%] women and 47 467 [52·5%] men, ethnicity data unavailable) and 10 091 individuals from 30 RCTs (mean age 39·6 years [SD 11·7]; 3724 [36·9%] women and 6367 [63·1%] men, 6022 White [59·7%]). We found good agreement in effectiveness of antipsychotics between Swedish and Finnish registries (HR ratio 0·97, 95% CI 0·88-1·08). Drug effectiveness versus no antipsychotic was larger in RCT-eligible than RCT-ineligible individuals (HR ratio 1·40 [1·24-1·59]). Efficacy versus placebo in RCTs was larger than effectiveness versus no antipsychotic in real-world (HR ratio 2·58 [2·02-3·30]). We found no evidence of differences between effectiveness and efficacy for between-drug comparisons (HR ratio vs oral haloperidol 1·17 [0·83-1·65], where HR ratio >1 means superior effectiveness in real-world to RCTs), except for LAI versus oral comparisons (HR ratio 0·73 [0·53-0·99], indicating superior effectiveness in real-world data relative to RCTs). The real-world network meta-analysis showed clozapine was most effective, followed by olanzapine LAI. The RCT network meta-analysis exhibited heterogeneity and inconsistency. The joint real-world and RCT network meta-analysis identified olanzapine as the most efficacious antipsychotic amongst those present in both RCTs and the real world registries. INTERPRETATION: LAI antipsychotics perform slightly better in the real world than according to RCTs. Otherwise, RCT evidence was in line with real-world evidence for most between-drug comparisons, but RCTs might overestimate effectiveness of antipsychotics observed in routine care settings. Our results further the understanding of the generalisability of RCT findings to clinical practice and can inform preferential prescribing guidelines. FUNDING: None.

Components and Delivery Formats of Cognitive Behavioral Therapy for Chronic Insomnia in Adults: A Systematic Review and Component Network Meta-Analysis.

Importance: Chronic insomnia disorder is highly prevalent, disabling, and costly. Cognitive behavioral therapy for insomnia (CBT-I), comprising various educational, cognitive, and behavioral strategies delivered in various formats, is the recommended first-line treatment, but the effect of each component and delivery method remains unclear. Objective: To examine the association of each component and delivery format of CBT-I with outcomes. Data Sources: PubMed, Cochrane Central Register of Controlled Trials, PsycInfo, and International Clinical Trials Registry Platform from database inception to July 21, 2023. Study Selection: Published randomized clinical trials comparing any form of CBT-I against another or a control condition for chronic insomnia disorder in adults aged 18 years and older. Insomnia both with and without comorbidities was included. Concomitant treatments were allowed if equally distributed among arms. Data Extraction and Synthesis: Two independent reviewers identified components, extracted data, and assessed trial quality. Random-effects component network meta-analyses were performed. Main Outcomes and Measures: The primary outcome was treatment efficacy (remission defined as reaching a satisfactory state) posttreatment. Secondary outcomes included all-cause dropout, self-reported sleep continuity, and long-term remission. Results: A total of 241 trials were identified including 31 452 participants (mean [SD] age, 45.4 [16.6] years; 21 048 of 31 452 [67%] women). Results suggested that critical components of CBT-I are cognitive restructuring (remission incremental odds ratio [iOR], 1.68; 95% CI, 1.28-2.20) third-wave components (iOR, 1.49; 95% CI, 1.10-2.03), sleep restriction (iOR, 1.49; 95% CI, 1.04-2.13), and stimulus control (iOR, 1.43; 95% CI, 1.00-2.05). Sleep hygiene education was not essential (iOR, 1.01; 95% CI, 0.77-1.32), and relaxation procedures were found to be potentially counterproductive(iOR, 0.81; 95% CI, 0.64-1.02). In-person therapist-led programs were most beneficial (iOR, 1.83; 95% CI, 1.19-2.81). Cognitive restructuring, third-wave components, and in-person delivery were mainly associated with improved subjective sleep quality. Sleep restriction was associated with improved subjective sleep quality, sleep efficiency, and wake after sleep onset, and stimulus control with improved subjective sleep quality, sleep efficiency, and sleep latency. The most efficacious combination-consisting of cognitive restructuring, third wave, sleep restriction, and stimulus control in the in-person format-compared with in-person psychoeducation, was associated with an increase in the remission rate by a risk difference of 0.33 (95% CI, 0.23-0.43) and a number needed to treat of 3.0 (95% CI, 2.3-4.3), given the median observed control event rate of 0.14. Conclusions and Relevance: The findings suggest that beneficial CBT-I packages may include cognitive restructuring, third-wave components, sleep restriction, stimulus control, and in-person delivery but not relaxation. However, potential undetected interactions could undermine the conclusions. Further large-scale, well-designed trials are warranted to confirm the contribution of different treatment components in CBT-I.

