Explore the vast range of titles available.

Background This study developed, calibrated and evaluated a machine learning (ML) classifier designed to reduce study identification workload in maintaining the Cochrane COVID-19 Study Register (CCSR), a continuously updated register of COVID-19 research studies. Methods A ML classifier for retrieving COVID-19 research studies (the ‘Cochrane COVID-19 Study Classifier’) was developed using a data set of title-abstract records ‘included’ in, or ‘excluded’ from, the CCSR up to 18th October 2020, manually labelled by information and data curation specialists or the Cochrane Crowd. The classifier was then calibrated using a second data set of similar records ‘included’ in, or ‘excluded’ from, the CCSR between October 19 and December 2, 2020, aiming for 99% recall. Finally, the calibrated classifier was evaluated using a third data set of similar records ‘included’ in, or ‘excluded’ from, the CCSR between the 4th and 19th of January 2021. Results The Cochrane COVID-19 Study Classifier was trained using 59,513 records (20,878 of which were ‘included’ in the CCSR). A classification threshold was set using 16,123 calibration records (6005 of which were ‘included’ in the CCSR) and the classifier had a precision of 0.52 in this data set at the target threshold recall >0.99. The final, calibrated COVID-19 classifier correctly retrieved 2285 (98.9%) of 2310 eligible records but missed 25 (1%), with a precision of 0.638 and a net screening workload reduction of 24.1% (1113 records correctly excluded). Conclusions The Cochrane COVID-19 Study Classifier reduces manual screening workload for identifying COVID-19 research studies, with a very low and acceptable risk of missing eligible studies. It is now deployed in the live study identification workflow for the Cochrane COVID-19 Study Register.

Background Explicit labelling of lower strength alcohol products could reduce alcohol consumption by attracting more people to buy and drink such products instead of higher strength ones. Alternatively, it may lead to more consumption due to a ‘self-licensing’ mechanism. Equivalent labelling of food or tobacco (for example “Low fat” or “Low tar”) could influence consumption of those products by similar mechanisms. This systematic review examined the effects of ‘Low alcohol’ and equivalent labelling of alcohol, food and tobacco products on selection, consumption, and perceptions of products among adults. Methods A systematic review was conducted based on Cochrane methods. Electronic and snowball searches identified 26 eligible studies. Evidence from 12 randomised controlled trials (all on food) was assessed for risk of bias, synthesised using random effects meta-analysis, and interpreted in conjunction with evidence from 14 non-randomised studies (one on alcohol, seven on food and six on tobacco). Outcomes assessed were: quantities of the product (i) selected or (ii) consumed (primary outcomes - behaviours), (iii) intentions to select or consume the product, (iv) beliefs associated with it consumption, (v) product appeal, and (vi) understanding of the label (secondary outcomes – cognitions). Results Evidence for impacts on the primary outcomes (i.e. amounts selected or consumed) was overall of very low quality, showing mixed effects, likely to vary by specific label descriptors, products and population characteristics. Overall very low quality evidence suggested that exposure to ‘Low alcohol’ and equivalent labelling on alcohol, food and tobacco products can shift consumer perceptions of products, with the potential to ‘self-licence’ excess consumption. Conclusions Considerable uncertainty remains about the effects of labels denoting low alcohol, and equivalent labels, on alcohol, food and tobacco selection and consumption. Independent, high-quality studies are urgently needed to inform policies on labelling regulations.

