Article type
Year
Abstract
Background: Correlation coefficient (r) is a measure of strength and direction of linear association between two variables or used as quantitative surrogate of individual variables. There have been no or few studies in which pooling of r values has been done and hence, we are presenting a meta-analysis of non interventional cohort studies by combining the r values.
Objectives: To meta-analyze correlation coefficients derived from cohort studies.
Methods: We conducted a systematic review on association of anthropometric measures during childhood and risk of becoming overweight or obese in later life. The studies included were prospective cohort or longitudinal studies. Tracking estimates such as the correlation coefficient are preferred compared to difference estimates when assessing body-mass index (BMI) tracking. Basic information recorded consisted of cohort size n, (mean) age at baseline, and at follow-up measurement. The effect size calculated was correlation coefficient (either Pearson's or Spearman's). Standard error was calculated using the information from cohort size and r values. Meta-analysis was done using inverse variance random-effects model. Stata/MP 12.2 software) was used to perform the meta-analysis.
Results: Twenty-eight studies provided correlation coefficient data on BMI tracking from 40,219 individuals for meta-analysis. Follow-up time ranged from two to 65 years. We derived standard errors from the available sample size and r value. Then r values were pooled using the inverse variance random-effects model. The estimate size with confidence intervals provided the net effect.
Conclusions: Pooling of correlation coefficients is a useful and feasible means of meta-analyzing quantitative data from cohort studies.
Acknowledgment: The study was supported by World Health Organization, Geneva, Switzerland and ICMR, New Delhi, India
Objectives: To meta-analyze correlation coefficients derived from cohort studies.
Methods: We conducted a systematic review on association of anthropometric measures during childhood and risk of becoming overweight or obese in later life. The studies included were prospective cohort or longitudinal studies. Tracking estimates such as the correlation coefficient are preferred compared to difference estimates when assessing body-mass index (BMI) tracking. Basic information recorded consisted of cohort size n, (mean) age at baseline, and at follow-up measurement. The effect size calculated was correlation coefficient (either Pearson's or Spearman's). Standard error was calculated using the information from cohort size and r values. Meta-analysis was done using inverse variance random-effects model. Stata/MP 12.2 software) was used to perform the meta-analysis.
Results: Twenty-eight studies provided correlation coefficient data on BMI tracking from 40,219 individuals for meta-analysis. Follow-up time ranged from two to 65 years. We derived standard errors from the available sample size and r value. Then r values were pooled using the inverse variance random-effects model. The estimate size with confidence intervals provided the net effect.
Conclusions: Pooling of correlation coefficients is a useful and feasible means of meta-analyzing quantitative data from cohort studies.
Acknowledgment: The study was supported by World Health Organization, Geneva, Switzerland and ICMR, New Delhi, India