Multiple imputation of incomplete multilevel data using Heckman selection models.

Missing data is a common problem in medical research, and is commonly addressed using multiple imputation. Although traditional imputation methods allow for valid statistical inference when data are missing at random (MAR), their implementation is problematic when the presence of missingness depends on unobserved variables, that is, the data are missing not at random (MNAR). Unfortunately, this MNAR situation is rather common, in observational studies, registries and other sources of real-world data. While several imputation methods have been proposed for addressing individual studies when data are MNAR, their application and validity in large datasets with multilevel structure remains unclear. We therefore explored the consequence of MNAR data in hierarchical data in-depth, and proposed a novel multilevel imputation method for common missing patterns in clustered datasets. This method is based on the principles of Heckman selection models and adopts a two-stage meta-analysis approach to impute binary and continuous variables that may be outcomes or predictors and that are systematically or sporadically missing. After evaluating the proposed imputation model in simulated scenarios, we illustrate it use in a cross-sectional community survey to estimate the prevalence of malaria parasitemia in children aged 2-10 years in five regions in Uganda.

External validation of the Dat'AIDS score: A risk score for predicting 5-year overall mortality in people living with HIV aged 60 years or older.

OBJECTIVE: To perform an external validation of the Dat'AIDS score for predicting 5-year overall mortality among people with HIV (PWH) aged 60 years or older. METHODS: This was a multi-centre prospective cohort study at all sites participating in the Swiss HIV Cohort Study (SHCS). We calculated the Dat'AIDS score in PWH aged 60 years or older at their first visit between 1 January 2015 and 1 January 2020. People living with HIV-2 and those whose Dat'AIDS score could not be calculated were excluded. Patients were followed until 1 January 2020. The primary endpoint was all-cause mortality. Vital status was collected throughout the study period. We obtained population and score descriptive statistics and assessed the score's discrimination and calibration. RESULTS: We included 2205 participants (82% male) of median [interquartile range (IQR)] age 62.0 (60.3-67.0) years, mostly with viraemia <50 copies/mL (92.7%). Median follow-up time was 15.9 years and median (IQR) CD4 cell count at enrolment was 586 (420-782) cells/µL. In all, 152 deaths were recorded during a total follow-up period of 7147 patient-years. The median (IQR) observed Dat'AIDS score was 3 (0-8). Discriminative capacities were good as the C-statistic was 0.73 (95% CI: 0.69-0.77) and consistent across all subgroups. Comparison of observed and expected survival probabilities showed good calibration. CONCLUSIONS: External validation of the Dat'AIDS score in patients aged 60 years or older showed that it could be a useful tool not only for research purposes, but also to identify older patients at a higher mortality risk and to tailor the most appropriate interventions.

A systematic review and meta-analysis investigating the relationship between metabolic syndrome and the incidence of thyroid diseases.