Background Restricting marketing of alcoholic products is purported to be a cost-effective intervention to reduce alcohol consumption. The strength of evidence supporting this claim is contested. This systematic review aimed to assess immediate effects of exposure to alcohol marketing on alcoholic beverage consumption and related cognitions. Methods Electronic searches of nine databases, supplemented with reference list searches and forward citation tracking, were used to identify randomised, experimental studies assessing immediate effects of exposure to alcohol marketing communications on objective alcohol consumption (primary outcome), explicit or implicit alcohol-related cognitions, or selection without purchasing (secondary outcomes). Study limitations were assessed using the Cochrane Risk of Bias tool. Random and fixed effects meta-analyses were conducted to estimate effect sizes. Results Twenty four studies met the eligibility criteria. A meta-analysis integrating seven studies (758 participants, all students) found that viewing alcohol advertisements increased immediate alcohol consumption relative to viewing non-alcohol advertisements (SMD = 0.20, 95 % CI = 0.05, 0.34). A meta-analysis integrating six studies (631 participants, all students) did not find that viewing alcohol portrayals in television programmes or films increased consumption (SMD = 0.16, 95 % CI = −0.05, 0.37). Meta-analyses of secondary outcome data found that exposure to alcohol portrayals increased explicit alcohol-related cognitions, but did not find that exposure to alcohol advertisements influenced explicit or implicit alcohol-related cognitions. Confidence in results is diminished by underpowered analyses and unclear risk of bias. Conclusions Viewing alcohol advertisements (but not alcohol portrayals) may increase immediate alcohol consumption by small amounts, equivalent to between 0.39 and 2.67 alcohol units for males and between 0.25 and 1.69 units for females. The generalizability of this finding beyond students and to other marketing channels remains to be established.

The third meeting of the International Collaboration for Automation of Systematic Reviews (ICASR) was held 17–18 October 2017 in London, England. ICASR is an interdisciplinary group whose goal is to maximize the use of technology for conducting rapid, accurate, and efficient systematic reviews of scientific evidence. The group seeks to facilitate the development and widespread acceptance of automated techniques for systematic reviews. The meeting’s conclusion was that the most pressing needs at present are to develop approaches for validating currently available tools and to provide increased access to curated corpora that can be used for validation. To that end, ICASR’s short-term goals in 2018–2019 are to propose and publish protocols for key tasks in systematic reviews and to develop an approach for sharing curated corpora for validating the automation of the key tasks.

Trastuzumab improves survival in HER2+ breast cancer patients, with some evidence of adverse cardiac side effects. Current recommendations are to give adjuvant trastuzumab for one year or until recurrence, although trastuzumab treatment for only 9 or 10 weeks has shown similar survival rates to 12-month treatment. We present here a multi-arm joint analysis examining the relative cost-effectiveness of different durations of adjuvant trastuzumab. Network meta-analysis (NMA) was used to examine which trials' data to include in the cost-effectiveness analysis (CEA). A network using FinHer (9 weeks vs. zero) and BCIRG006 (12 months vs. zero) trials offered the only jointly randomisable network so these trials were used in the CEA. The 3-arm CEA compared costs and quality-adjusted life-years (QALYs) associated with zero, 9-week and 12-month adjuvant trastuzumab durations in early breast cancer, using a decision tree followed by a Markov model that extrapolated the results to a lifetime time horizon. Pairwise incremental cost-effectiveness ratios (ICERs) were also calculated for each pair of regimens and used in budget impact analysis, and the Bucher method was used to check face validity of the findings. Addition of the PHARE trial (6 months vs. 12 months) to the network, in order to create a 4-arm CEA including the 6-month regimen, was not possible as late randomisation in this trial resulted in recruitment of a different patient population as evidenced by the NMA findings. The CEA results suggest that 9 weeks' trastuzumab is cost-saving and leads to more QALYs than 12 months', i.e. the former dominates the latter. The cost-effectiveness acceptability frontier (CEAF) favours zero trastuzumab at willingness-to-pay levels below £2,500/QALY and treatment for 9 weeks above this threshold. The combination of the NMA and Bucher investigations suggests that the 9-week duration is as efficacious as the 12-month duration for distant-disease-free survival and overall survival, and safer in terms of fewer adverse cardiac events. Our CEA results suggest that 9-week trastuzumab dominates 12-month trastuzumab in cost-effectiveness terms at conventional thresholds of willingness to pay for a QALY, and the 9-week regimen is also suggested to be as clinically effective as the 12-month regimen according to the NMA and Bucher analyses. This finding agrees with the results of the E2198 head-to-head study that compared 10 weeks' with 14 months' trastuzumab and found no significant difference. Appropriate trial design and reporting is critical if results are to be synthesisable with existing evidence, as selection bias can lead to recruitment of a different patient population from existing trials. Our analysis was not based on head-to-head trials' data, so the results should be viewed with caution. Short-duration trials would benefit from recruiting larger numbers of participants to reduce uncertainty in the synthesised results.

