A quick detour to a medical RCT, which affects me having newly turned 60. NEJM just published the 1st large RCT of colonoscopy screening to prevent colorectal cancer. Quick take: High-quality RCT shows screening benefit, but much smaller than suggested by prior non-RCTs.

Program & Study Design:

• The study randomly assigned 85,000 adults ages 55-64 in Poland, Norway, and Sweden to a treatment (T) group invited to colonoscopy screening, vs a control (C) group not invited. The RCT was very well-conducted (e.g., large sample, baseline balance between treatment and control, negligible attrition).

Findings:

• The RCT found that the treatment group had 18% lower risk of colorectal cancer (CRC) than the control group over the next 10 years (which was statistically significant), and 10% lower risk of CRC mortality (which wasn't statistically significant). A NEJM editorial described these smallish effects as "surprising and disappointing."

• However, only 42% of treatment group members actually received a colonoscopy. The authors report a "per protocol" analysis to estimate the treatment effect on those who got a colonoscopy, but I agree with Dr. Vinay Prasad that it's not a reliable analysis.

• Fortunately, the study appendix reports a valid (instrumental variables) analysis of the effect on T members who actually got a colonoscopy: It reduced their risk of CRC by 34% and of CRC mortality by 28%. The effect on CRC was statistically significant; the effect on mortality wasn't (so is only suggestive).

Comment:

• These benefits are much smaller than those found in less-rigorous, non-randomized studies. A meta-analysis of such studies had suggested screening reduced CRC risk by 89% and CRC mortality by 64%. This underscores the importance of RCT evidence.

• Bottom line for me personally: I will get a colonoscopy based on the finding that it reduces CRC risk by 34% and that it might reduce CRC mortality by 28% (the latter finding being only suggestive).