Extending DerSimonian and Laird's methodology to perform multivariate random effects meta-analyses

A non-iterative extension of the multivariate random effects meta-analysis

Multivariate methods in meta-analysis are becoming popular and more accepted in biomedical research despite computational issues in some of the techniques. A number of approaches, both iterative and non-iterative, have been proposed including the multivariate DerSimonian and Laird method by Jackson et al. (2010), which is non-iterative. In this study, we propose an extension of the method by Hartung and Makambi (2002) and Makambi (2001) to multivariate situations. A comparison of the bias and mean square error from a simulation study indicates that, in some circumstances, the proposed approach perform better than the multivariate DerSimonian-Laird approach. An example is presented to demonstrate the application of the proposed approach.

Meta-analysis of sequencing studies with heterogeneous genetic associations

Recent advances in sequencing technologies have made it possible to explore the influence of rare variants on complex diseases and traits. Meta-analysis is essential to this exploration because large sample sizes are required to detect rare variants. Several methods are available to conduct meta-analysis for rare variants under fixed-effects models, which assume that the genetic effects are the same across all studies. In practice, genetic associations are likely to be heterogeneous among studies because of differences in population composition, environmental factors, phenotype and genotype measurements, or analysis method. We propose random-effects models which allow the genetic effects to vary among studies and develop the corresponding meta-analysis methods for gene-level association tests. Our methods take score statistics, rather than individual participant data, as input and thus can accommodate any study designs and any phenotypes. We produce the random-effects versions of all commonly used gene-level association tests, including burden, variable threshold, and variance-component tests. We demonstrate through extensive simulation studies that our random-effects tests are substantially more powerful than the fixed-effects tests in the presence of moderate and high between-study heterogeneity and achieve similar power to the latter when the heterogeneity is low. The usefulness of the proposed methods is further illustrated with data from National Heart, Lung, and Blood Institute Exome Sequencing Project (NHLBI ESP). The relevant software is freely available.

Alcohol consumption over time and risk of death: a systematic review and meta-analysis

The results from the few cohort studies that have measured usual alcohol consumption over time have not been summarized. We therefore conducted a systematic review and meta-analysis to quantify mortality risk. Pertinent studies were identified by searching the Medline, Web of Science, Cumulative Index to Nursing and Allied Health Literature (CINAHL) Plus, and Scopus databases through August 2012 using broad search criteria. Studies reporting relative mortality risks for quantitatively defined categories of alcohol consumption over time were eligible. Nine cohort studies published during 1991-2010 (comprising 62,950 participants and 10,490 deaths) met the inclusion criteria. For men, there was weak evidence of lower mortality risk with low levels of alcohol intake over time but higher mortality risk for those with intakes over 40 g/day compared with abstainers using a random-effects model (P for nonlinearity = 0.02). The pooled relative risks were 0.90 (95% confidence interval: 0.81, 0.99) for 1-29 g/day, 1.19 (95% confidence interval: 0.89, 1.58) for 30-59 g/day, and 1.52 (95% confidence interval: 0.78, 2.98) for 60 or more g/day compared with abstention. There was moderate between-study heterogeneity but no evidence of publication bias. Studies including women were extremely scarce. Our findings include a curvilinear association between drinking over time and mortality risk for men overall and widespread disparity in methods used to capture exposure and report results.

BMI and all-cause mortality in older adults: a meta-analysis

BACKGROUND

Whether the association between body mass index (BMI) and all-cause mortality for older adults is the same as for younger adults is unclear.

OBJECTIVE

The objective was to determine the association between BMI and all-cause mortality risk in adults ≥65 y of age.

DESIGN

A 2-stage random-effects meta-analysis was performed of studies published from 1990 to 2013 that reported the RRs of all-cause mortality for community-based adults aged ≥65 y.

RESULTS

Thirty-two studies met the inclusion criteria; these studies included 197,940 individuals with an average follow-up of 12 y. With the use of a BMI (in kg/m2) of 23.0-23.9 as the reference, there was a 12% greater risk of mortality for a BMI range of 21.0-21.9 and a 19% greater risk for a range of 20.0-20.9 [BMI of 21.0-21.9; HR (95% CI): 1.12 (1.10, 1.13); BMI of 20.0-20.9; HR (95% CI): 1.19 (1.17, 1.22)]. Mortality risk began to increase for BMI >33.0 [BMI of 33.0-33.9; HR (95% CI): 1.08 (1.00, 1.15)]. Self-reported anthropometric measurements, adjustment for intermediary factors, and exclusion of early deaths or preexisting disease did not markedly alter the associations, although there was a slight attenuation of the association in never-smokers.

CONCLUSIONS

For older populations, being overweight was not found to be associated with an increased risk of mortality; however, there was an increased risk for those at the lower end of the recommended BMI range for adults. Because the risk of mortality increased in older people with a BMI <23.0, it would seem appropriate to monitor weight status in this group to address any modifiable causes of weight loss promptly with due consideration of individual comorbidities.

Body mass index and risk of renal cell cancer: a dose-response meta-analysis of published cohort studies

Obesity is accepted as one of the major risk factors for renal cell cancer (RCC). However, conflicting results persist for the pooled risks based on the results from case-control and cohort studies combined, and the exact shape of the dose-response relationship has not been clearly defined yet. To help elucidate the role of obesity, PubMed and Embase databases were searched for published cohort studies on associations between body mass index (BMI) and risk of RCC. Random-effects models and dose-response meta-analyses were used to pool study results. Subgroup analyses were conducted by the available characteristics of studies and participants. Cohort studies (21) with 15,144 cases and 9,080,052 participants were identified. Compared to normal weight, the pooled relative risks and the corresponding 95% confidence intervals of RCC were 1.28(1.24-1.33) for preobesity and 1.77(1.68-1.87) for obesity, respectively. A nonlinear dose-response relationship was also found for RCC risk with BMI (p = 0.000), and the risk increased by 4% for each 1 kg/m(2) increment in BMI. There was no significant between-study heterogeneity among studies (I(2) = 35.6% for preobesity and I(2) = 44.2% for obesity, respectively). Subgroup analysis showed a basically consistent result with the overall analysis. These results suggest that increased BMI are associated with increased risk of RCC both for men and women.

A multivariate model for the meta-analysis of study level survival data at multiple times

Motivated by our meta-analytic dataset involving survival rates after treatment for critical leg ischemia, we develop and apply a new multivariate model for the meta-analysis of study level survival data at multiple times. Our data set involves 50 studies that provide mortality rates at up to seven time points, which we model simultaneously, and we compare the results to those obtained from standard methodologies. Our method uses exact binomial within-study distributions and enforces the constraints that both the study specific and the overall mortality rates must not decrease over time. We directly model the probabilities of mortality at each time point, which are the quantities of primary clinical interest. We also present I² statistics that quantify the impact of the between-study heterogeneity, which is very considerable in our data set. © 2014 The Authors. Research Synthesis Methods published by John Wiley & Sons, Ltd.

