Screening

One of the topics we've covered is screening. I found this really interesting because it conflicted with my preconceived notions about screening. I originally thought knowledge and testing is always good. If we have a screen for cancer, we should use it; after all, cancer is a fatal disease, and catching it earlier should translate to better outcomes. Isn't it better to know than not to know?

Reading the syllabus sections have convinced me otherwise. There are too many factors in determining the efficacy and utility of a screening test. The best way to know if a screen works is to run a randomized control trial with mortality or morbidity as the outcome. The problem is screening has inherent biases; if you catch a disease earlier, it will appear that the person survived longer regardless of whether catching it earlier had any effect on overall outcome (if there aren't good treatments, then screening has few advantages). Screening might catch things that would not otherwise have progressed to disease (mammograms have found lots of ductal carcinoma in situ but it is unclear how many of those would have lead to invasive carcinoma).

Indeed, this is compounded by the fact that treatment for cancer is associated with morbidity and mortality. Chemotherapy is hardly benign, and a physician must really think about the oath to do no harm. The efficacy ("positive predictive value") of a screening test depends not only on its sensitivity and specificity, but also on the population prevalence of that disease. Even if a screen is extraordinarily specific for a disease, if the prevalence of that disease is very low (ie. cancer), then there will be more false positives than true positives. The utility of a screen depends also on the follow-up procedure. For example, there's a fairly good screen for ovarian cancer (transvaginal ultrasound), but the confirmatory test is surgery, and 13 oophorectomies would need to be done to find one case of ovarian cancer (compare this to 838 mammograms needed to prevent a death due to breast cancer in women 50-74). Lastly, screens must be acceptable to patients; people getting screened are healthy and the threshold for tolerating procedures varies a lot (no one wants a colonoscopy).

I was pretty surprised about all of this; it really did change my view on screening tests. Furthermore, cost-effectiveness comes into play since screens are for large populations; we need to prove that the risks we are taking (money, working up false positives, diversion of resources) are really worth the benefit. In the end, I think only cervical cancer (pap smear), breast cancer (mammogram), and colon cancer (colonoscopy or others) really meet the criteria for tests with a net benefit.