Confidence Intervals for Heterogeneity Measures in Meta-analysis

Takkouche, Bahi; Khudyakov, Polyna; Costa, Julián; Spiegelman, Donna

doi:10.1093/aje/kwt060

Cited by 26 publications

(41 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The variance of these two estimators decreased as K increased, resulting in confidence intervals that were less likely to included the true value of R b . The logit transformation did not improve the coverage for the confidence intervals, similar to previously reports for the confidence intervals based uponR I [12]. To illustrate the use ofR b and of its confidence intervals in determining the impact of heterogeneity and to compare it with the two other measures of heterogeneity,R I and I 2 , we considered three metaanalyses, two of which were taken from the metafor R package [13], chosen to cover a range of values with regard to the extent of heterogeneity, number of studies, type and the magnitude of the estimates, and variability in the within-study variances.…”

Section: Results Of the Simulation Studysupporting

confidence: 86%

A new measure of between‐studies heterogeneity in meta‐analysis

Crippa

Khudyakov

Wang

et al. 2016

Statistics in Medicine

View full text Add to dashboard Cite

Assessing the magnitude of heterogeneity in a meta-analysis is important for determining the appropriateness of combining results. The most popular measure of heterogeneity, I(2) , was derived under an assumption of homogeneity of the within-study variances, which is almost never true, and the alternative estimator, R^I, uses the harmonic mean to estimate the average of the within-study variances, which may also lead to bias. This paper thus presents a new measure for quantifying the extent to which the variance of the pooled random-effects estimator is due to between-studies variation, R^b, that overcomes the limitations of the previous approach. We show that this measure estimates the expected value of the proportion of total variance due to between-studies variation and we present its point and interval estimators. The performance of all three heterogeneity measures is evaluated in an extensive simulation study. A negative bias for R^b was observed when the number of studies was very small and became negligible as the number of studies increased, while R^I and I(2) showed a tendency to overestimate the impact of heterogeneity. The coverage of confidence intervals based upon R^b was good across different simulation scenarios but was substantially lower for R^I and I(2) , especially for high values of heterogeneity and when a large number of studies were included in the meta-analysis. The proposed measure is implemented in a user-friendly function available for routine use in r and sas. R^b will be useful in quantifying the magnitude of heterogeneity in meta-analysis and should supplement the p-value for the test of heterogeneity obtained from the Q test. Copyright © 2016 John Wiley & Sons, Ltd.

show abstract

Section: Results Of the Simulation Studysupporting

confidence: 86%

A new measure of between‐studies heterogeneity in meta‐analysis

Crippa

Khudyakov

Wang

et al. 2016

Statistics in Medicine

View full text Add to dashboard Cite

show abstract

“…We tested for effect modification by age, physical activity, family history of diabetes, and BMI, by including cross-product terms. The analysis was carried out separately for each cohort, and the cohort-specific HRs were combined using a fixed-effects model; the Cochrane Q statistic [ 26 ], the I 2 statistic [ 27 ], and the between-study coefficient of variation [ 28 , 29 ] were used to assess heterogeneity among the cohorts. All statistical tests were two-sided (α = 0.05).…”

Section: Methodsmentioning

confidence: 99%

Plant-Based Dietary Patterns and Incidence of Type 2 Diabetes in US Men and Women: Results from Three Prospective Cohort Studies

et al. 2016

View full text Add to dashboard Cite

BackgroundPlant-based diets have been recommended to reduce the risk of type 2 diabetes (T2D). However, not all plant foods are necessarily beneficial. We examined the association of an overall plant-based diet and hypothesized healthful and unhealthful versions of a plant-based diet with T2D incidence in three prospective cohort studies in the US.Methods and FindingsWe included 69,949 women from the Nurses’ Health Study (1984–2012), 90,239 women from the Nurses’ Health Study 2 (1991–2011), and 40,539 men from the Health Professionals Follow-Up Study (1986–2010), free of chronic diseases at baseline. Dietary data were collected every 2–4 y using a semi-quantitative food frequency questionnaire. Using these data, we created an overall plant-based diet index (PDI), where plant foods received positive scores, while animal foods (animal fats, dairy, eggs, fish/seafood, poultry/red meat, miscellaneous animal-based foods) received reverse scores. We also created a healthful plant-based diet index (hPDI), where healthy plant foods (whole grains, fruits, vegetables, nuts, legumes, vegetable oils, tea/coffee) received positive scores, while less healthy plant foods (fruit juices, sweetened beverages, refined grains, potatoes, sweets/desserts) and animal foods received reverse scores. Lastly, we created an unhealthful plant-based diet index (uPDI) by assigning positive scores to less healthy plant foods and reverse scores to healthy plant foods and animal foods.We documented 16,162 incident T2D cases during 4,102,369 person-years of follow-up. In pooled multivariable-adjusted analysis, both PDI and hPDI were inversely associated with T2D (PDI: hazard ratio [HR] for extreme deciles 0.51, 95% CI 0.47–0.55, p trend < 0.001; hPDI: HR for extreme deciles 0.55, 95% CI 0.51–0.59, p trend < 0.001). The association of T2D with PDI was considerably attenuated when we additionally adjusted for body mass index (BMI) categories (HR 0.80, 95% CI 0.74–0.87, p trend < 0.001), while that with hPDI remained largely unchanged (HR 0.66, 95% CI 0.61–0.72, p trend < 0.001). uPDI was positively associated with T2D even after BMI adjustment (HR for extreme deciles 1.16, 95% CI 1.08–1.25, p trend < 0.001). Limitations of the study include self-reported diet assessment, with the possibility of measurement error, and the potential for residual or unmeasured confounding given the observational nature of the study design.ConclusionsOur study suggests that plant-based diets, especially when rich in high-quality plant foods, are associated with substantially lower risk of developing T2D. This supports current recommendations to shift to diets rich in healthy plant foods, with lower intake of less healthy plant and animal foods.