Dose-response relationship of serum uric acid levels with risk of stroke mortality

INTRODUCTION

Results from the recent meta-analysis suggested that higher serum uric acid (SUA) levels are positively associated with risk of stroke. However, the relationship of SUA levels with risk of stroke is still unclear.

MATERIALS AND METHODS

Data from prospective cohort studies on SUA levels and risk of stroke mortality was used. Dose-response relationship was assessed by restricted cubic spline model and multivariate random effect meta-regression.

RESULTS

A non-linear relationship (Pfor non-linearity = 0.004) of SUA levels with risk of stroke mortality was found for men, and the relative risk (RR) with 95% confidence interval (CI) of stroke mortality was 1.00 (0.99-1.01), 0.99 (0.94-1.04), 0.98 (0.91-1.06), 1.00 (0.90-1.12), 1.17 (1.09-1.24) and 1.52 (1.33-1.78) for 2, 3, 4, 5, 6 and 7 mg/dL of SUA levels, respectively. For women, the departure from linearity was not significant (Pfor non-linearity = 0.67), and the RR (95 %CI) of stroke mortality was 1.02 (0.99-1.04), 1.10 (0.97-1.20), 1.15 (0.96-1.37), 1.25 (1.09-1.44), 1.39 (1.28-1.50) for 2, 3, 4, 5, 6 mg/dL of SUA levels, respectively.

CONCLUSIONS

Different dose-response relationships of SUA levels with risk of stroke mortality might exist for men and for women. Dose-response relationship of SUA levels with risk of stroke incidence needs to be explored.

Statistics for quantifying heterogeneity in univariate and bivariate meta-analyses of binary data: the case of meta-analyses of diagnostic accuracy

Heterogeneity in diagnostic meta-analyses is common because of the observational nature of diagnostic studies and the lack of standardization in the positivity criterion (cut-off value) for some tests. So far the unexplained heterogeneity across studies has been quantified by either using the I(2) statistic for a single parameter (i.e. either the sensitivity or the specificity) or visually examining the data in a receiver-operating characteristic space. In this paper, we derive improved I(2) statistics measuring heterogeneity for dichotomous outcomes, with a focus on diagnostic tests. We show that the currently used estimate of the 'typical' within-study variance proposed by Higgins and Thompson is not able to properly account for the variability of the within-study variance across studies for dichotomous variables. Therefore, when the between-study variance is large, the 'typical' within-study variance underestimates the expected within-study variance, and the corresponding I(2) is overestimated. We propose to use the expected value of the within-study variation in the construction of I(2) in cases of univariate and bivariate diagnostic meta-analyses. For bivariate diagnostic meta-analyses, we derive a bivariate version of I(2) that is able to account for the correlation between sensitivity and specificity. We illustrate the performance of these new estimators using simulated data as well as two real data sets.

Accuracy of computational soft tissue predictions in orthognathic surgery from three-dimensional photographs 6 months after completion of surgery: a preliminary study of 13 patients

BACKGROUND

This study aimed at a preliminary evaluation of the accuracy of computed three-dimensional (3D) predictions in orthognathic surgery by comparing predicted and real postoperative results.

METHODS

Pre- and postoperative 3D photographs and time-matching computed tomography (CT) and cone-beam CT scans of the face of 13 patients with dentofacial deformities were analyzed. Three-dimensional photographs were fused with preoperative CT data using dedicated software (3dMDvultus, version 2.2.0.8). Postoperative CT data were superposed on the preoperative skull. With an activated rendering function, the osteotomies were simulated in the preoperative CT data and the bony segments moved to their real postoperative position, resulting in a textured soft tissue prediction. This computed skin surface was compared with the real postoperative result by dividing the face into a surgically treated lower half and an untreated upper half. A statistical quantitative analysis of the surfaces was performed.

RESULTS

The mean differences between surfaces were +0.27 mm for the untreated upper half and -0.64 mm for the surgically treated lower half (p < 0.001). Averaged distributions of absolute errors showed more discrepancies between predicted and real postoperative results in the lower half of the face. Errors exceeding 3 mm were encountered in 4 % of the upper halves versus 29.8 % of the lower halves (p < 0.001).

CONCLUSIONS

The accuracy of a specific software platform for predicting 3D soft tissue changes after surgery was insufficient.

LEVEL OF EVIDENCE IV

This journal requires that authors assign a level of evidence to each article. For a full description of these Evidence-Based Medicine ratings, please refer to the Table of Contents or the online Instructions to Authors www.springer.com/00266 .

Red meat and processed meat consumption and all-cause mortality: a meta-analysis

High consumption of red meat and processed meat has been associated with increased risk of several chronic diseases. We conducted a meta-analysis to summarize the evidence from prospective studies on red meat and processed meat consumption in relationship to all-cause mortality. Pertinent studies were identified by searching PubMed through May 2013 and by reviewing the reference lists of retrieved articles. Prospective studies that reported relative risks with 95% confidence intervals for the association of red meat or processed meat consumption with all-cause mortality were eligible. Study-specific results were combined by using a random-effects model. Nine prospective studies were included in the meta-analysis. The summary relative risks of all-cause mortality for the highest versus the lowest category of consumption were 1.10 (95% confidence interval (CI): 0.98, 1.22; n = 6 studies) for unprocessed red meat, 1.23 (95% CI: 1.17, 1.28; n = 6 studies) for processed meat, and 1.29 (95% CI: 1.24, 1.35; n = 5 studies) for total red meat. In a dose-response meta-analysis, consumption of processed meat and total red meat, but not unprocessed red meat, was statistically significantly positively associated with all-cause mortality in a nonlinear fashion. These results indicate that high consumption of red meat, especially processed meat, may increase all-cause mortality.

Dietary fiber intake and risk of type 2 diabetes: a dose-response analysis of prospective studies

Observational studies suggest an association between dietary fiber intake and risk of type 2 diabetes, but the results are inconclusive. We conducted a meta-analysis of prospective studies evaluating the associations of dietary fiber intake and risk of type 2 diabetes. Relevant studies were identified by searching EMBASE (from 1974 to April 2013) and PubMed (from 1966 to April 2013). The fixed or random-effect model was selected based on the homogeneity test among studies. In addition, a 2-stage random-effects dose-response meta-analysis was performed. We identified 17 prospective cohort studies of dietary fiber intake and risk of type 2 diabetes involving 19,033 cases and 488,293 participants. The combined RR (95 % CI) of type 2 diabetes for intake of total dietary fiber, cereal fiber, fruit fiber and insoluble fiber was 0.81 (0.73-0.90), 0.77 (0.69-0.85), 0.94 (0.88-0.99) and 0.75 (0.63-0.89), respectively. A nonlinear relationship was found of total dietary fiber intake with risk of type 2 diabetes (P for nonlinearity < 0.01), and the RRs (95 % CI) of type 2 diabetes were 0.98 (0.90-1.06), 0.97 (0.87-1.07), 0.89 (0.80-0.99), 0.76 (0.65-0.88), and 0.66 (0.53-0.82) for 15, 20, 25, 30, and 35 g/day. The departure from nonlinear relationship was not significant (P for nonlinearity = 0.72), and the risk of type 2 diabetes decreased by 6 % (RR 0.94, 95 % CI 0.93-0.96) for 2 g/day increment in cereal fiber intake. Findings from this meta-analysis indicate that the intakes of dietary fiber may be inversely associated with risk of type 2 diabetes.