PURPOSE: To assess the prospective association between metabolic syndrome (MetS), its components, and incidence of thyroid disorders by conducting a systematic review and meta-analysis. METHODS: A systematic search was performed in Ovid Medline, Embase.com, and Cochrane CENTRAL from inception to February 22, 2023. Publications from prospective studies were included if they provided data on baseline MetS status or one of its components and assessed the incidence of thyroid disorders over time. A random effects meta-analysis was conducted to calculate the odds ratio (OR) for developing thyroid disorders. RESULTS: After full-text screening of 2927 articles, seven studies met our inclusion criteria. Two of these studies assessed MetS as an exposure (N = 71,727) and were included in our meta-analysis. The association between MetS at baseline and incidence of overt hypothyroidism at follow-up yielded an OR of 0.78 (95% confidence interval [CI]: 0.52-1.16 for two studies, I2 = 0%). Pooled analysis was not possible for subclinical hypothyroidism, due to large heterogeneity (I2 = 92.3%), nor for hyperthyroidism, as only one study assessed this association. We found evidence of an increased risk of overt (RR: 3.10 (1.56-4.64, I2 = 0%) and subclinical hypothyroidism (RR 1.50 (1.05-1.94), I2 = 0%) in individuals with obesity at baseline. There was a lower odds of developing overt hyperthyroidism in individuals with prediabetes at baseline (OR: 0.68 (0.47-0.98), I2 = 0%). CONCLUSIONS: We were unable to draw firm conclusions regarding the association between MetS and the incidence of thyroid disorders due to the limited number of available studies and the presence of important heterogeneity in reporting results. However, we did find an association between obesity at baseline and incidence of overt and subclinical hypothyroidism.

Imiquimod for Cervical and Vaginal Intraepithelial Neoplasia: A Systematic Review and Meta-analysis.

OBJECTIVE: To evaluate the treatment efficacy and the risk of adverse events of imiquimod for cervical intraepithelial neoplasia (CIN) and vaginal intraepithelial neoplasia (VAIN), compared with placebo or no intervention. DATA SOURCES: We searched Cochrane, PubMed, ISRCTN registry, ClinicalTrials.gov , and the World Health Organization International Clinical Trials Registry Platform up to November 23, 2022. METHODS OF STUDY SELECTION: We included randomized controlled trials and prospective nonrandomized studies with control arms that investigated the efficacy of imiquimod for histologically confirmed CIN or VAIN. The primary outcomes were histologic regression of the disease (primary efficacy outcome) and treatment discontinuation due to side effects (primary safety outcome). We estimated pooled odds ratios (ORs) of imiquimod, compared with placebo or no intervention. We also conducted a meta-analysis of the proportions of patients with adverse events in the imiquimod arms. TABULATION, INTEGRATION, AND RESULTS: Four studies contributed to the pooled OR for the primary efficacy outcome. An additional four studies were available for meta-analyses of proportions in the imiquimod arm. Imiquimod was associated with increased probability of regression (pooled OR 4.05, 95% CI 2.08-7.89). Pooled OR for CIN in the three studies was 4.27 (95% CI 2.11-8.66); results of one study were available for VAIN (OR, 2.67, 95% CI 0.36-19.71). Pooled probability for primary safety outcome in the imiquimod arm was 0.07 (95% CI 0.03-0.14). The pooled probabilities (95% CI) of secondary outcomes were 0.51 (0.20-0.81) for fever, 0.53 (0.31-0.73) for arthralgia or myalgia, 0.31 (0.18-0.47) for abdominal pain, 0.28 (0.09-0.61) for abnormal vaginal discharge or genital bleeding, 0.48 (0.16-0.82) for vulvovaginal pain, and 0.02 (0.01-0.06) for vaginal ulceration. CONCLUSION: Imiquimod was found to be effective for CIN, whereas data on VAIN were limited. Although local and systemic complications are common, treatment discontinuation is infrequent. Thus, imiquimod is potentially an alternative therapy to surgery for CIN. SYSTEMATIC REVIEW REGISTRATION: PROSPERO, CRD42022377982.

Predicting outcomes at the individual patient level: what is the best method?

