Julie McLellan

Cancer outcomes and all-cause mortality in adults allocated to metformin: systematic review and collaborative meta-analysis of randomised clinical trials.

Aims/hypothesisObservational studies suggest that metformin may reduce cancer risk by approximately one-third. We examined cancer outcomes and all-cause mortality in published randomised controlled trials (RCTs).MethodsRCTs comparing metformin with active glucose-lowering therapy or placebo/usual care, with minimum 500 participants and 1-year follow-up, were identified by systematic review. Data on cancer incidence and all-cause mortality were obtained from publications or by contacting investigators. For two trials, cancer incidence data were not available; cancer mortality was used as a surrogate. Summary RRs, 95% CIs and I2statistics for heterogeneity were calculated by fixed effects meta-analysis.ResultsOf 4,039 abstracts identified, 94 publications described 14 eligible studies. RRs for cancer were available from 11 RCTs with 398 cancers during 51,681 person-years. RRs for all-cause mortality were available from 13 RCTs with 552 deaths during 66,447 person-years. Summary RRs for cancer outcomes in people randomised to metformin compared with any comparator were 1.02 (95% CI 0.82, 1.26) across all trials, 0.98 (95% CI 0.77, 1.23) in a subgroup analysis of active-comparator trials and 1.36 (95% CI 0.74, 2.49) in a subgroup analysis of placebo/usual care comparator trials. The summary RR for all-cause mortality was 0.94 (95% CI 0.79, 1.12) across all trials.Conclusions/interpretationMeta-analysis of currently available RCT data does not support the hypothesis that metformin lowers cancer risk by one-third. Eligible trials also showed no significant effect of metformin on all-cause mortality. However, limitations include heterogeneous comparator types, absent cancer data from two trials, and short follow-up, especially for mortality.

Effectiveness of school-based smoking prevention curricula: systematic review and meta-analysis

Objective To assess effectiveness of school-based smoking prevention curricula keeping children never-smokers. Design Systematic review, meta-analysis. Data: MEDLINE (1966+), EMBASE (1974+), Cinahl, PsycINFO (1967+), ERIC (1982+), Cochrane CENTRAL, Health Star, Dissertation Abstracts, conference proceedings. Data synthesis: pooled analyses, fixed-effects models, adjusted ORs. Risk of bias assessed with Cochrane Risk of Bias tool. Setting 50 randomised controlled trials (RCTs) of school-based smoking curricula. Participants Never-smokers age 5–18 (n=143 495); follow-up ≥6 months; all countries; no date/language limitations. Interventions Information, social influences, social competence, combined social influences/competence and multimodal curricula. Outcome measure Remaining a never-smoker at follow-up. Results Pooling all curricula, trials with follow-up ≤1 year showed no statistically significant differences compared with controls (OR 0.91 (0.82 to 1.01)), though trials of combined social competence/social influences curricula had a significant effect on smoking prevention (7 trials, OR 0.59 (95% CI 0.41 to 0.85)). Pooling all trials with longest follow-up showed an overall significant effect in favour of the interventions (OR 0.88 (0.82 to 0.95)), as did the social competence (OR 0.65 (0.43 to 0.96)) and combined social competence/social influences curricula (OR 0.60 (0.43 to 0.83)). No effect for information, social influences or multimodal curricula. Principal findings were not sensitive to inclusion of booster sessions in curricula or to whether they were peer-led or adult-led. Differentiation into tobacco-only or multifocal curricula had a similar effect on the primary findings. Few trials assessed outcomes by gender: there were significant effects for females at both follow-up periods, but not for males. Conclusions RCTs of baseline never-smokers at longest follow-up found an overall significant effect with average 12% reduction in starting smoking compared with controls, but no effect for all trials pooled at ≤1 year. However, combined social competence/social influences curricula showed a significant effect at both follow-up periods. Systematic review registration Cochrane Tobacco Review Group CD001293.

Optimal strategies for monitoring lipid levels in patients at risk or with cardiovascular disease: a systematic review with statistical and cost-effectiveness modelling