Systematic review of the effect of pneumococcal conjugate vaccine dosing schedules on immunogenicity

BACKGROUND

Despite the breadth of studies demonstrating benefits of pneumococcal conjugate vaccine (PCV), uncertainty remains regarding the optimal PCV dosing schedule in infants.

METHODS

We conducted a systematic literature review of PCV immunogenicity published from 1994 to 2010 (supplemented post hoc with studies from 2011). Studies included for analysis evaluated ≥2 doses of 7-valent or higher product (excluding Aventis-Pasteur PCV11) administered to nonhigh-risk infants ≤6 months of age. Impact of PCV schedule on geometric mean antibody concentration (GMC) and proportion of subjects over 0.35 mcg/mL were assessed at various time points; the GMC 1 month postdose 3 (for various dosing regimens) for serotypes 1, 5, 6B, 14, 19F and 23F was assessed in detail using random effects linear regression, adjusted for product, acellular diphtheria-tetanus-pertussis/whole-cell diphtheria- tetanus-pertussis coadministration, laboratory method, age at first dose and geographic region.

RESULTS

From 61 studies, we evaluated 13 two-dose (2+0) and 65 three-dose primary schedules (3+0) without a booster dose, 11 "2+1" (2 primary plus booster) and 42 "3+1" schedules. The GMC after the primary series was higher following 3-dose schedules compared with 2-dose schedules for all serotypes except for serotype 1. Pre- and postbooster GMCs were generally similar regardless of whether 2 or 3 primary doses were given. GMCs were significantly higher for all serotypes when dose 3 was administered in the second year (2+1) compared with ≤6 months of age (3+0).

CONCLUSIONS

While giving the third dose in the second year of life produces a higher antibody response than when given as part of the primary series in the first 6 months, the lower GMC between the 2-dose primary series and booster may result in less disease protection for infants in that interval than those who completed the 3-dose primary series. Theoretical advantages of higher antibodies induced by giving the third dose in the second year of life, such as increased protection against serotype 1 disease, longer duration of protection or more rapid induction of herd effects, need to be evaluated in practice.

Tea consumption and risk of type 2 diabetes: a dose-response meta-analysis of cohort studies

Tea consumption has inconsistently been shown to be associated with the risk of type 2 diabetes (T2D). The aim of the present study was to conduct a dose-response meta-analysis of cohort studies assessing the association between consumption of tea and risk of developing T2D. Pertinent studies were identified by searching PubMed, Web of Science and EMBASE through 31 March 2013. A total of sixteen cohorts from fifteen articles that reported 37,445 cases of diabetes among 545,517 participants were included. A significant linearly inverse association between tea consumption and T2D risk was found (P for linear trend = 0.02). An increase of 2 cups/d in tea consumption was found to be associated with a 4.6 (95% CI 0.9, 8.1) % reduced risk of T2D. On the basis of the dose-response meta-analysis, the predicted relative risks of diabetes for 0, 1, 2, 3, 4, 5 and 6 cups of tea consumed per d were 1.00 (referent), 0.97 (95% CI 0.94, 1.01), 0.95(95% CI 0.92, 0.98), 0.93 (95% CI 0.88, 0.98), 0.90 (95% CI 0.85, 0.96), 0.88 (95 % CI 0.83, 0.93) and 0.85 (95% CI 0.80, 0.91), respectively. There was a statistically significant heterogeneity within the selected studies (Q= 45.32, P< 0.001, I 2= 60.3 %). No evidence of substantial small-study bias was found (P= 0·46). Our findings suggest that tea consumption could be linearly inversely associated with T2D risk. Future well-designed observational studies that account for different characteristics of tea such as tea types, preparation methods and tea strength are needed to fully characterise such an association.

An empirical comparison of univariate and multivariate meta-analyses for categorical outcomes

Treatment effects for multiple outcomes can be meta-analyzed separately or jointly, but no systematic empirical comparison of the two approaches exists. From the Cochrane Library of Systematic Reviews, we identified 45 reviews, including 1473 trials and 258,675 patients, that contained two or three univariate meta-analyses of categorical outcomes for the same interventions that could also be analyzed jointly. Eligible were meta-analyses with at least seven trials reporting all outcomes for which the cross-classification tables were exactly recoverable (e.g., outcomes were mutually exclusive, or one was a subset of the other). This ensured known correlation structures. Outcomes in 40 reviews had an is-subset-of relationship, and those in 5 were mutually exclusive. We analyzed these data with univariate and multivariate models based on discrete and approximate likelihoods. Discrete models were fit in the Bayesian framework using slightly informative priors. The summary effects for each outcome were similar with univariate and multivariate meta-analyses (both using the approximate and discrete likelihoods); however, the multivariate model with the discrete likelihood gave smaller between-study variance estimates, and narrower predictive intervals for new studies. When differences in the summary treatment effects were examined, the multivariate models gave similar summary estimates but considerably longer (shorter) uncertainty intervals because of positive (negative) correlation between outcome treatment effects. It is unclear whether any of the examined reviews would change their overall conclusions based on multivariate versus univariate meta-analyses, because extra-analytical and context-specific considerations contribute to conclusions and, secondarily, because numerical differences were often modest.

Meta-analysis and The Cochrane Collaboration: 20 years of the Cochrane Statistical Methods Group

The Statistical Methods Group has played a pivotal role in The Cochrane Collaboration over the past 20 years. The Statistical Methods Group has determined the direction of statistical methods used within Cochrane reviews, developed guidance for these methods, provided training, and continued to discuss and consider new and controversial issues in meta-analysis. The contribution of Statistical Methods Group members to the meta-analysis literature has been extensive and has helped to shape the wider meta-analysis landscape.In this paper, marking the 20th anniversary of The Cochrane Collaboration, we reflect on the history of the Statistical Methods Group, beginning in 1993 with the identification of aspects of statistical synthesis for which consensus was lacking about the best approach. We highlight some landmark methodological developments that Statistical Methods Group members have contributed to in the field of meta-analysis. We discuss how the Group implements and disseminates statistical methods within The Cochrane Collaboration. Finally, we consider the importance of robust statistical methodology for Cochrane systematic reviews, note research gaps, and reflect on the challenges that the Statistical Methods Group faces in its future direction.

Coffee and caffeine intake and incidence of type 2 diabetes mellitus: a meta-analysis of prospective studies

PURPOSE

Coffee and caffeine have been linked to type 2 diabetes mellitus (T2DM). A dose-response meta-analysis of prospective studies was conducted to assess the association between coffee and caffeine intake and T2DM incidence.

