Summary results and analysis for odds ratio

By Yashika Kapoor & Priya Chetty on December 25, 2017

Results can be obtained by clicking on “Run Analysis”. The image below shows the result for meta-analysis of the odds ratio data, for both random and fixed effects model. As seen the relative weighting of studies in the two models are quite different. The random effects model takes into account both the interstudy variance and sample size, resulting in well-distributed weight. Also, as the studies do not follow identical research design, hence random effect model selection ensues.

Basic summary statistic results
Image 1: Basic summary statistic results

Interpretation of summary statistics

The summary effect size for the random model is 5.376 (95% CI = 2.28-12.65) with p-value = 0.00 (<0.05), thus leading to acceptance of the alternative hypothesis. This indicates the significant association between prevalence of gallstones and gallbladder carcinoma, with odds of an individual developing gallbladder cancer, are 5 fold in case of gallstones.

The I2 statistic for heterogeneity, as shown below, is 96.08 (96.08%), p = 0.00, resulting in acceptance of the alternative hypothesis. Thereby, indicating the presence of ‘significant’ heterogeneity within the studies in taken for present review.

Model statistics
Image2: Model statistics

The forest plot, shown below, has the line of no effect marked as 1, as the outcome variables are binary such as odds ratio.

Forest plot for odds ratio data
Image3: Forest plot for odds ratio data

Furthermore, customization of the forest plot by using the options is shown below.

Assigning header to the forest plot
Image 4: Assigning header to the forest plot

Besides assigning a title to the forest plot, changing the study symbols, customize space in forest plot and changing line thickness for anchors and confidence intervals is also possible.

Other options for customization
Image 5: Other options for customization

Assesment of publication bias

The assessment of funnel plot as shown below, indicates the asymmetrical distribution of effect estimates, away from the central line. This suggests a lack of inclusion of relevant trials.

Funnel plot for odds ratio data
Image 6: Funnel plot for odds ratio data

The formal test statistics for intercept and rank correlation test sufficiently indicate the statistical basis for bias. The tests statistics values are as shown below. The results for rank correlation test show values of p>0.05 thereby indicating the acceptance of null hypothesis and the presence of bias.

Rank correlation test statistics
Image 7: Rank correlation test statistics

The regression test results indicate significant deviation from zero and pronounced the asymmetry. Also, the p-value > 0.05, which indicates the presence of significant asymmetry and bias.

Intercept test statistics
Image 8: Intercept test statistics

Cancer risk associated with cholecystitis

The article series beginning from Performing manual data entry in CMA spreadsheet up till now elaborates the following:

  • Data entry – Manual and Import procedures.
  • Performing meta-analysis for Events rate and Odds ratio data.
  • Visualization, customization, and interpretation of results.

The illustration of different procedures via different data type and studies highlights the significance of study selection. The investigation into the association of gallstones with gallbladder carcinoma via the different set of studies reports different results. On one hand, the event rate data did not yield significant association, whereas odds ratio analysis result was opposite.

The plausible explanation for the same is that the studies in event rates data were highly heterogeneous. The high degree of heterogeneity and interstudy variance produced insignificant summary effect estimate. Thus, the selection of studies for systematic review is to be done carefully to obtain significant results.

Discuss

1 thought on “Summary results and analysis for odds ratio”