show abstract

“…Binomial proportion CIs for individual studies were calculated using the Clopper-Pearson method, which allows for asymmetry. 23 Between-study heterogeneity was evaluated using standard χ 2 tests and the I 2 statistic. 21 I 2 statistics were calculated to describe the percentages of total variation across studies caused by heterogeneity.…”

Section: Methodsmentioning

confidence: 99%

Prevalence of depression and depressive symptoms among outpatients: a systematic review and meta-analysis

et al. 2017

View full text Add to dashboard Cite

ObjectivesDepression and depressive symptoms are common mental disorders that have a considerable effect on patients’ health-related quality of life and satisfaction with medical care, but the prevalence of these conditions varies substantially between published studies. The aim of this study is to conduct a systematic review and meta-analysis to provide a precise estimate of the prevalence of depression or depressive symptoms among outpatients in different clinical specialties.DesignSystematic review and meta-analysis.Data sources and eligibility criteriaThe PubMed and PsycINFO, EMBASE and Cochrane Library databases were searched to identify observational studies that contained information on the prevalence of depression and depressive symptoms in outpatients. All studies included were published before January 2016. Data characteristics were extracted independently by two investigators. The point prevalence of depression or depressive symptoms was measured using validated self-report questionnaires or structured interviews. Assessments were pooled using a random-effects model. Differences in study-level characteristics were estimated by meta-regression analysis. Heterogeneity was assessed using standard χ2 tests and the I2 statistic. The study protocol has been registered with PROSPERO under number CRD42017054738.ResultsEighty-three cross-sectional studies involving 41 344 individuals were included in this study. The overall pooled prevalence of depression or depressive symptoms was 27.0% (10 943/41 344 individuals; 95% CI 24.0% to 29.0%), with significant heterogeneity between studies (p<0.0001, τ2=0.3742, I2=96.7%). Notably, a significantly higher prevalence of depression and depressive symptoms was observed in outpatients than in the healthy controls (OR 3.16, 95% CI 2.66 to 3.76, I2=72.0%, χ2=25.33). The highest depression/depressive symptom prevalence estimates occurred in studies of outpatients from otolaryngology clinics (53.0%), followed by dermatology clinics (39.0%) and neurology clinics (35.0%). Subgroup analyses showed that the prevalence of depression and depressive symptoms in different specialties varied from 17.0% to 53.0%. The prevalence of depression and depressive symptoms was higher among outpatients in developing countries than in outpatients from developed countries. Moreover, the prevalence of depression and depressive symptoms in outpatients slightly decreased from 1996 to 2010. Regarding screening instruments, the Beck Depression Inventory led to a higher estimate of the prevalence of depression and depressive symptoms (1316/4702, 36.0%, 95% CI 29.0% to 44.0%, I2=94.8%) than the Hospital Anxiety and Depression Scale (1003/2025, 22.0%, 95% CI 12.0% to 35.0%, I2=96.6%).ConclusionOur study provides evidence that a significant proportion of outpatients experience depression or depressive symptoms, highlighting the importance of developing effective management strategies for the early identification and treatment of these conditions among outpatients in clinical practice. The substantial hete...

show abstract

Confidence Intervals for Heterogeneity Measures in Meta-analysis

Cited by 26 publications

References 16 publications

A new measure of between‐studies heterogeneity in meta‐analysis

A new measure of between‐studies heterogeneity in meta‐analysis

Plant-Based Dietary Patterns and Incidence of Type 2 Diabetes in US Men and Women: Results from Three Prospective Cohort Studies

Prevalence of depression and depressive symptoms among outpatients: a systematic review and meta-analysis

Contact Info

Product

Resources

About