METHODS

Pertinent studies were identified by a search of PubMed and EMBASE. The fixed- or random-effect pooled measure was selected based on between-study heterogeneity. Dose-response relationship was assessed by restricted cubic spline.

RESULTS

Compared with the lowest level, the pooled relative risk (95 % CI) of T2DM was 0.71 (0.67-0.76) for the highest level of coffee intake (26 articles involving 50,595 T2DM cases and 1,096,647 participants), 0.79 (0.69-0.91) for the highest level of decaffeinated coffee intake (10 articles involving 29,165 T2DM cases and 491,485 participants) and 0.70 (0.65-0.75) for the highest level of caffeine intake (6 articles involving 9,302 T2DM cases and 321,960 participants). The association of coffee, decaffeinated coffee and caffeine intake with T2DM incidence was stronger for women than that for men. A stronger association of coffee intake with T2DM incidence was found for non-smokers and subjects with body mass index <25 kg/m(2). Dose-response analysis suggested that incidence of T2DM decreased by 12 % [0.88 (0.86-0.90)] for every 2 cups/day increment in coffee intake, 11 % [0.89 (0.82-0.98)] for every 2 cups/day increment in decaffeinated coffee intake and 14 % [0.86 (0.82-0.91)] for every 200 mg/day increment in caffeine intake.

CONCLUSIONS

Coffee and caffeine intake might significantly reduce the incidence of T2DM.

Association of inorganic arsenic exposure with type 2 diabetes mellitus: a meta-analysis

BACKGROUND

The association of long-term effects of inorganic arsenic (iAs) exposure with type 2 diabetes mellitus (T2DM) risk remains controversial.

METHODS

A literature search was performed in PubMed, China National Knowledge Infrastructure and Web of Knowledge for relevant available articles published in English or Chinese from 1 January 1990 to 5 June 2013. Case-control, cohort or cross-sectional studies evaluating iAs and T2DM were included. The DerSimonian and Laird random effect model was adopted as the pooling method. Dose-response relationship was assessed by restricted cubic spline model and multivariate random-effect meta-regression.

RESULTS

Of the 569 articles identified through searching databases, 17 published articles with 2,243,745 participants for iAs in drinking water and 21 083 participants for total arsenic (tAs) in urine were included for this meta-analysis. The pooled relative risk with 95% CI of T2DM for the highest versus lowest category of iAs exposure level in drinking water was 1.75 (1.20 to 2.54). After removing three studies that had a strong effect on heterogeneity, the pooled relative risk was 1.23 (1.12 to 1.36). Dose-response analysis suggested T2DM risk increased by 13% (1.13 (1.00 to 1.27)) for every 100 µg/L increment of iAs in drinking water. Significant association of T2DM risk with tAs in urine was also found 1.28 (1.14 to 1.44).

CONCLUSIONS

This meta-analysis indicates that long-term iAs exposure might be positively associated with T2DM risk.

Short-term associations between fine and coarse particulate matter and hospitalizations in Southern Europe: results from the MED-PARTICLES project

BACKGROUND

Evidence on the short-term effects of fine and coarse particles on morbidity in Europe is scarce and inconsistent.

OBJECTIVES

We aimed to estimate the association between daily concentrations of fine and coarse particles with hospitalizations for cardiovascular and respiratory conditions in eight Southern European cities, within the MED-PARTICLES project.

METHODS

City-specific Poisson models were fitted to estimate associations of daily concentrations of particulate matter with aerodynamic diameter ≤ 2.5 μm (PM2.5), ≤ 10 μm (PM10), and their difference (PM2.5-10) with daily counts of emergency hospitalizations for cardiovascular and respiratory diseases. We derived pooled estimates from random-effects meta-analysis and evaluated the robustness of results to co-pollutant exposure adjustment and model specification. Pooled concentration-response curves were estimated using a meta-smoothing approach.

RESULTS

We found significant associations between all PM fractions and cardiovascular admissions. Increases of 10 μg/m3 in PM2.5, 6.3 μg/m3 in PM2.5-10, and 14.4 μg/m3 in PM10 (lag 0-1 days) were associated with increases in cardiovascular admissions of 0.51% (95% CI: 0.12, 0.90%), 0.46% (95% CI: 0.10, 0.82%), and 0.53% (95% CI: 0.06, 1.00%), respectively. Stronger associations were estimated for respiratory hospitalizations, ranging from 1.15% (95% CI: 0.21, 2.11%) for PM10 to 1.36% (95% CI: 0.23, 2.49) for PM2.5 (lag 0-5 days).

CONCLUSIONS

PM2.5 and PM2.5-10 were positively associated with cardiovascular and respiratory admissions in eight Mediterranean cities. Information on the short-term effects of different PM fractions on morbidity in Southern Europe will be useful to inform European policies on air quality standards.

A refined method for multivariate meta-analysis and meta-regression

Making inferences about the average treatment effect using the random effects model for meta-analysis is problematic in the common situation where there is a small number of studies. This is because estimates of the between-study variance are not precise enough to accurately apply the conventional methods for testing and deriving a confidence interval for the average effect. We have found that a refined method for univariate meta-analysis, which applies a scaling factor to the estimated effects’ standard error, provides more accurate inference. We explain how to extend this method to the multivariate scenario and show that our proposal for refined multivariate meta-analysis and meta-regression can provide more accurate inferences than the more conventional approach. We explain how our proposed approach can be implemented using standard output from multivariate meta-analysis software packages and apply our methodology to two real examples. © 2013 The Authors. Statistics in Medicine published by John Wiley & Sons, Ltd.

Association of vitamin B6, vitamin B12 and methionine with risk of breast cancer: a dose-response meta-analysis

Background:

Epidemiological studies evaluating the association of vitamin B₆, vitamin B₁₂ and methionine with breast cancer risk have produced inconsistent results.

Methods:

Pertinent studies were identified by a search in PubMed and Web of Knowledge. Random-effect model was used. Dose–response relationship was assessed by restricted cubic spline.

Results:

The combined relative risk (95% confidence interval) of breast cancer for the highest vs lowest category of serum pyridoxal 5′-phosphate (PLP, active form of vitamin B₆) levels and dietary methionine intake was 0.80 (0.66–0.98, P=0.03) and 0.94 (0.89–0.99, P=0.03), respectively, and the associations of breast cancer with higher serum PLP levels and dietary methionine intake were significant among post-menopausal women, but not among pre-menopausal women. The inverse association between breast cancer risk and dietary vitamin B₆ intake, serum vitamin B₁₂ levels and dietary vitamin B₁₂ intake was not significant overall. Linear dose–response relationship was found, and the risk of breast cancer decreased by 23% (P<0.00) for every 100 pmol ml⁻¹ increment in PLP levels and 4% (P=0.05) for every 1 g per day increment in dietary methionine intake, respectively.

Conclusion:

Serum PLP levels and methionine intake might be inversely associated with breast cancer risk, especially among postmenopausal women, which need to be confirmed.

