ABSTRACT
Publication bias can result from the propensity of researchers to document what is unusual. This can distort the inferences drawn in systematic reviews. To measure the distortion, it has been suggested that a second analysis be done; using weights proportional to the size of the population from which the samples are drawn. We re-evaluate data from a published meta-analysis on prevalence of hepatitis B in India, to see how this approach alters the results. Prevalence of hepatitis B among tribal and non-tribal populations in different States was analyzed. Weights were then assigned according to population of the State. The overall country prevalence was then calculated. Using population-weights it is estimated that the point-prevalence of hepatitis B among non-tribal populations is 3.07% [95% CI: 2.5-3.64]. Among tribal populations it is 11.85% (CI 10.76-12.93). Overall prevalence was 3.70 (CI: 3.17-4.24) (corresponding to a chronic carrier rate of 2.96%). The present analysis using population-weights has resulted in the estimated prevalence among non tribal populations increasing by 24% and that among tribal populations decreasing by 25.5% when compared to figures of the meta-analysis published earlier. The advantages and drawbacks of this procedure are discussed.
Subject(s)
Hepatitis B/ethnology , Data Interpretation, Statistical , Humans , India/epidemiology , Meta-Analysis as Topic , Population Groups/statistics & numerical data , Prevalence , Publication Bias , Residence CharacteristicsABSTRACT
OBJECTIVE: To evaluate the point prevalence of Hepatitis B in India. DESIGN: Meta-analysis of data on point prevalence from different parts of the country. DATA SOURCES: Searches were made in Medline, Cochrane Library and Best bets and previous reviews. A limited hand search of cross references was also done. Finally a consultation with experts was held to enlarge the references base. REVIEW METHODS: Studies reporting prevalence of HBsAg were selected. Data from high risk groups were excluded. MAIN RESULTS: 54 papers reporting data on 61 populations were identified. The true prevalence for each study was calculated from the reported prevalence using the specificity and sensitivity of the test employed. The true prevalence in non-tribal populations is 2.4% (95% CI: 2.2%-2.7%). True prevalence among tribal populations is 15.9% (CI: 11.4%-20.4%). CONCLUSION: These figures may be useful in estimation of the burden of the disease in the country and for projecting the cost-benefits of immunization.
