The main issue I see for iNZight is the estimation of the prevalence of exposure. For a cohort study that is representative of the underlying population, the prevalence of exposure can be estimated from the total study population. For a case-control study, that can be thought of as a heavily stratified sample of the population (by case-status) the prevalence will be a weighted average of the cases and controls, but if a rare disease, much closer to the prevalence of exposure in the controls.
If the prevalence information were taken either from the total population or controls in a case-control study, then this would enable the estimation to be carried out using only the information in a 2x2 contingency table.
Variance estimation for PAR is not so important (at least for me), but here's a discussion, FYI...
Here's a couple of sources on PAR.
The detail - for statisticians… https://journals.sagepub.com/doi/pdf/10.1177/096228020101000302
The basics - for epidemiologists… https://bjsm.bmj.com/content/bjsports/52/4/212.full.pdf
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3734469/pdf/ijerph-10-02932.pdf
The main issue I see for iNZight is the estimation of the prevalence of exposure. For a cohort study that is representative of the underlying population, the prevalence of exposure can be estimated from the total study population. For a case-control study, that can be thought of as a heavily stratified sample of the population (by case-status) the prevalence will be a weighted average of the cases and controls, but if a rare disease, much closer to the prevalence of exposure in the controls.
If the prevalence information were taken either from the total population or controls in a case-control study, then this would enable the estimation to be carried out using only the information in a 2x2 contingency table.
Variance estimation for PAR is not so important (at least for me), but here's a discussion, FYI...
https://www.mayo.edu/research/documents/tech-report-54/doc-20429047