Unhealthy diet and low levels of physical activity are common behavioural factors in the aetiology of many non-communicable diseases. Recent years have witnessed an upsurge of policy and research interest in the use of taxes and other economic instruments to improve population health. To assemble, configure and analyse empirical research studies available to inform the public health case for using economic instruments to promote dietary and physical activity behaviour change. We conducted a systematic scoping review of evidence for the effects of specific interventions to change, or general exposure to variations in, prices or income on dietary and physical activity behaviours and corollary outcomes. Systematic electronic searches and parallel snowball searches retrieved >1 million study records. Text mining technologies were used to prioritise title-abstract records for screening. Eligible studies were selected, classified and analysed in terms of key characteristics and principal findings, using a narrative, configuring synthesis focused on implications for policy and further research. We identified 880 eligible studies, including 192 intervention studies and 768 studies that incorporated evidence for prices or income as correlates or determinants of target outcomes. Current evidence for the effects of economic instruments and exposures on diet and physical activity is limited in quality and equivocal in terms of its policy implications. Direct evidence for the effects of economic instruments is heavily skewed towards impacts on diet, with a relative lack of evidence for impacts on physical activity. The evidence-based case for using economic instruments to promote dietary and physical activity behaviour change may be less compelling than some proponents have claimed. Future research should include measurement of people's actual behavioural responses using study designs capable of generating reliable causal inferences regarding intervention effects. Policy implementation needs to be carefully aligned with evaluation planning and design.

Unsustainable practices in the land use sector contribute to climate change through the release of greenhouse gases. Payment for environmental services (PESs) provide economic incentives to reduce the negative environmental impacts of land use and are a popular approach to mitigate climate change in low‐ and middle‐income countries. Some PES programmes also aim to improve socioeconomic outcomes and reduce poverty. This systematic review examines the effect of programmes on environmental and socioeconomic outcomes. We identified 44 quantitative impact evaluations and 60 qualitative studies of PES programmes for inclusion in the review, to assess both the effects of PES and identify context, design and implementation features that may influence PES effectiveness. The studies covered 18 programmes from 12 countries in Latin America and the Caribbean, East Asia and Pacific, South Asia and Sub‐Saharan Africa. The review finds that PES may increase household income, reduce deforestation and improve forest cover, but the findings are, however, based on low and very low quality evidence from a small number of programmes and should be treated with caution. Qualitative evidence indicates that several factors influence whether PES programmes are likely to be effective in different contexts and suggests that the inclusion of strong governance structures and the effective targeting of both locations and participants may improve intervention effectiveness. Funders, implementing agencies and researchers should collaborate to develop a coordinated programme of rigorous, mixed‐methods impact evaluation implemented across contexts. Until such evidence is available, PES programmes remain a high‐risk strategy for climate change mitigation.

This study developed, calibrated, and evaluated a machine learning classifier designed to reduce study identification workload in Cochrane for producing systematic reviews. A machine learning classifier for retrieving randomized controlled trials (RCTs) was developed (the “Cochrane RCT Classifier”), with the algorithm trained using a data set of title–abstract records from Embase, manually labeled by the Cochrane Crowd. The classifier was then calibrated using a further data set of similar records manually labeled by the Clinical Hedges team, aiming for 99% recall. Finally, the recall of the calibrated classifier was evaluated using records of RCTs included in Cochrane Reviews that had abstracts of sufficient length to allow machine classification. The Cochrane RCT Classifier was trained using 280,620 records (20,454 of which reported RCTs). A classification threshold was set using 49,025 calibration records (1,587 of which reported RCTs), and our bootstrap validation found the classifier had recall of 0.99 (95% confidence interval 0.98–0.99) and precision of 0.08 (95% confidence interval 0.06–0.12) in this data set. The final, calibrated RCT classifier correctly retrieved 43,783 (99.5%) of 44,007 RCTs included in Cochrane Reviews but missed 224 (0.5%). Older records were more likely to be missed than those more recently published. The Cochrane RCT Classifier can reduce manual study identification workload for Cochrane Reviews, with a very low and acceptable risk of missing eligible RCTs. This classifier now forms part of the Evidence Pipeline, an integrated workflow deployed within Cochrane to help improve the efficiency of the study identification processes that support systematic review production. •Systematic review processes need to become more efficient.•Machine learning is sufficiently mature for real-world use.•A machine learning classifier was built using data from Cochrane Crowd.•It was calibrated to achieve very high recall.•It is now live and in use in Cochrane review production systems.