Dose-response meta-analysis on coffee, tea and caffeine consumption with risk of Parkinson's disease

AIMS

A dose-response meta-analysis was carried out between Parkinson's disease (PD) risk, and coffee, tea and caffeine consumption.

METHODS

A comprehensive search was carried out to identify eligible studies. The fixed or random effect model was used based on heterogeneity test. The dose-response relationship was assessed by restricted cubic spline.

RESULTS

A total of 13 articles involving 901 764 participants for coffee, eight articles involving 344 895 participants for tea and seven articles involving 492 724 participants for caffeine were included. A non-linear relationship was found between coffee consumption and PD risk overall, and the strength of protection reached the maximum at approximately 3 cups/day (smoking-adjusted relative risk: 0.72, 95% confidence interval 0.65-0.81). A linear relationship was found between tea and caffeine consumption, and PD risk overall, and the smoking-adjusted risk of PD decreased by 26% and 17% for every two cups/day and 200 mg/day increments, respectively. The association of coffee and tea consumption with PD risk was stronger for men than that for women, and the association of caffeine consumption with PD risk was stronger for ever users of hormones than that for never users of hormones among postmenopausal women. The aforementioned associations were weaker for USA relative to Europe or Asia.

CONCLUSIONS

A linear dose-relationship for decreased PD risk with tea and caffeine consumption was found, whereas the strength of protection reached a maximum at approximately 3 cups/day for coffee consumption overall. Further studies are required to confirm the findings.

Serum uric acid levels and incidence of impaired fasting glucose and type 2 diabetes mellitus: a meta-analysis of cohort studies

AIMS

A meta-analysis of cohort studies was conducted to assess the association between serum uric acid (SUA) levels and incidence of impaired fasting glucose (IFG) and type 2 diabetes mellitus (T2DM).

METHODS

A comprehensive search was conducted to identify eligible studies. The fixed or random effect pooled measure was selected based on between-study heterogeneity. Dose-response relationship was assessed by restricted cubic spline model and multivariate random-effect meta-regression.

RESULTS

Twelve studies with fifteen results were included involving 6340 cases and 62,834 participants. The pooled multivariate-adjusted relative risk (RR) (95%CI) of IFG and T2DM for the highest vs. lowest level of SUA was 1.54 (1.41-1.68), I(2)=42.2%. The association was consistent and significant across subgroup analysis. A nonlinear relationship was found of SUA levels with incidence of IFG and T2DM (P<0.01), and the multivariate-adjusted RRs (95%CI) of IFG and T2DM were 1.02 (0.95-1.10), 1.04 (0.94-1.15), 1.10 (0.99-1.22), 1.25 (1.16-1.35), 1.43 (1.31-1.55), 1.50 (1.38-1.63) and 1.49 (1.34-1.67) for 2.5, 3.5, 4.5, 5.5, 6.5, 7.5 and 8.5mg/dl of SUA. The RR (95%CI) of T2DM for the highest vs. lowest level of SUA was 1.67 (1.51-1.86), and a nonlinear relationship was also found between SUA levels and incidence of T2DM.

CONCLUSIONS

SUA levels are positively associated with incidence of IFG and T2DM, and the association might be nonlinear.

Diagnostic accuracy of circulating tumor cells detection in gastric cancer: systematic review and meta-analysis

BACKGROUND

Circulating tumor cells (CTCs) detection has previously been used for diagnosing gastric cancer. However, the previous studies failed to make an agreement whether the detection of CTCs contributes to the diagnosis of gastric cancer.

METHODS

A systematic review and meta-analysis was performed to evaluate the overall accuracy of CTCs detection for diagnosing gastric cancer. PubMed, Embase and the Wanfang database were searched in all languages published up to Oct 2012. The pooled sensitivity (SEN), specificity (SPE), positive and negative likelihood ratios (PLR and NLR, respectively), diagnostic odds ratio (DOR) and summary receiver operating characteristic (sROC) curve were calculated to evaluate the overall test performance.

RESULTS

Twenty studies were included in this systematic review and meta-analysis. The diagnostic value of CTCs detection for the gastric cancer was calculated to evaluate the overall test performance. The summary estimates of The pooled sensitivity, specificity, positive and negative likelihood ratios, diagnostic odds ratio were 0.42 (95% confidence interval (CI), 0.21-0.67), 0.99 (95% CI, 0.96-1.00), 58.2 (95% CI, 9.8-345.9), 0.58 (95% CI, 0.38-0.89), and 100 (95% CI, 15-663), respectively. The summary receiver operating characteristic curve was 0.97 (95% CI, 0.95-0.98). Deek's funnel plot asymmetry test found no evidence of study publication bias in the current study (P = 0.49).

CONCLUSION

This systematic review suggests that CTCs detection alone cannot be recommended as a screening test for gastric cancer. However, it might be used as a noninvasive method for the confirmation of the gastric cancer diagnosis.

Black tea, green tea and risk of breast cancer: an update

Previous meta-analysis indicated conflicting results in case–control versus cohort studies on the association of green tea with breast cancer risk, and conflicting results were also found in case–control versus cohort studies in another meta-analysis on the association of black tea with breast cancer risk. Many studies were published after the previous meta-analysis. Besides, the dose-response relationship of tea consumption with breast cancer risk is unclear. Thus the association of tea consumption with breast cancer risk was assessed incorporating new publications. Summary relative risk (RR) for highest versus lowest level of tea consumption was calculated based on fixed or random effect models. Dose-response relationship was assessed by restricted cubic spline model and multivariate random-effect meta-regression. The combined results from 9 studies suggested no significant association between green tea consumption and breast cancer risk (RR = 0.82, 95% CI = 0.64-1.04). No significant association was found among cohort studies and case-control studies after sensitivity analysis, respectively. A linear but not significant dose-response association was found between green tea consumption and breast cancer risk. The combined results from 25 studies demonstrated no significant association between black tea consumption and breast cancer risk (RR = 0.98, 95% CI = 0.93-1.03), and no significant association was found in subgroup analysis. A linear but not significant dose-response association was found between black tea consumption and breast cancer risk. Based on the current evidence, black tea and green tea might not contribute significantly to breast cancer risk, respectively.

Meta-analysis of contemporary short- and long-term mortality rates in patients diagnosed with critical leg ischaemia

BACKGROUND

Critical leg ischaemia (CLI) has been associated with high mortality rates. There is a lack of contemporary data on both short- and long-term mortality rates in patients diagnosed with CLI.

METHODS

This was a systematic literature search for studies prospectively reporting mortality in patients diagnosed with CLI. Meta-analysis and meta-regression models were developed to determine overall mortality rates and specific patient-related factors that were associated with death.

RESULTS

A total of 50 studies were included in the analysis The estimated probability of all-cause mortality in patients with CLI was 3·7 per cent at 30 days, 17·5 per cent at 1 year, 35·1 per cent at 3 years and 46·2 per cent at 5 years. Men had a statistically significant survival benefit at 30 days and 3 years. The presence of ischaemic heart disease, tissue loss and older age resulted in a higher probability of death at 3 years.

