The treatment recommendation based on a network meta-analysis (NMA) is usually the single treatment with the highest expected value (EV) on an evaluative function. We explore approaches that recommend multiple treatments and that penalise uncertainty, making them suitable for risk-averse decision-makers. We introduce loss-adjusted EV (LaEV) and compare it to GRADE and three probability-based rankings. We define properties of a valid ranking under uncertainty and other desirable properties of ranking systems. A two-stage process is proposed: the first identifies treatments superior to the reference treatment; the second identifies those that are also within a minimal clinically important difference (MCID) of the best treatment. Decision rules and ranking systems are compared on stylised examples and 10 NMAs used in NICE (National Institute of Health and Care Excellence) guidelines. Only LaEV reliably delivers valid rankings under uncertainty and has all the desirable properties. In 10 NMAs comparing between 5 and 41 treatments, an EV decision maker would recommend 4–14 treatments, and LaEV 0–3 (median 2) fewer. GRADE rules give rise to anomalies, and, like the probability-based rankings, the number of treatments recommended depends on arbitrary probability cutoffs. Among treatments that are superior to the reference, GRADE privileges the more uncertain ones, and in 3/10 cases, GRADE failed to recommend the treatment with the highest EV and LaEV. A two-stage approach based on MCID ensures that EV- and LaEV-based rules recommend a clinically appropriate number of treatments. For a risk-averse decision maker, LaEV is conservative, simple to implement, and has an independent theoretical foundation.

A fundamental pillar of science is the estimation of the effect size of associations. However, this task is sometimes difficult and error-prone. To facilitate this process, the R package metaConvert automatically calculates and flexibly converts multiple effect size measures. It applies more than 120 formulas to convert any relevant input data into Cohen’s d, Hedges’ g, mean difference, odds ratio, risk ratio, incidence rate ratio, correlation coefficient, Fisher’s r-to-z transformed correlation coefficient, variability ratio, coefficient of variation ratio, or number needed to treat. Researchers unfamiliar with R can use this software through a browser-based graphical interface (https://metaconvert.org/). We hope this suite will help researchers in the life sciences and other disciplines estimate and convert effect sizes more easily and accurately.

With the increasing volume of scientific literature, there is a need to streamline the screening process for titles and abstracts in systematic reviews, reduce the workload for reviewers, and minimize errors. This study validated artificial intelligence (AI) tools, specifically Llama 3 70B via Groq’s application programming interface (API) and ChatGPT-4o mini via OpenAI’s API, for automating this process in biomedical research. It compared these AI tools with human reviewers using 1,081 articles after duplicate removal. Each AI model was tested in three configurations to assess sensitivity, specificity, predictive values, and likelihood ratios. The Llama 3 model’s LLA_2 configuration achieved 77.5% sensitivity and 91.4% specificity, with 90.2% accuracy, a positive predictive value (PPV) of 44.3%, and a negative predictive value (NPV) of 97.9%. The ChatGPT-4o mini model’s CHAT_2 configuration showed 56.2% sensitivity, 95.1% specificity, 92.0% accuracy, a PPV of 50.6%, and an NPV of 96.1%. Both models demonstrated strong specificity, with CHAT_2 having higher overall accuracy. Despite these promising results, manual validation remains necessary to address false positives and negatives, ensuring that no important studies are overlooked. This study suggests that AI can significantly enhance efficiency and accuracy in systematic reviews, potentially revolutionizing not only biomedical research but also other fields requiring extensive literature reviews.

Systematic reviews are essential for evidence-based health care, but conducting them is time- and resource-consuming. To date, efforts have been made to accelerate and (semi-)automate various steps of systematic reviews through the use of artificial intelligence (AI) and the emergence of large language models (LLMs) promises further opportunities. One crucial but complex task within systematic review conduct is assessing the risk of bias (RoB) of included studies. Therefore, the aim of this study was to test the LLM Claude 2 for RoB assessment of 100 randomized controlled trials, published in English language from 2013 onwards, using the revised Cochrane risk of bias tool (‘RoB 2’; involving judgements for five specific domains and an overall judgement). We assessed the agreement of RoB judgements by Claude with human judgements published in Cochrane reviews. The observed agreement between Claude and Cochrane authors ranged from 41% for the overall judgement to 71% for domain 4 (‘outcome measurement’). Cohen’s κ was lowest for domain 5 (‘selective reporting’; 0.10 (95% confidence interval (CI): −0.10–0.31)) and highest for domain 3 (‘missing data’; 0.31 (95% CI: 0.10–0.52)), indicating slight to fair agreement. Fair agreement was found for the overall judgement (Cohen’s κ: 0.22 (95% CI: 0.06–0.38)). Sensitivity analyses using alternative prompting techniques or the more recent version Claude 3 did not result in substantial changes. Currently, Claude’s RoB 2 judgements cannot replace human RoB assessment. However, the potential of LLMs to support RoB assessment should be further explored.

