Opinion | Screening Colonoscopy: Where Do We Go From Here?

Screening for colorectal cancer (CRC) is both effective and cost effective. One of the earliest tests for colorectal cancer screening is detecting occult blood in stool followed by colonoscopy if the test is positive. A randomized clinical trial of annual fecal occult blood testing in the U.S. demonstrated an 18% reduction in colorectal incidence over 18 years of follow-up and 33% reduction in colorectal cancer mortality sustained through 30 years of follow-up. Screening with colonoscopy gained popularity in the early 2000s, and observational studies have reported a 60-85% reduction in colorectal cancer incidence.

In the U.S., colonoscopy is now the most common screening modality for colorectal cancer. It has the advantage of detecting not only early cancers but also precursor lesions such as polyps, some of which, if left in place, could turn into cancer. Hence, colonoscopy is an early cancer detection and cancer prevention tool. Despite the widespread enthusiasm for colonoscopy screening, randomized clinical trial data for its benefit were lacking — until now.

The results of a recent colonoscopy trial have attracted considerable attention given results that, on the surface, suggest colonoscopy is less effective than previously thought. However, the study has several limitations, and will do more harm than good if it promotes a chilling effect in screening among patients.

Interpreting the Latest Colonoscopy Study

The NordICC trial was a pragmatic randomized trial to test the effectiveness of colonoscopy. Between 2009-2014, the study randomized 84,585 individuals in a 1:2 ratio of colonoscopy to usual care (no screening). The study reported an 18% reduction in CRC incidence and no reduction in CRC mortality with colonoscopy screening, results that have surprised and dismayed the readers.

There are three important considerations in interpreting the results: First, the study was an intention to screen trial, which means individuals randomized to the colonoscopy arm were invited to undergo screening. This is how randomized trials are and should be done, to preserve randomization and balance known and unknown confounders. However, the uptake of colonoscopy was very low, at only 42%. Compare this to the Minnesota fecal occult blood screening randomized trial, where uptake of fecal occult blood testing and subsequent colonoscopy was over 84%. Since those in the arm randomized to colonoscopy are counted in the denominator regardless of completion, one can see how the comparison can show a diluted effect with low uptake of colonoscopy. This is when we look at the per protocol, or the compliance adjusted estimates. In this trial, those that adhered had a 40% reduction in CRC incidence. Still not as high as expected, but better than 18%.

Second, after decades of research on the topic, we know that colonoscopy is highly operator dependent, and that only high-quality colonoscopies are able to detect and remove the early cancers and premalignant polyps that confer protection from subsequent colorectal risk. As a result, in the U.S. we have developed a set of quality indicators that have been validated against risk of developing colorectal cancer in the near future. The best studied ones are cecal intubation rate and adenoma detection rate (ADR). The bar we have set is 95% and 25%, respectively, for these two indicators.

In the U.S. we perform a lot of colonoscopies (more than 15 million per year), and we do them well. U.S. endoscopists, who perform this procedure, meet and exceed these benchmarks for the most part. However, the endoscopists in the NordICC study fell short of these benchmarks. Nearly 30% had ADR below 25% and cecal intubation rates less than 95%. The study reports on only 35 endoscopists that performed 30 or more colonoscopies during the study period, and it is not known how many more endoscopists participated who may be low volume and have unknown quality indicators. This does call into question the generalizability of the study to the U.S. in particular.

Third, the study is underpowered to detect a difference in CRC mortality between the two arms. Colonoscopy detects premalignant polyps, which otherwise would take 8 to 20 years to become cancer, and another 5 to 10 years to advance, spread, and become fatal. Hence the benefit in reducing death from such lesions is expected to be realized over a long period of time. With enrollment ending in 2014, and follow-up in 2020 (national death registries generally lag 2 years behind), the follow-up time period is insufficient to expect a difference in risk of death. The authors are continuing to follow these individuals, and we are likely to see more information on CRC mortality in 5 and 10 years. Therefore, the conclusion of no benefit in CRC mortality is premature.

The last, and perhaps most important, aspect of the study is the questionable generalizability to the whole U.S. population. The way the study was designed, individuals were not consented (or told) they were being randomized to colonoscopy or usual care. Hence the authors have no information on the risk factors of these individuals, such as their body mass index, baseline risk of colorectal cancer, smoking history, and dietary and other lifestyle factors that we know are associated with risk of colorectal cancer, and attenuated in many subgroups of the U.S. population. Hence the benefit may not reflect the effect in an at-risk population in the U.S. We know the endoscopists and the colonoscopy practice are certainly not generalizable. Taken together, it’s hard to see how this may apply to our practice.

Placing the Study in a Broader Context

Randomized clinical trials are the holy grail of scientific evidence. Estimates from observational studies are generally inflated, and time and time again we have seen these estimates become smaller and more conservative when the rigors of a randomized trial with concealed allocation are applied. In the context of colonoscopy, we expected the estimates to shrink somewhat, and the 95% confidence estimates to get tighter compared to results from observational studies — but this study may have missed the mark due to its limitations.

What we want is the study to be taken in the context of the body of evidence on effectiveness of colonoscopy, and for people to understand that it is one more data point, with more to come from ongoing trials. We also need the public to understand that screening for colorectal cancer reduces the risk of developing and dying from CRC, and that colonoscopy is a highly effective modality. Clinicians must ensure their patients understand this, and continue to encourage screening. Let’s not throw the baby out with the bathwater just yet.

Aasma Shaukat, MD, MPH, is the Robert M. and Mary H. Glickman Professor of Medicine and Gastroenterology and the director of outcomes research in the Division of Gastroenterology & Hepatology at NYU Langone in New York City. She is a practicing gastroenterologist and researcher in colon cancer screening and quality indicators of colonoscopy.