Background There is limited information about how transgender, gender diverse, and Two-Spirit (trans) people have been represented and studied by researchers. The objectives of this study are to (1) map and describe trans research in the social sciences, sciences, humanities, health, education, and business, (2) identify evidence gaps and opportunities for more responsible research with trans people, (3) assess the use of text mining for study identification, and (4) increase access to trans research for key stakeholders through the creation of a web-based evidence map. Methods Study design was informed by community consultations and pilot searches. Eligibility criteria were established to include all original research of any design, including trans people or their health information, and published in English in peer-reviewed journals. A complex electronic search strategy based on relevant concepts in 15 databases was developed to obtain a broad range of results linked to transgender, gender diverse, and Two-Spirit individuals and communities. Searches conducted in early 2015 resulted in 25,242 references after removal of duplicates. Based on the number of references, resources, and an objective to capture upwards of 90% of the existing literature, this study is a good candidate for text mining using Latent Dirichlet Allocation to improve efficiency of the screening process. The following information will be collected for evidence mapping: study topic, study design, methods and data sources, recruitment strategies, sample size, sample demographics, researcher name and affiliation, country where research was conducted, funding source, and year of publication. Discussion The proposed research incorporates an extensive search strategy, text mining, and evidence map; it therefore has the potential to build on knowledge in several fields. Review results will increase awareness of existing trans research, identify evidence gaps, and inform strategic research prioritization. Publishing the map online will improve access to research for key stakeholders including community members, policy makers, and healthcare providers. This study will also contribute to knowledge in the area of text mining for study identification by providing an example of how semi-automation performs for screening on title and abstract and on full text.

Background Meta-research studies investigating methods, systems, and processes designed to improve the efficiency of systematic review workflows can contribute to building an evidence base that can help to increase value and reduce waste in research. This study demonstrates the use of an economic evaluation framework to compare the costs and effects of four variant approaches to identifying eligible studies for consideration in systematic reviews. Methods A cost-effectiveness analysis was conducted using a basic decision-analytic model, to compare the relative efficiency of ‘safety first’, ‘double screening’, ‘single screening’ and ‘single screening with text mining’ approaches in the title-abstract screening stage of a ‘case study’ systematic review about undergraduate medical education in UK general practice settings. Incremental cost-effectiveness ratios (ICERs) were calculated as the ‘incremental cost per citation ‘saved’ from inappropriate exclusion’ from the review. Resource use and effect parameters were estimated based on retrospective analysis of ‘review process’ meta-data curated alongside the ‘case study’ review, in conjunction with retrospective simulation studies to model the integrated use of text mining. Unit cost parameters were estimated based on the ‘case study’ review’s project budget. A base case analysis was conducted, with deterministic sensitivity analyses to investigate the impact of variations in values of key parameters. Results Use of ‘single screening with text mining’ would have resulted in title-abstract screening workload reductions (base case analysis) of >60 % compared with other approaches. Across modelled scenarios, the ‘safety first’ approach was, consistently, equally effective and less costly than conventional ‘double screening’. Compared with ‘single screening with text mining’, estimated ICERs for the two non-dominated approaches (base case analyses) ranged from £1975 (‘single screening’ without a ‘provisionally included’ code) to £4427 (‘safety first’ with a ‘provisionally included’ code) per citation ‘saved’. Patterns of results were consistent between base case and sensitivity analyses. Conclusions Alternatives to the conventional ‘double screening’ approach, integrating text mining, warrant further consideration as potentially more efficient approaches to identifying eligible studies for systematic reviews. Comparable economic evaluations conducted using other systematic review datasets are needed to determine the generalisability of these findings and to build an evidence base to inform guidance for review authors.