Data synthesis and statistical analysis
The comparison tests were categorized into two groups: 1) Heated
(vegetable oil) and 2) control group. Hence, there are six separate
comparison tests for each respective oil to determine its summary effect
estimate. Review Manager 5.4.1 was used to perform the statistical
analyses with the extracted mean ± standard deviation and sample size as
data inputs. A random-effect model was used as the studies were
heterogeneous from each other in the scope of intervention, population
and outcome, hence making it impossible to impose a fixed-effect model
that pools all of the study findings to arrive at a summary estimate.
The selected studies were very different by the results of adverse
health effects caused by consuming thermally oxidized vegetable oil but
other areas such as the type of oil used in interventions are
sufficiently similar nonetheless. Weighted mean differences, the
statistical significance of each included paper, and heterogeneity
between studies were highlighted. Chi-square test was used to determine
significant heterogeneity (P < 0.1). In addition to that,
I2 statistic was used to quantify heterogeneity
through the percentage of total variation across studies attributable to
heterogeneity rather than chance. Cut off points of percentage are such:
>75% (high heterogeneity), 50 – 75% (moderate
heterogeneity), and <50% (low heterogeneity).