CONCLUSION

Early mortality rates in patients diagnosed with CLI have improved slightly compared with previous historical data, but long-term mortality rates are still high.

Dietary calcium intake and risk of stroke: a dose-response meta-analysis

BACKGROUND

The findings from epidemiologic studies of calcium intake and risk of stroke have been conflicting.

OBJECTIVE

The objective was to conduct a dose-response meta-analysis of prospective studies to assess the association between dietary calcium intake and stroke risk.

DESIGN

Relevant studies were identified by searching PubMed and EMBASE databases until 11 December 2012 and by reviewing the reference lists of relevant articles. Observational prospective studies that reported RRs and 95% CIs for the association of calcium intake with stroke incidence or mortality were eligible. Study-specific RRs were combined by using a random-effects model.

RESULTS

Eleven prospective studies, including 9095 cases of stroke, were included in the meta-analysis. Evidence of a nonlinear association between dietary calcium intake and risk of stroke was found. In a stratified analysis, calcium intake was inversely associated with risk of stroke in populations with a low to moderate average calcium intake (<700 mg/d; RR for a 300-mg/d increase in calcium intake: 0.82; 95% CI: 0.76, 0.88) but was weakly positively associated with risk in populations with a high calcium intake (≥700 mg/d; corresponding RR: 1.03; 95% CI: 1.01, 1.06). An inverse association between calcium intake and risk of stroke was observed only in Asian populations (n = 4; RR for a 300-mg/d increase in calcium intake: 0.78; 95% CI: 0.71, 0.87).

CONCLUSION

These findings suggest that dietary calcium intake may be inversely associated with stroke in populations with low to moderate calcium intakes and in Asian populations.

Individual participant data meta-analysis for a binary outcome: one-stage or two-stage?

BACKGROUND

A fundamental aspect of epidemiological studies concerns the estimation of factor-outcome associations to identify risk factors, prognostic factors and potential causal factors. Because reliable estimates for these associations are important, there is a growing interest in methods for combining the results from multiple studies in individual participant data meta-analyses (IPD-MA). When there is substantial heterogeneity across studies, various random-effects meta-analysis models are possible that employ a one-stage or two-stage method. These are generally thought to produce similar results, but empirical comparisons are few.

OBJECTIVE

We describe and compare several one- and two-stage random-effects IPD-MA methods for estimating factor-outcome associations from multiple risk-factor or predictor finding studies with a binary outcome. One-stage methods use the IPD of each study and meta-analyse using the exact binomial distribution, whereas two-stage methods reduce evidence to the aggregated level (e.g. odds ratios) and then meta-analyse assuming approximate normality. We compare the methods in an empirical dataset for unadjusted and adjusted risk-factor estimates.

RESULTS

Though often similar, on occasion the one-stage and two-stage methods provide different parameter estimates and different conclusions. For example, the effect of erythema and its statistical significance was different for a one-stage (OR = 1.35, [Formula: see text]) and univariate two-stage (OR = 1.55, [Formula: see text]). Estimation issues can also arise: two-stage models suffer unstable estimates when zero cell counts occur and one-stage models do not always converge.

CONCLUSION

When planning an IPD-MA, the choice and implementation (e.g. univariate or multivariate) of a one-stage or two-stage method should be prespecified in the protocol as occasionally they lead to different conclusions about which factors are associated with outcome. Though both approaches can suffer from estimation challenges, we recommend employing the one-stage method, as it uses a more exact statistical approach and accounts for parameter correlation.

Inhaled foreign bodies in children: a global perspective on their epidemiological, clinical, and preventive aspects

CONTEXT

While several articles describe clinical management of foreign bodies injuries in the upper air tract, little epidemiological evidence is available from injury databases.

OBJECTIVE

This article aims to understand the burden of airway FB injuries in high-, low-, and middle-income countries as emerging from scientific literature.

DATA SOURCES

One thousand six hundred ninety-nine published articles 1978-2008.

STUDY SELECTION

A free text search on PubMed database ((foreign bodies) or (foreign body)) and ((aspiration) or (airways) or (tracheobronchial) or (nasal) or (inhalation) or (obstruction) or (choking) or (inhaled) or (aspirations) or (nose) or (throat) or (asphyxiation)) and ((children) or (child)).

DATA EXTRACTION

Information on reported injuries according to country, time period, children sex and age, FB type, site of obstruction, symptoms, signs, diagnostic and therapeutic procedures, delay at the diagnosis, complications, number of deaths.

RESULTS

Serious complications occur both in high-income and low-middle income countries in a considerable proportion of cases (10% and 20%, respectively). Similarly, death is not infrequent (5-7% of cases).

CONCLUSIONS

Few countries have good systematic data collection and there's a lack of sensibility in parents and clinicians in terms of acknowledge of the choking risk. On the contrary, international surveillance systems able to collect information in a standardized way need to be implemented.

Estimating within-study covariances in multivariate meta-analysis with multiple outcomes

Multivariate meta-analysis allows the joint synthesis of effect estimates based on multiple outcomes from multiple studies, accounting for the potential correlations among them. However, standard methods for multivariate meta-analysis for multiple outcomes are restricted to problems where the within-study correlation is known or where individual participant data are available. This paper proposes an approach to approximating the within-study covariances based on information about likely correlations between underlying outcomes. We developed methods for both continuous and dichotomous data and for combinations of the two types. An application to a meta-analysis of treatments for stroke illustrates the use of the approximated covariance in multivariate meta-analysis with correlated outcomes.

Coffee and caffeine intake and breast cancer risk: an updated dose-response meta-analysis of 37 published studies

OBJECTIVE

We conducted an updated meta-analysis to summarize the evidence from published studies regarding the association of coffee and caffeine intake with breast cancer risk.

METHODS

Pertinent studies were identified by a search of PubMed and by reviewing the reference lists of retrieved articles. The fixed or random effect model was used based on heterogeneity test. The dose-response relationship was assessed by restricted cubic spline model and multivariate random-effect meta-regression.

RESULTS

37 published articles, involving 59,018 breast cancer cases and 966,263 participants, were included in the meta-analysis. No significant association was found between breast cancer risk and coffee (RR=0.97, P=0.09), decaffeinated coffee (RR=0.98, P=0.55) and caffeine (RR=0.99, P=0.73), respectively. And the association was still not significant when combining coffee and caffeine (coffee/caffeine) (RR=0.97, P=0.09). However, an inverse association of coffee/caffeine with breast cancer risk was found for postmenopausal women (RR=0.94, P=0.02), and a strong and significant association of coffee with breast cancer risk was found for BRCA1 mutation carriers (RR=0.69, P<0.01). A linear dose-response relationship was found for breast cancer risk with coffee and caffeine, and the risk of breast cancer decreased by 2% (P=0.05) for every 2 cups/day increment in coffee intake, and 1% (P=0.52) for every 200mg/day increment in caffeine intake, respectively.

