I want to analyse a cohort study where individuals where followed for a number of years and to examine factors that influence occurence of disease (by a Cox model or logistic regression). One factor is family history (occurence of disease in patients, assessed at baseline). What adjustment is needed to allow for the fact that some parents may not yet be diagnosed - if its a disease occuring a lot in middle or old age (eg DM, CVD, cancer) - this is an issue I'd have thought when adjusting for age as younger individuals would likely have younger parents who although may be free od disease now may well be diagnosed in years to come.
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
Presumably all your index individuals are old enough to start seeing the disease in that group, and if so, it seems unlikely that you have parents of these individuals who are too young to have developed the disease. Although, it seems reasonable to be concerned that the parents may still develop the disease after baseline since there is certainly some a distribution of age at onset.
There are three potential options I can see (not mutually exclusive):
Adjust for parent's age, not just index individual's age, at baseline to deal with your concern about some parents being too young.
If you have information on the parents over follow-up, adjust for parent's disease status at over time as a time-varying covariate. Here, the tricky issue is whether you believe that the index patient's disease status could influence the parent's disease status -- if so, you will need to use methods like inverse probability weighting to adjust for potential treatment-confounder feedback.
Imagine you were going to do a randomized trial of the exposures you are interested in. In that imaginary trial, how would you define your eligibility criteria? Would you set an age limit on the index individuals and/ or their parents? Would you require any eligibility criteria related to the parent's disease status? If so, apply those same criteria to your cohort study through restriction.
Also, you may need to think carefully about why you are interested in family history, especially if you use option 2 above -- do you consider it a marker of genes or environment or both? Imagine a situation where an index individual develops the disease at say 40 and their parent develops it one year earlier at say 62. Would that be an indication of the same mechanistic role of family history as someone who develops the disease at age 63 when their parent had developed the disease 20 years earlier at age 65?