BACKGROUND Various lipid measurements in monitoring/screening programmes can be used, alone or in cardiovascular risk scores, to guide treatment for prevention of cardiovascular disease (CVD). Because some changes in lipids are due to variability rather than true change, the value of lipid-monitoring strategies needs evaluation. OBJECTIVE To determine clinical value and cost-effectiveness of different monitoring intervals and different lipid measures for primary and secondary prevention of CVD. DATA SOURCES We searched databases and clinical trials registers from 2007 (including the Cochrane Central Register of Controlled Trials, MEDLINE, EMBASE, the Clinical Trials Register, the Current Controlled Trials register, and the Cumulative Index to Nursing and Allied Health Literature) to update and extend previous systematic reviews. Patient-level data from the Clinical Practice Research Datalink and St Lukes Hospital, Japan, were used in statistical modelling. Utilities and health-care costs were drawn from the literature. METHODS In two meta-analyses, we used prospective studies to examine associations of lipids with CVD and mortality, and randomised controlled trials to estimate lipid-lowering effects of atorvastatin doses. Patient-level data were used to estimate progression and variability of lipid measurements over time, and hence to model lipid-monitoring strategies. Results are expressed as rates of true-/false-positive and true-/false-negative tests for high lipid or high CVD risk. We estimated incremental costs per quality-adjusted life-year. RESULTS A total of 115 publications reported strength of association between different lipid measures and CVD events in 138 data sets. The summary adjusted hazard ratio per standard deviation of total cholesterol (TC) to high-density lipoprotein (HDL) cholesterol ratio was 1.25 (95% confidence interval 1.15 to 1.35) for CVD in a primary prevention population but heterogeneity was high (I(2) = 98%); similar results were observed for non-HDL cholesterol, apolipoprotein B and other ratio measures. Associations were smaller for other single lipid measures. Across 10 trials, low-dose atorvastatin (10 and 20 mg) effects ranged from a TC reduction of 0.92 mmol/l to 2.07 mmol/l, and low-density lipoprotein reduction of between 0.88 mmol/l and 1.86 mmol/l. Effects of 40 mg and 80 mg were reported by one trial each. For primary prevention, over a 3-year period, we estimate annual monitoring would unnecessarily treat 9 per 1000 more men (28 vs. 19 per 1000) and 5 per 1000 more women (17 vs. 12 per 1000) than monitoring every 3 years. However, annual monitoring would also undertreat 9 per 1000 fewer men (7 vs. 16 per 1000) and 4 per 1000 fewer women (7 vs. 11 per 1000) than monitoring at 3-year intervals. For secondary prevention, over a 3-year period, annual monitoring would increase unnecessary treatment changes by 66 per 1000 men and 31 per 1000 women, and decrease undertreatment by 29 per 1000 men and 28 per 1000 men, compared with monitoring every 3 years. In cost-effectiveness, strategies with increased screening/monitoring dominate. Exploratory analyses found that any unknown harms of statins would need utility decrements as large as 0.08 (men) to 0.11 (women) per statin user to reverse this finding in primary prevention. LIMITATION Heterogeneity in meta-analyses. CONCLUSIONS While acknowledging known and potential unknown harms of statins, we find that more frequent monitoring strategies are cost-effective compared with others. Regular lipid monitoring in those with and without CVD is likely to be beneficial to patients and to the health service. Future research should include trials of the benefits and harms of atorvastatin 40 and 80 mg, large-scale surveillance of statin safety, and investigation of the effect of monitoring on medication adherence. STUDY REGISTRATION This study is registered as PROSPERO CRD42013003727. FUNDING The National Institute for Health Research Health Technology Assessment programme.

Performance of point-of-care HbA1c test devices: implications for use in clinical practice – a systematic review and meta-analysis

Abstract Background: Point-of-care (POC) devices could be used to measure hemoglobin A1c (HbA1c) in the doctors’ office, allowing immediate feedback of results to patients. Reports have raised concerns about the analytical performance of some of these devices. We carried out a systematic review and meta-analysis using a novel approach to compare the accuracy and precision of POC HbA1c devices. Methods: Medline, Embase and Web of Science databases were searched in June 2015 for published reports comparing POC HbA1c devices with laboratory methods. Two reviewers screened articles and extracted data on bias, precision and diagnostic accuracy. Mean bias and variability between the POC and laboratory test were combined in a meta-analysis. Study quality was assessed using the QUADAS2 tool. Results: Two researchers independently reviewed 1739 records for eligibility. Sixty-one studies were included in the meta-analysis of mean bias. Devices evaluated were A1cgear, A1cNow, Afinion, B-analyst, Clover, Cobas b101, DCA 2000/Vantage, HemoCue, Innovastar, Nycocard, Quo-Lab, Quo-Test and SDA1cCare. Nine devices had a negative mean bias which was significant for three devices. There was substantial variability in bias within devices. There was no difference in bias between clinical or laboratory operators in two devices. Conclusions: This is the first meta-analysis to directly compare performance of POC HbA1c devices. Use of a device with a mean negative bias compared to a laboratory method may lead to higher levels of glycemia and a lower risk of hypoglycaemia. The implications of this on clinical decision-making and patient outcomes now need to be tested in a randomized trial.

Systematic review and metaanalysis comparing the bias and accuracy of the modification of diet in renal disease and chronic kidney disease epidemiology collaboration equations in community-based populations