OBJECTIVE: When developing prediction models, researchers commonly employ a single model which uses all the available data (end-to-end approach). Alternatively, a similarity-based approach has been previously proposed, in which patients with similar clinical characteristics are first grouped into clusters, then prediction models are developed within each cluster. The potential advantage of the similarity-based approach is that it may better address heterogeneity in patient characteristics. However, it remains unclear whether it improves the overall predictive performance. We illustrate the similarity-based approach using data from people with depression and empirically compare its performance with the end-to-end approach. METHODS: We used primary care data collected in general practices in the UK. Using 31 predefined baseline variables, we aimed to predict the severity of depressive symptoms, measured by Patient Health Questionnaire-9, 60 days after initiation of antidepressant treatment. Following the similarity-based approach, we used k-means to cluster patients based on their baseline characteristics. We derived the optimal number of clusters using the Silhouette coefficient. We used ridge regression to build prediction models in both approaches. To compare the models' performance, we calculated the mean absolute error (MAE) and the coefficient of determination (R2) using bootstrapping. RESULTS: We analysed data from 16 384 patients. The end-to-end approach resulted in an MAE of 4.64 and R2 of 0.20. The best-performing similarity-based model was for four clusters, with MAE of 4.65 and R2 of 0.19. CONCLUSIONS: The end-to-end and the similarity-based model yielded comparable performance. Due to its simplicity, the end-to-end approach can be favoured when using demographic and clinical data to build prediction models on pharmacological treatments for depression.

Cardiovascular disease and type 2 diabetes in older adults: a combined protocol for an individual participant data analysis for risk prediction and a network meta-analysis of novel anti-diabetic drugs.

Introduction: Older and multimorbid adults with type 2 diabetes (T2D) are at high risk of cardiovascular disease (CVD) and chronic kidney disease (CKD). Estimating risk and preventing CVD is a challenge in this population notably because it is underrepresented in clinical trials. Our study aims to (1) assess if T2D and haemoglobin A1c (HbA1c) are associated with the risk of CVD events and mortality in older adults, (2) develop a risk score for CVD events and mortality for older adults with T2D, (3) evaluate the comparative efficacy and safety of novel antidiabetics. Methods and analysis: For Aim 1, we will analyse individual participant data on individuals aged ≥65 years from five cohort studies: the Optimising Therapy to Prevent Avoidable Hospital Admissions in Multimorbid Older People study; the Cohorte Lausannoise study; the Health, Aging and Body Composition study; the Health and Retirement Study; and the Survey of Health, Ageing and Retirement in Europe. We will fit flexible parametric survival models (FPSM) to assess the association of T2D and HbA1c with CVD events and mortality. For Aim 2, we will use data on individuals aged ≥65 years with T2D from the same cohorts to develop risk prediction models for CVD events and mortality using FPSM. We will assess model performance, perform internal-external cross validation, and derive a point-based risk score. For Aim 3, we will systematically search randomized controlled trials of novel antidiabetics. Network meta-analysis will be used to determine comparative efficacy in terms of CVD, CKD, and retinopathy outcomes, and safety of these drugs. Confidence in results will be judged using the CINeMA tool. Ethics and dissemination: Aims 1 and 2 were approved by the local ethics committee (Kantonale Ethikkommission Bern); no approval is required for Aim 3. Results will be published in peer-reviewed journals and presented in scientific conferences.

Developing prediction models when there are systematically missing predictors in individual patient data meta-analysis.

Clinical prediction models are widely used in modern clinical practice. Such models are often developed using individual patient data (IPD) from a single study, but often there are IPD available from multiple studies. This allows using meta-analytical methods for developing prediction models, increasing power and precision. Different studies, however, often measure different sets of predictors, which may result to systematically missing predictors, that is, when not all studies collect all predictors of interest. This situation poses challenges in model development. We hereby describe various approaches that can be used to develop prediction models for continuous outcomes in such situations. We compare four approaches: a "restrict predictors" approach, where the model is developed using only predictors measured in all studies; a multiple imputation approach that ignores study-level clustering; a multiple imputation approach that accounts for study-level clustering; and a new approach that develops a prediction model in each study separately using all predictors reported, and then synthesizes all predictions in a multi-study ensemble. We explore in simulations the performance of all approaches under various scenarios. We find that imputation methods and our new method outperform the restrict predictors approach. In several scenarios, our method outperformed imputation methods, especially for few studies, when predictor effects were small, and in case of large heterogeneity. We use a real dataset of 12 trials in psychotherapies for depression to illustrate all methods in practice, and we provide code in R.

Second-generation antipsychotics and seizures - a systematic review and meta-analysis of serious adverse events in randomized controlled trials.