Researchers would often like to leverage data from a collection of sources (e.g., meta-analyses of randomized trials, multi-center trials, pooled analyses of observational cohorts) to estimate causal effects in a target population of interest. However, because different data sources typically represent different underlying populations, traditional meta-analytic methods may not produce causally interpretable estimates that apply to any reasonable target population. In this article, we present the CausalMetaR R package, which implements robust and efficient methods to estimate causal effects in a given internal or external target population using multi-source data. The package includes estimators of average and subgroup treatment effects for the entire target population. To produce efficient and robust estimates of causal effects, the package implements doubly robust and non-parametric efficient estimators and supports using flexible data-adaptive (e.g., machine learning techniques) methods and cross-fitting techniques to estimate the nuisance models (e.g., the treatment model, the outcome model). We briefly review the methods, describe the key features of the package, and demonstrate how to use the package through an example. The package aims to facilitate causal analyses in the context of meta-analysis.

While the Institute of Education Science’s ERIC is often recommended for comprehensive literature searching in the field of education, there are several other specialized education databases to discover education literature. This study investigates journal coverage overlaps between four specialized education databases: Education Source (EBSCO), Education Database (ProQuest), ERIC (Institute of Education Sciences), and Educator’s Reference Complete (Gale). Out of a total of 4,695 unique journals analyzed, there are 2,831 journals uniquely covered by only one database, as well as many journals covered by only two or three databases. Findings show that evidence synthesis projects and literature reviews benefit from the careful selection of multiple specialized education databases and that ERIC is insufficient as the primary education database for comprehensive searching in the field.

In health technology assessment, matching-adjusted indirect comparison (MAIC) is the most common method for pairwise comparisons that control for imbalances in baseline characteristics across trials. One of the primary challenges in MAIC is the need to properly account for the additional uncertainty introduced by the matching process. Limited evidence and guidance are available on variance estimation in MAICs. Therefore, we conducted a comprehensive Monte Carlo simulation study to evaluate the performance of different statistical methods across 108 scenarios. Four general approaches for variance estimation were compared in both anchored and unanchored MAICs of binary and time-to-event outcomes: (1) conventional estimators (CE) using raw weights; (2) CE using weights rescaled to the effective sample size (ESS); (3) robust sandwich estimators; and (4) bootstrapping. Several variants of sandwich estimators and bootstrap methods were tested. Performance was quantified on the basis of empirical coverage probabilities for 95% confidence intervals and variability ratios. Variability was underestimated by CE + raw weights when population overlap was poor or moderate. Despite several theoretical limitations, CE + ESS weights accurately estimated uncertainty across most scenarios. Original implementations of sandwich estimators had a downward bias in MAICs with a small ESS, and finite sample adjustments led to marked improvements. Bootstrapping was unstable if population overlap was poor and the sample size was limited. All methods produced valid coverage probabilities and standard errors in cases of strong population overlap. Our findings indicate that the sample size, population overlap, and outcome type are important considerations for variance estimation in MAICs.

Network meta-analysis (NMA) incorporates all available evidence into a general statistical framework for comparing multiple treatments. Standard NMAs make three major assumptions, namely homogeneity, similarity, and consistency, and violating these assumptions threatens an NMA's validity. In this article, we suggest a graphical approach to assessing these assumptions and distinguishing between qualitative and quantitative versions of these assumptions. In our plot, the absolute effect of each treatment arm is plotted against the level of effect modifiers, and the three assumptions of NMA can then be visually evaluated. We use four hypothetical scenarios to show how violating these assumptions can lead to different consequences and difficulties in interpreting an NMA. We present an example of an NMA evaluating steroid use to treat septic shock patients to demonstrate how to use our graphical approach to assess an NMA's assumptions and how this approach can help with interpreting the results. We also show that all three assumptions of NMA can be summarized as an exchangeability assumption. Finally, we discuss how reporting of NMAs can be improved to increase transparency of the analysis and interpretability of the results.

The development of new statistical models for the meta-analysis of diagnostic test accuracy studies is still an ongoing field of research, especially with respect to summary receiver operating characteristic (ROC) curves. In the recently published updated version of the "Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy", the authors point to the challenges of this kind of meta-analysis and propose two approaches. However, both of them come with some disadvantages, such as the nonstraightforward choice of priors in Bayesian models or the requirement of a two-step approach where parameters are estimated for the individual studies, followed by summarizing the results. As an alternative, we propose a novel model by applying methods from time-to-event analysis. To this task we use the discrete proportional hazard approach to treat the different diagnostic thresholds, that provide means to estimate sensitivity and specificity and are reported by the single studies, as categorical variables in a generalized linear mixed model, using both the logit- and the asymmetric cloglog-link. This leads to a model specification with threshold-specific discrete hazards, avoiding a linear dependency between thresholds, discrete hazard, and sensitivity/specificity and thus increasing model flexibility. We compare the resulting models to approaches from the literature in a simulation study. While the estimated area under the summary ROC curve is estimated comparably well in most approaches, the results depict substantial differences in the estimated sensitivities and specificities. We also show the practical applicability of the models to data from a meta-analysis for the screening of type 2 diabetes.