Nuria Riera-Guardia

Purpose: Ascertaining migraine in electronic healthcare data is challenging because of likely diagnosis underrecording and treatment with over‐the‐counter analgesics, which cannot be used as disease proxies. Algorithm‐identified migraine prevalence may depend on algorithm characteristics and target population. Methods: To describe migraine‐identifying algorithms implemented in electronic healthcare data sources and summarize validation results and observed migraine prevalence, we searched PubMed for peer‐reviewed, English‐language, original research articles that identified migraine in adults using electronic algorithms in electronic healthcare data. We summarized algorithms, validation results, and migraine prevalence (PROSPERO: CRD42023491279). Results: Of 360 unique titles and abstracts, 50 articles (14%) were selected for full‐text review; of them, 41 articles (82%) were finally included: 16 were studies conducted in Europe, 13 in North America, and 12 in Asia. Sixteen studies (39%) identified migraine only using diagnosis codes, 5 (12%) only treatments, 9 (22%) diagnosis and/or treatment codes, and 11 (27%) diagnosis codes, treatments, and setting (e.g., primary care, specialist consultation). Reported migraine prevalence in the general population ranged between 4% and 17%. Only two studies reported validation results: one identified prevention‐eligible patients with migraine (positive predictive value [PPV] = 97%), and one identified migraine on the basis of calculated probabilities with PPVs between 74% and 92%. Conclusion: Finding patients with migraine is feasible in various types of data sources; preferred algorithms vary; algorithm performance is mostly unknown. Identifying chronic migraine or other complex types of migraine requires combining diagnosis codes, treatments, and care settings, which is possible in only some data sources.

Ascertaining migraine in electronic healthcare data is challenging because of likely diagnosis underrecording and treatment with over-the-counter analgesics, which cannot be used as disease proxies. Algorithm-identified migraine prevalence may depend on algorithm characteristics and target population. To describe migraine-identifying algorithms implemented in electronic healthcare data sources and summarize validation results and observed migraine prevalence, we searched PubMed for peer-reviewed, English-language, original research articles that identified migraine in adults using electronic algorithms in electronic healthcare data. We summarized algorithms, validation results, and migraine prevalence (PROSPERO: CRD42023491279). Of 360 unique titles and abstracts, 50 articles (14%) were selected for full-text review; of them, 41 articles (82%) were finally included: 16 were conducted in Europe, 13 in North America, and 12 in Asia. Sixteen studies (39%) identified migraine only using diagnosis codes, 5 (12%) only treatments, 9 (22%) diagnosis and/or treatment codes, and 11 (27%) diagnosis codes, treatments, and setting (e.g., primary care, specialist consultation). Reported migraine prevalence in the general population ranged between 4% and 17%. Only 2 studies reported validation results: one identified prevention-eligible patients with migraine (positive predictive value [PPV] = 97%), and one identified migraine on the basis of calculated probabilities with PPVs between 74% and 92%. Finding patients with migraine is feasible in various types of data sources; preferred algorithms vary; algorithm performance is mostly unknown. Identifying chronic migraine or other complex types of migraine requires combining diagnosis codes, treatments, and care settings, which is possible in only some data sources.

The results of observational studies evaluating and comparing the cardiovascular safety of glitazones, metformin and sufonylureas are inconsistent.To conduct and evaluate heterogeneity in a meta-analysis of observational studies on the risk of acute myocardial infarction (AMI) or stroke in patients with type 2 diabetes using non-insulin blood glucose–lowering drugs (NIBGLD). We systematically identified and reviewed studies evaluating NIBGLD in patients with type 2 diabetes indexed in Medline, Embase, or the Cochrane Library that met prespecified criteria. The quality of included studies was assessed with the RTI item bank. Results were combined using fixed- and random-effects models, and the Higgins I 2 statistic was used to evaluate heterogeneity. Sensitivity analyses by study quality were conducted. The summary relative risk (sRR) (95 % CI) of AMI for rosiglitazone versus pioglitazone was 1.13 (1.04–1.24) [I 2 = 55 %]. In the sensitivity analysis, heterogeneity was reduced [I 2 = 16 %]. The sRR (95 % CI) of stroke for rosiglitazone versus pioglitazone was 1.18 (1.02–1.36) [I 2 = 42 %]. There was strong evidence of heterogeneity related to study quality in the comparisons of rosiglitazone versus metformin and rosiglitazone versus sulfonylureas (I 2 ≥ 70 %). The sRR (95 % CI) of AMI for sulfonylurea versus metformin was 1.24 (1.14–1.34) [I 2 = 41 %] and for pioglitazone versus metformin was 1.02 (0.75–1.38) [I 2 = 17 %]. Sensitivity analyses decreased heterogeneity in most comparisons. Sulfonylureas increased the risk of AMI by 24 % compared with metformin; an imprecise point estimate indicated no difference in risk of AMI when comparing pioglitazone with metformin. The presence of heterogeneity precluded any conclusions on the other comparisons. The quality assessment was valuable in identifying methodological problems in the individual studies and for analysing potential sources of heterogeneity.