Seizures are suspected to be side effects of antipsychotics. To examine a possible causal relationship, we compared the risk of seizures on second-generation antipsychotics to the risk on placebo in randomized controlled clinical trials (RCTs) across diagnostic groups. The primary outcome was any seizure reported as International Conference on Harmonisation-Good Clinical Practice (ICH-GCP)-defined serious adverse event (SAEs). The risk ratio (RR) with antipsychotics versus placebo was synthesized in a pairwise common effects Mantel-Haenszel meta-analysis. For 314 of 597 idenitified placebo-controlled RCTs information about all SAEs could be retrieved from publications, original investigators, pharmaceutical companies and the European Medical Agency. In those, 37 seizures occurred in 42,600 participants on antipsychotics (0.09%) and 28 in 25,042 participants on placebo (0.11%). The meta-analytic results (RR 0,68; 95% Confidence Interval 0.41-1.12) indicated a reduced risk on antipsychotics with a confidence interval including no difference (i.e. RR=1). Neither in sensitivity analyses (excluding events in the safety-follow-up of trials or first-generation antipsychotics; using odds ratios) nor in subgroup analyses (on specific antipsychotics, drug combinations, diagnostic categories, age groups, and study duration) there was evidence for an increased risk on antipsychotics, except for some weak indications of an increased risk on antipsychotics in older and/or demented participants (RRs 1.11 and 1.48, respectively, but with 95% CIs of 0.35-3.49 and 0.41-5.26 including no difference and subgroup tests with p=0.54 and p=0.66 not indicating differences between age groups or diagnostic categories). Consequently, there are no indications that second-generation antipsychotics cause seizures in middle-aged adults and children in most diagnostic groups; rather our results provide some weak evidence for a protective effect. However, there was no data on SAEs available for clozapine, for which observational studies provide the strongest associations with increased seizure rates, and for older and/or demented patients a small additional risk on antipsychotics cannot be excluded.

Measuring the performance of prediction models to personalize treatment choice.

When data are available from individual patients receiving either a treatment or a control intervention in a randomized trial, various statistical and machine learning methods can be used to develop models for predicting future outcomes under the two conditions, and thus to predict treatment effect at the patient level. These predictions can subsequently guide personalized treatment choices. Although several methods for validating prediction models are available, little attention has been given to measuring the performance of predictions of personalized treatment effect. In this article, we propose a range of measures that can be used to this end. We start by defining two dimensions of model accuracy for treatment effects, for a single outcome: discrimination for benefit and calibration for benefit. We then amalgamate these two dimensions into an additional concept, decision accuracy, which quantifies the model's ability to identify patients for whom the benefit from treatment exceeds a given threshold. Subsequently, we propose a series of performance measures related to these dimensions and discuss estimating procedures, focusing on randomized data. Our methods are applicable for continuous or binary outcomes, for any type of prediction model, as long as it uses baseline covariates to predict outcomes under treatment and control. We illustrate all methods using two simulated datasets and a real dataset from a trial in depression. We implement all methods in the R package predieval. Results suggest that the proposed measures can be useful in evaluating and comparing the performance of competing models in predicting individualized treatment effect.

Using individual participant data to improve network meta-analysis projects.

A network meta-analysis combines the evidence from existing randomised trials about the comparative efficacy of multiple treatments. It allows direct and indirect evidence about each comparison to be included in the same analysis, and provides a coherent framework to compare and rank treatments. A traditional network meta-analysis uses aggregate data (eg, treatment effect estimates and standard errors) obtained from publications or trial investigators. An alternative approach is to obtain, check, harmonise and meta-analyse the individual participant data (IPD) from each trial. In this article, we describe potential advantages of IPD for network meta-analysis projects, emphasising five key benefits: (1) improving the quality and scope of information available for inclusion in the meta-analysis, (2) examining and plotting distributions of covariates across trials (eg, for potential effect modifiers), (3) standardising and improving the analysis of each trial, (4) adjusting for prognostic factors to allow a network meta-analysis of conditional treatment effects and (5) including treatment-covariate interactions (effect modifiers) to allow relative treatment effects to vary by participant-level covariate values (eg, age, baseline depression score). A running theme of all these benefits is that they help examine and reduce heterogeneity (differences in the true treatment effect between trials) and inconsistency (differences in the true treatment effect between direct and indirect evidence) in the network. As a consequence, an IPD network meta-analysis has the potential for more precise, reliable and informative results for clinical practice and even allows treatment comparisons to be made for individual patients and targeted populations conditional on their particular characteristics.