BACKGROUND The majority of patients with chronic kidney disease are diagnosed and monitored in primary care. Glomerular filtration rate (GFR) is a key marker of renal function, but direct measurement is invasive; in routine practice, equations are used for estimated GFR (eGFR) from serum creatinine. We systematically assessed bias and accuracy of commonly used eGFR equations in populations relevant to primary care. CONTENT MEDLINE, EMBASE, and the Cochrane Library were searched for studies comparing measured GFR (mGFR) with eGFR in adult populations comparable to primary care and reporting both the Modification of Diet in Renal Disease (MDRD) and the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equations based on standardized creatinine measurements. We pooled data on mean bias (difference between eGFR and mGFR) and on mean accuracy (proportion of eGFR within 30% of mGFR) using a random-effects inverse-variance weighted metaanalysis. We included 48 studies of 26875 patients that reported data on bias and/or accuracy. Metaanalysis of within-study comparisons in which both formulae were tested on the same patient cohorts using isotope dilution-mass spectrometry-traceable creatinine showed a lower mean bias in eGFR using CKD-EPI of 2.2 mL/min/1.73 m2 (95% CI, 1.1-3.2; 30 studies; I2 = 74.4%) and a higher mean accuracy of CKD-EPI of 2.7% (1.6-3.8; 47 studies; I2 = 55.5%). Metaregression showed that in both equations bias and accuracy favored the CKD-EPI equation at higher mGFR values. SUMMARY Both equations underestimated mGFR, but CKD-EPI gave more accurate estimates of GFR.

Kidney age, not kidney disease

KEY POINTS Although a variety of conditions and syndromes may affect the kidneys over either chronic or acute time frames, the term “chronic kidney disease” (CKD) is used to describe a decrease in the filtration ability of the glomerular capillaries in the kidney. The most prevalent forms of CKD

Essential components in natriuretic peptide-guided management of heart failure: an intervention synthesis

Aim To identify the key components of natriuretic peptide (NP)-guided treatment interventions which reduced hospitalisation in patients with heart failure (HF). Methods and results We extracted detailed information on the components of interventions from studies of NP-guided treatment of HF identified in a previous systematic review. We used meta-regression techniques to assess univariate associations between components and the strength of the reduction in HF hospitalisations and all-cause mortality. A Bayesian meta-analysis approach was used to re-estimate study-level effects in order to identify the study with the most effective NP-guided monitoring intervention. Finally, we examined the intervention options common to the studies in which the 95% credible interval excluded no effect. We identified eight components of NP-guided treatment from ten studies. Univariate comparisons produced mainly equivocal results, but single trial choice and common components analysis led to similar conclusions. Using a predefined treatment protocol, setting a stringent NP target (N-terminal pro-B-type natriuretic peptide of 1000 pg/mL or B-type natriuretic peptide 100 pg/mL) and including a relative targetwere potential key components to reducing HF hospitalisations using NP-guided therapy. Conclusion This analysis provides a description of the key components of NP-guided treatment which could help policy makers develop specific recommendations for HF management. Our research suggests that NP-guided interventions could be simplified, but more research in relevant health settings, such as primary care, is required.

21 Restricted meta-analyses versus full meta-analyses: threshold number of studies based on study sample size

Objectives Systematic reviews and meta-analyses are well established as the highest level of evidence in health care, but can be costly in time, money and labour. Meta-analyses have historically advocated ‘more is better’ to approximate the true effect. However, restricting the number of studies in meta-analyses would reduce the time taken to do a review. It has been suggested that sample size is an adequate indicator of effect estimate and larger studies have estimates closer to the true effect. This research’s objective was to generate a minimum threshold number of studies based on study sample size, to use in meta-analyses without comprising their overall conclusions. Method Using a derivation dataset from the Cochrane library, meta-analyses were ranked by sample size. Pooled estimates for the individual meta-analyses were recalculated to obtain a threshold number of larger studies (study sample size) where the change in the effect estimate 95% confidence interval width had stabilised and was minimal. This threshold number of studies was tested for concordance between the original meta-analysis and the restricted meta-analysis in a validation dataset of meta-analyses. Comparisons were made between the paired meta-analyses using two methods: levels of agreement in direction of effect and statistical significance, and correlation of effect estimates. Results The research suggests where studies are ranked by study sample size, nine studies are sufficient to draw the same meta-analysis conclusion in terms of total agreement between restricted and the original meta-analysis 80% of the time. Correlation of the effect estimates was 0.96. Where restricted meta-analysis results are statistically insignificant caution should be taken as there may be a higher chance of disagreement between the paired meta-analyses. Conclusions This research adds to the body of evidence that supports the view that it is possible to use restricted meta-analyses without jeopardising the integrity of the final findings for a given research question.

Extraction of unadjusted estimates of prognostic association for meta-analysis: simulation methods as good alternatives to trend and direct method estimation

OBJECTIVES Systematic reviews and meta-analysis are the standard methods to assess the association between prognostic markers and major events/conditions. However, the summary measures reported are not always explicitly presented and therefore different indirect methods of extracting estimates have been proposed. The aim of this study is to present two new alternative methods for obtaining summary statistics to be included in a meta-analysis of prognostic studies based on simulating individual patient data and to compare them with the already known generalized least squares for trend (glst) estimation method and direct method. STUDY DESIGN AND SETTINGS We have checked the performance of these methods using a between study comparison, including 122 studies, and a within study comparison, based on data from one of the studies. RESULTS The results obtained in this study show that glst estimation method appears to overestimate the effect size when reported information is incomplete. For the within-study comparison, the closest approximation to the direct estimates was obtained using the approach based on simulating individual patient data. CONCLUSION The proposed simulation methods are a good alternative when other well-known indirect methods cannot be used.