Background: The study objective was to compare the Newcastle–Ottawa Scale (NOS) and the RTI item bank (RTI-IB) and estimate interrater agreement using the RTI-IB within a systematic review on the cardiovascular safety of glucose-lowering drugs. Methods: We tailored both tools and added four questions to the RTI-IB. Two reviewers assessed the quality of the 44 included studies with both tools, (independently for the RTI-IB) and agreed on which responses conveyed low, unclear, or high risk of bias. For each question in the RTI-IB (n=31), the observed interrater agreement was calculated as the percentage of studies given the same bias assessment by both reviewers; chance-adjusted interrater agreement was estimated with the first-order agreement coefficient (AC1) statistic. Results: The NOS required less tailoring and was easier to use than the RTI-IB, but the RTI-IB produced a more thorough assessment. The RTI-IB includes most of the domains measured in the NOS. Median observed interrater agreement for the RTI-IB was 75% (25th percentile [p25] =61%; p75 =89%); median AC1 statistic was 0.64 (p25 =0.51; p75 =0.86). Conclusion: The RTI-IB facilitates a more complete quality assessment than the NOS but is more burdensome. The observed agreement and AC1 statistic in this study were higher than those reported by the RTI-IB's developers.

Patients with type 2 diabetes mellitus (T2DM) are at high risk of heart failure. A summary of the effects of blood glucose-lowering drugs other than glitazones on the risk of heart failure in routine clinical practice is lacking. The objective of this study was to conduct a systematic review and meta-analysis of observational studies on the risk of heart failure when using blood glucose-lowering drugs. We systematically identified and reviewed cohort and case–control studies in which the main exposure of interest was noninsulin blood glucose-lowering medications in patients with T2DM. We searched Medline, Embase, and the Cochrane Library to identify publications meeting prespecified eligibility criteria. The quality of included studies was assessed with the Newcastle-Ottawa Scale and the RTI item bank. Results were combined using fixed and random-effects models when at least 3 independent data points were available for a drug-drug comparison. The summary relative risk of heart failure in rosiglitazone users versus pioglitazone users (95% CI) was 1.16 (1.05-1.28) (5 cohort studies). Heterogeneity was present (I 2 = 66%). For new users (n = 4) the summary relative risk was 1.21 (1.14-1.30) and the heterogeneity was reduced (I 2 = 31%);. The summary relative risk for rosiglitazone versus metformin was 1.36 (95% CI, 1.17-1.59) (n = 3). The summary relative risk (95% CI) of heart failure in sulfonylureas users versus metformin users was 1.17 (95% CI, 1.06-1.29) (5 cohort studies; I 2 = 24%) and 1.22 (1.02-1.46) when restricted to new users (2 studies). Information on other comparisons was very scarce. Information on dose and duration of treatment effects was lacking for most comparisons. Few studies accounted for disease severity; therefore, confounding by indication might be present in the majority of the within-study comparisons of this meta-analysis. Use of glitazones and sulfonylureas was associated with an increased risk of heart failure compared with metformin use. However, indication bias cannot be ruled out. Ongoing large multidatabase studies will help to evaluate the risk of heart failure in treated patients with diabetes, including those using newer blood glucose-lowering therapies.

We quantified the contribution to pharmacoepidemiologic research of studies conducted in automated health care databases over the last decade. We reviewed abstracts accepted at the International Conference of Pharmacoepidemiology in 2000 and 2011 to identify studies on drug safety, drug utilization, validation of database data, and disease epidemiology conducted in databases. We abstracted data on study objectives, database name and country, and number of databases used in each study. The total number of ICPE abstracts doubled from 389 in 2000 to 806 in 2011. Abstracts on database studies contributed 36 percent (139 of 389) of all abstracts in 2000 and 45 percent (360 of 806) in 2011. The most common study objective in both years was to evaluate drug safety. Abstracts on validation of database data increased from 2 to 39. Multiple-database abstracts increased from 9 to 43. The number of countries contributing to database abstracts increased from 14 to 22; the increase in countries was primarily in Europe and the Asian Pacific region. This study indicates that the use of health care databases in pharmacoepidemiologic research is growing in many countries. The different types of database studies and the number of studies conducted using multiple databases is also increasing, suggesting that larger study populations and greater collaboration among investigators are becoming more common.

As part of the Safety of Non‐Steroidal Anti‐Inflammatory Drugs (SOS) Project, we reviewed the incidence of cardiovascular (CV) and gastrointestinal (GI) events associated with the use of this category of drugs. We collected data from published meta‐analyses (MAs) of clinical trials of nonsteroidal anti‐inflammatory drugs (NSAIDs). The Medline, Cochrane, ISI, and SCOPUS databases were systematically searched for MAs of NSAID clinical trials that could potentially contain data on adverse incidents such as myocardial infarction (MI), cerebrovascular events (CeVs), stroke, thromboembolic events (ThEs), heart failure (HF), gastrointestinal bleeding (GIB), and perforation, ulcer, and bleeding (PUB). From 1,733 identified references, 29 MAs were selected for the review. This allowed 109 estimations of incidence rates of CV adverse events and 26 estimations of incidence rates for GI adverse events. No data were found on hemorrhagic stroke or LGIB. Coxibs were studied in more MAs than traditional NSAIDs were (21 MAs for coxibs vs. 7 for traditional NSAIDs; one meta‐analysis studied both). Many NSAIDs were not considered in any of the MAs. Our systematic review of MAs included information on the incidence of CV and GI events and identified important knowledge gaps regarding, in particular, the CV safety of traditional NSAIDs. Clinical Pharmacology & Therapeutics (2011) 89 6, 855–866. doi:10.1038/clpt.2011.45