Individual Patient Data Meta-analysis of Drug-eluting Versus Bare-metal Stents for Percutaneous Coronary Intervention in Chronic Versus Acute Coronary Syndromes.

New-generation drug-eluting stents (DES) strongly reduce restenosis and repeat revascularization compared with bare-metal stents (BMS) for percutaneous coronary intervention. There is residual uncertainty as to whether other prognostically relevant outcomes are affected by DES versus BMS concerning initial presentation (chronic coronary syndrome [CCS] vs acute coronary syndrome [ACS]). We performed an individual patient data meta-analysis of randomized trials comparing new-generation DES versus BMS (CRD42017060520). The primary outcome was the composite of cardiac death or myocardial infarction (MI). Outcomes were examined at maximum follow-up and with a 1-year landmark. Risk estimates are expressed as hazard ratio (HR) with 95% confidence interval (CI). A total of 22,319 patients were included across 14 trials; 7,691 patients (34.5%) with CCS and 14,628 patients (65.5%) with ACS. We found evidence that new-generation DES versus BMS consistently reduced the risk of cardiac death or MI in both patients with CCS (HR 0.83, 95% CI 0.70 to 0.98, p <0.001) and ACS (HR 0.83, 95% CI 0.75 to 0.92, p <0.001) (p-interaction = 0.931). This benefit was mainly driven by a similar reduction in the risk of MI (p-interaction = 0.898) for both subsets (HRCCS 0.80, 95% CI 0.65 to 0.97; HRACS 0.79, 95% CI 0.70 to 0.89). In CCS and ACS, we found a time-dependent treatment effect, with the benefit from DES accumulating during 1-year follow-up, without offsetting effects after that. In conclusion, patients with CCS were slightly underrepresented in comparative clinical trials. Still, they benefited similarly to patients with ACS from new-generation DES instead of BMS with a sustained reduction of cardiac death or MI because of lower event rates within 1 year.

Personalized Prediction of Alzheimer's Disease and Its Treatment Effects by Donepezil: An Individual Participant Data Meta-Analysis of Eight Randomized Controlled Trials.

BACKGROUND: Patient characteristics may predict the progression of Alzheimer's disease (AD) and may moderate the effects of donepezil. OBJECTIVE: To build a personalized prediction model for patients with AD and to estimate patient-specific treatment effects of donepezil, using individual patient characteristics. METHODS: We systematically searched for all double-masked randomized controlled trials comparing oral donepezil and pill placebo in the treatment of AD and requested individual participant data through its developer, Eisai. The primary outcome was cognitive function at 24 weeks, measured with the Alzheimer's Disease Assessment Scale-cognitive component (ADAS-cog). We built a Bayesian meta-analytical prediction model for patients receiving placebo and we performed an individual patient data meta-analysis to estimate patient-level treatment effects. RESULTS: Eight studies with 3,156 participants were included. The Bayesian prediction model suggested that more severe cognitive and global function at baseline and younger age were associated with worse cognitive function at 24 weeks. The individual participant data meta-analysis showed that, on average, donepezil was superior to placebo in cognitive function (ADAS-cog scores, -3.2; 95% Credible Interval (CrI) -4.2 to -2.1). In addition, our results suggested that antipsychotic drug use at baseline might be associated with a lower effect of donepezil in ADAS-cog (2.0; 95% CrI, -0.02 to 4.3). CONCLUSION: Although our results suggested that donepezil is somewhat efficacious for cognitive function for most patients with AD, use of antipsychotic drugs may be associated with lower efficacy of the drug. Future research with larger sample sizes, more patient covariates, and longer treatment duration is needed.