CONCLUSIONS

Findings from this meta-analysis suggested that coffee/caffeine might be weakly associated with breast cancer risk for postmenopausal women, and the association for BRCA1 mutation carriers deserves further investigation.

A matrix-based method of moments for fitting the multivariate random effects model for meta-analysis and meta-regression

Multivariate meta-analysis is becoming more commonly used. Methods for fitting the multivariate random effects model include maximum likelihood, restricted maximum likelihood, Bayesian estimation and multivariate generalisations of the standard univariate method of moments. Here, we provide a new multivariate method of moments for estimating the between-study covariance matrix with the properties that (1) it allows for either complete or incomplete outcomes and (2) it allows for covariates through meta-regression. Further, for complete data, it is invariant to linear transformations. Our method reduces to the usual univariate method of moments, proposed by DerSimonian and Laird, in a single dimension. We illustrate our method and compare it with some of the alternatives using a simulation study and a real example.

Bayesian multivariate meta-analysis with multiple outcomes

There has been a recent growth in developments of multivariate meta-analysis. We extend the methodology of Bayesian multivariate meta-analysis to the situation when there are more than two outcomes of interest, which is underexplored in the current literature. Our objective is to meta-analyse summary data from multiple outcomes simultaneously, accounting for potential dependencies among the data. One common issue is that studies do not all report all of the outcomes of interests, and we take an approach relying on marginal modelling of only the reported data. We employ a separation prior for the between-study variance-covariance matrix, which offers an improvement on the conventional inverse-Wishart prior, showing robustness in estimation and flexibility in incorporating prior information. Particular challenges arise when the number of outcomes is large relative to the number of studies because the number of parameters in the variance-covariance matrix can become substantial and there can be very little information with which to estimate between-study correlation coefficients. We explore assumptions that reduce the number of parameters in this matrix, including assumptions of homogenous variances, homogenous correlations for certain outcomes and positive correlation coefficients. We illustrate the methods with an example data set from the Cochrane Database of Systematic Reviews.

Reducing and meta-analysing estimates from distributed lag non-linear models

Background

The two-stage time series design represents a powerful analytical tool in environmental epidemiology. Recently, models for both stages have been extended with the development of distributed lag non-linear models (DLNMs), a methodology for investigating simultaneously non-linear and lagged relationships, and multivariate meta-analysis, a methodology to pool estimates of multi-parameter associations. However, the application of both methods in two-stage analyses is prevented by the high-dimensional definition of DLNMs.

Methods

In this contribution we propose a method to synthesize DLNMs to simpler summaries, expressed by a reduced set of parameters of one-dimensional functions, which are compatible with current multivariate meta-analytical techniques. The methodology and modelling framework are implemented in R through the packages dlnm and mvmeta.

Results

As an illustrative application, the method is adopted for the two-stage time series analysis of temperature-mortality associations using data from 10 regions in England and Wales. R code and data are available as supplementary online material.

Discussion and Conclusions

The methodology proposed here extends the use of DLNMs in two-stage analyses, obtaining meta-analytical estimates of easily interpretable summaries from complex non-linear and delayed associations. The approach relaxes the assumptions and avoids simplifications required by simpler modelling approaches.

Physical activity and risk of breast cancer: a meta-analysis of prospective studies

We conducted a meta-analysis to summarize the evidence from prospective studies regarding the association between physical activity and breast cancer risk. A comprehensive search was conducted to identify eligible studies. The fixed or random effect model was used based on heterogeneity test. The dose-response relationship was assessed by restricted cubic spline model and multivariate random-effect meta-regression. Overall, 31 studies with 63,786 cases were included, and the combined relative risk (RR) with 95 % CI of breast cancer was 0.88 (0.85-0.91). In subgroup analysis by activity type, data from 27 studies including 37,568 cases for non-occupational activity (including recreational activity and household activity) and seven studies including 28,268 cases for occupational activity were used, and the RR (95 % CI) of breast cancer was 0.87 (0.83-0.91) and 0.90 (0.83-0.97), respectively. The inverse association was consistent among all subgroups analyses. Stronger association was found for subjects with BMI <25 kg/m(2) [0.72 (0.65-0.81)], premenopausal women [0.77 (0.72-0.84)], and estrogen and progesterone receptor-negative breast cancer [0.80 (0.73-0.87)]. Dose-response analysis suggested that the risk of breast cancer decreased by 2 % (P < 0.00) for every 25 metabolic equivalent (MET)-h/week increment in non-occupational physical activity, 3 % (P < 0.00) for every 10 MET-h/week (roughly equivalent to 4 h/week of walking in 2 miles/h or 1 h/week of running in 6 miles/h) increment in recreational activity, and 5 % (P < 0.00) for every 2 h/week increment in moderate plus vigorous recreational activity, respectively. Physical activity could significantly reduce the risk of breast cancer.

Meta-analysis of two-arm studies: Modeling the intervention effect from survival probabilities

Pooling the hazard ratios is not always feasible in meta-analyses of two-arm survival studies, because the measure of the intervention effect is not systematically reported. An alternative approach proposed by Moodie et al. is to use the survival probabilities of the included studies, all collected at a single point in time: the intervention effect is then summarised as the pooled ratio of the logarithm of survival probabilities (which is an estimator of the hazard ratios when hazards are proportional). In this article, we propose a generalization of this method. By using survival probabilities at several points in time, this generalization allows a flexible modeling of the intervention over time. The method is applicable to partially proportional hazards models, with the advantage of not requiring the specification of the baseline survival. As in Moodie et al.'s method, the study-level factors modifying the survival functions can be ignored as long as they do not modify the intervention effect. The procedures of estimation are presented for fixed and random effects models. Two illustrative examples are presented.

Quantifying the impact of between-study heterogeneity in multivariate meta-analyses

Measures that quantify the impact of heterogeneity in univariate meta-analysis, including the very popular I(2) statistic, are now well established. Multivariate meta-analysis, where studies provide multiple outcomes that are pooled in a single analysis, is also becoming more commonly used. The question of how to quantify heterogeneity in the multivariate setting is therefore raised. It is the univariate R(2) statistic, the ratio of the variance of the estimated treatment effect under the random and fixed effects models, that generalises most naturally, so this statistic provides our basis. This statistic is then used to derive a multivariate analogue of I(2), which we call I(R)(2). We also provide a multivariate H(2) statistic, the ratio of a generalisation of Cochran's heterogeneity statistic and its associated degrees of freedom, with an accompanying generalisation of the usual I(2) statistic, I(H)(2). Our proposed heterogeneity statistics can be used alongside all the usual estimates and inferential procedures used in multivariate meta-analysis. We apply our methods to some real datasets and show how our statistics are equally appropriate in the context of multivariate meta-regression, where study level covariate effects are included in the model. Our heterogeneity statistics may be used when applying any procedure for fitting the multivariate random effects model.

Meta-analysis of prospective trials determining the short- and mid-term effect of elective open and endovascular repair of abdominal aortic aneurysms on quality of life

Background

Repair of an abdominal aortic aneurysm (AAA) is undertaken to prevent rupture. Intervention is by either open repair (OR) or a more minimally invasive endovascular repair (EVAR). Quality-of-life (QoL) analysis is an important health outcome and a number of single studies have assessed QoL following OR and EVAR. This was a meta-analysis of published studies to assess the effect of an intervention on QoL in patients with an AAA.

Methods

A systematic literature search was undertaken for studies prospectively reporting QoL analysis in patients with an AAA undergoing elective intervention. A multivariable meta-analysis model was developed in which the outcomes were mean changes in QoL scores over time, both for all AAA repairs (OR and EVAR) and comparing OR with EVAR.

Results

Data were collated from 16 studies (14 OR, 12 EVAR). The results suggested that treating an AAA had an effect on patient-reported QoL, evident from the statistically significant changes predominantly in domains assessing physical ability and pain. QoL was affected most within the first 3 months after any form of intervention, and was more pronounced following OR. Furthermore, a deterioration in the Physical Component Summary score following an AAA repair (either OR or EVAR) was evident at 12 months after intervention.

Conclusion

Treating an AAA deleteriously affects patient-reported QoL over the first year following intervention.

Comparisons of the strength of associations with future type 2 diabetes risk among anthropometric obesity indicators, including waist-to-height ratio: a meta-analysis

The aim of this meta-analysis was to compare the association of waist-to-height ratio (WHtR) with risk of incident diabetes with the associations of 3 other conventional obesity indicators (body mass index (BMI), waist circumference (WC), and waist-to-hip ratio (WHR)) with risk of incident diabetes. Literature searches in MEDLINE (January 1950 to April 27, 2011) and EMBASE (January 1974 to April 27, 2011) were conducted for prospective studies that made it possible to estimate the relative risk of diabetes per 1-standard deviation increase in WHtR, in addition to the RR of BMI, WC, or WHR. Strength of the estimated pooled relative risk for a 1-standard deviation increase of each indicator (expressed as RR(WHtR), RR(BMI), RR(WC), and RR(WHR)) was compared with a bivariate random-effects model. Pooled relative risks of the 15 eligible studies with 6,472 diabetes cases were 1.62 (95% CI: 1.48, 1.78) for RR(WHtR), 1.55 (95% CI: 1.43, 1.69) for RR(BMI), 1.63 (95% CI: 1.49, 1.79) for RR(WC), and 1.52 (95% CI: 1.40, 1.66) for RR(WHR). WHtR had an association stronger than that of BMI (P<0.001) or WHR (P<0.001). The present meta-analysis showed that WHtR has a modestly but statistically greater importance than BMI and WHR in prediction of diabetes. Nevertheless, measuring height in addition to WC appeared to have no additional benefit.

Bivariate random effects models for meta-analysis of comparative studies with binary outcomes: methods for the absolute risk difference and relative risk

Multivariate meta-analysis is increasingly utilised in biomedical research to combine data of multiple comparative clinical studies for evaluating drug efficacy and safety profile. When the probability of the event of interest is rare, or when the individual study sample sizes are small, a substantial proportion of studies may not have any event of interest. Conventional meta-analysis methods either exclude such studies or include them through ad hoc continuality correction by adding an arbitrary positive value to each cell of the corresponding 2 × 2 tables, which may result in less accurate conclusions. Furthermore, different continuity corrections may result in inconsistent conclusions. In this article, we discuss a bivariate Beta-binomial model derived from Sarmanov family of bivariate distributions and a bivariate generalised linear mixed effects model for binary clustered data to make valid inferences. These bivariate random effects models use all available data without ad hoc continuity corrections, and accounts for the potential correlation between treatment (or exposure) and control groups within studies naturally. We then utilise the bivariate random effects models to reanalyse two recent meta-analysis data sets.

A multivariate meta-analysis approach for reducing the impact of outcome reporting bias in systematic reviews

Multivariate meta-analysis allows the joint synthesis of multiple correlated outcomes from randomised trials, and is an alternative to a separate univariate meta-analysis of each outcome independently. Usually not all trials report all outcomes; furthermore, outcome reporting bias (ORB) within trials, where an outcome is measured and analysed but not reported on the basis of the results, may cause a biased set of the evidence to be available for some outcomes, potentially affecting the significance and direction of meta-analysis results. The multivariate approach, however, allows one to 'borrow strength' across correlated outcomes, to potentially reduce the impact of ORB. Assuming ORB missing data mechanisms, we aim to investigate the magnitude of bias in the pooled treatment effect estimates for multiple outcomes using univariate meta-analysis, and to determine whether the 'borrowing of strength' from multivariate meta-analysis can reduce the impact of ORB. A simulation study was conducted for a bivariate fixed effect meta-analysis of two correlated outcomes. The approach is illustrated by application to a Cochrane systematic review. Results show that the 'borrowing of strength' from a multivariate meta-analysis can reduce the impact of ORB on the pooled treatment effect estimates. We also examine the use of the Pearson correlation as a novel approach for dealing with missing within-study correlations, and provide an extension to bivariate random-effects models that reduce ORB in the presence of heterogeneity.

Systematic review and meta-analysis of detecting galactomannan in bronchoalveolar lavage fluid for diagnosing invasive aspergillosis

BACKGROUND

Bronchoalveolar lavage (BAL) galactomannan (GM) assay has been used for diagnosing invasive aspergillosis (IA). We aimed to derive a definitive estimate of the overall accuracy of BAL-GM for diagnosing IA.

METHODS AND RESULTS

We undertook a systematic review of thirty diagnostic studies that evaluated the BAL-GM assay for diagnosing IA. PubMed and CBM (China Biological Medicine Database) databases were searched for relevant studies published in all languages up until Feb 2012. The pooled diagnostic odds ratio (DOR) and summary receiver operating characteristic (SROC) were constructed for each cutoff value. Additionally, pooled sensitivity (SEN), specificity (SPE), and positive and negative likelihood ratios (PLR and NLR, respectively) were calculated for summarizing overall test performance. Thirty studies were included in this meta-analysis. The summary estimates of pooled DOR, SEN, SPE, PLR, and NLR of the BAL-GM assay (cutoff value 0.5) for proven or probable IA were 52.7 (95% confidence interval (CI) 31.8-87.3), 0.87 (95% CI 0.79-0.92), 0.89 (95% CI 0.85-0.92), 8.0 (95% CI 5.7-11.1) and 0.15 (95% CI 0.10-0.23) respectively. The SROC was 0.94 (95% CI 0.92-0.96). Compared with cutoff value of 0.5, it has higher DOR, SPE and PLR, and similar SEN and NLR with cutoff value of 1.0, which indicated the optimal cutoff value might be 1.0. Compared with BAL-GM, serum GM has a lower SEN and higher SPE, while PCR displays a lower SEN and a similar SPE.

CONCLUSION

With the optimal cutoff value of 1.0, the BAL-GM assay has higher SEN compared to PCR and serum GM test. It is a useful adjunct in the diagnosis of proven and probable IA.