In this article, we demonstrate that prevention of CRC, in particular prevention of distal CRC by a single FS offered at age 55–64 years is substantially stronger than suggested by published results of RCTs which included in their analyses CRC cases that were already prevalent at recruitment and therefore not preventable anymore. Our analyses suggest that people undergoing FS at around 60 years of age can reduce their risk to develop incident distal CRC within the following 15 years by approximately two thirds rather than approximately half as suggested by published results, which included prevalent cases at recruitment in their outcome measures. Whereas inclusion of prevalent CRCs at recruitment attenuated effect estimates of endoscopic screening on CRC incidence in conventional analysis, the earlier detection of such prevalent cases should rather be considered as an additional major asset of screening on top of a stronger than previously assumed preventive effect on CRC incidence [15].
Our analyses assumed equal prevalence at recruitment of distal CRC among participants in the intervention groups and the control groups and among screened and unscreened participants in the intervention groups. The former assumption is plausible due to the randomized study designs and the large sample size. The latter assumption is also plausible for the two trials included in our analysis, as published cumulative incidence curves were almost identical for unscreened participants in the intervention group and participants in the control group, indicating that use of the FS offer was not related to CRC risk in these two trials.
Nevertheless, it could be theoretically possible that not all distal CRCs that were prevalent at the time of recruitment became clinically manifest and diagnosed during the follow up in the unscreened participants in the intervention group and the control group. Even though their proportion would be expected to be very small, given the long follow-up period (> 15 years) and a mean sojourn time of preclinical CRC in the order of 3 to 6 years [16,17,18], the numbers and proportion of prevalent cancers at recruitment among all reported distal CRC could have been slightly smaller, and the numbers of truly incident cancers could have been slightly higher in these subgroups than assumed in our analyses. This would imply that underestimation of reduction of truly incident cases among screened participants (in whom prevalent cases at recruitment were disclosed by FS) may have been even stronger than suggested by our analysis.
In theory, screening might also also have led to some overdiagnoses of cases that would otherwise never have become diagnosed at lifetime. However, such overdiagnoses are expected to be rare for the age groups included in the trials [19]. Of greater concern may be imperfect sensitivity of FS to detect precursors of CRC and lack of re-screening which may account for the majority of the remaining truly incident cases.
For the effect estimates on total CRC incidence the additional assumption was made that the ratio of total and distal CRC prevalence at recruitment was the same as the ratio of the observed incidence total and distal CRC incidence in the absence of screening which implies that mean sojourn time in preclinical state would be the same for distal and proximal CRC. To account for potential variation in mean sojourn time according to cancer site, we conducted sensitivity analyses assuming 10% higher or lower total numbers of prevalent cancers at recruitment which yielded very similar results as the base case analyses (see Supplementary Table 3). Given that the majority of CRCs are located in the distal colon and rectum, such 10% differences in the overall number of prevalent cases at recruitment would reflect very large differences in the mean sojourn time of proximal and distal cancers. Even major variations in such sojourn times which appears unlikely would therefore have only a rather minor impact on our results.
We focused our analyses on RCT estimates of the effectiveness of FS. As previously demonstrated [12, 20,21,22], the concerns regarding underestimation of screening effects on CRC incidence by inclusion of non-preventable prevalent cancers at recruitment in the outcome measure similarly apply to the NordICC study, the so far only RCT reporting on long-term outcomes of screening colonoscopy [6]. In that trial, however, major differences in reported cumulative incidence of CRC between unscreened participants in the intervention group and participants in the control group suggest major selection effects in use of screening colonoscopy, making derivation of „prevalence-corrected“ effect estimates somewhat more complex. However, a recent modelling study suggested „prevalence-corrected“ effect estimates of screening colonoscopy on total CRC incidence to be very similar to the ones for distal CRC incidence derived for FS in this article [20].
Among the four large FS trials [2,3,4,5, 7,8,9,10,11], we chose the UKFSST [2, 7] and the Italian SCORE trial [3, 10] for demonstrating our point as these trials focused on the effects of a single screening FS at age 55–64 and reported the necessary data from both intention-to-screen and per-protocol analysis in detail, including cumulative incidence curves showing virtually identical cumulative incidence of unscreened participants in the intervention group and participants in the control group. The latter enabled straightforward derivation of “prevalence-bias corrected” effect estimates under plausible assumptions. Overall results of these two trials are largely comparable to results of the other two large FS trials from Norway and the US [4, 5, 8, 9], suggesting that the order of magnitude of incidence reduction by FS screening may be generalizable to other countries.
Our analyses focused on CRC incidence, one of the two primary outcomes of the FS screening trials for which underestimation of effects by inclusion of prevalent cases is relevant. We did not address effects on CRC mortality, the other primary outcome of the trials. In contrast to CRC incidence, there is no concern about including cases that were already prevalent at the time of recruitment in analyses for the mortality outcomes. In the contrary, screening by FS conveys its preventive effects on CRC mortality through both earlier detection of prevalent CRC cases and prevention of truly incident CRC cases [23, 24]. Interestingly, the “prevalence bias corrected estimates” of CRC incidence reduction derived in our analysis are very similar to the estimates of CRC mortality reduction reported from both trials. For example, for the > 17 year follow-up of the UKFSST, our “prevalence bias corrected estimates” of incidence reduction (intention-to-screen analysis: 30% for any CRC, 50% for distal CRC; per-protocol analysis: 40% for any CRC, 66% for distal CRC) almost perfectly match the estimates of CRC mortality reduction reported from that study (intention-to-screen analysis: 30% for any CRC, 46% for distal CRC; per-protocol analysis: 41% for any CRC, 66% for distal CRC).
Our findings of much stronger preventive effects of screening endoscopies with detection and removal of CRC precursors than those suggested by the published RCT results are in line with observations of strong decreases in CRC incidence in countries with widespread offers and use of screening colonoscopy, such as the US and Germany [25, 26]. For example, CRC incidence has almost halved in the last three decades in the US, where use of screening colonoscopy has become very common, with meanwhile more than 60% of people above 50 years of age having had a colonoscopy in the past 10 years [27]. This strong decline in incidence was achieved despite unfavourable trends in CRC risk factors, such as the increase in prevalence of obesity [28]. Furthermore, the decline was selectively seen in age groups with widespread use of screening colonoscopy, whereas CRC incidence was rising in younger age groups [26]. Although screening FS is expected to be slightly less effective than screening colonoscopy with respect to reduction of total CRC risk among those who use the screening offer, the necessary capacities and resources and high adherence rates may be easier to achieve for screening FS. Furthermore, combination of FS with screening by fecal immunochemical test could ensure early detection (or even prevention) of the vast majority of proximal cancers, in addition to the strong prevention of distal cancers [29, 30]. Along with possible extension of FS screening intervals from previously recommended 5 years to 10 years [31], this could make screnning FS a particularly effective and cost-effective CRC screening strategy that would be feasible even in countries lacking screening colonoscopy resources.
In summary, our analysis provides evidence for substantially stronger preventive effects of CRC screening by FS than those suggested by the published RCT results. It is plausible to assume that preventive effects of other CRC screening approaches have likewise been substantially underestimated by the type of prevalence bias addressed in our analysis. The substantially stronger than previously assumed preventive effects of CRC screening may have important implications on key parameters of CRC programs, such as cost-effectiveness and benefit-harm ratio, and may help to better define target populations, age range and intervals of recommended screening offers which should be carefully further evaluated by pertinent modeling studies. Most importantly, however, our results should encourage more widespread roll-out of CRC programs and use of CRC screening offers, which are among the most effective approaches in cancer prevention available to date [32].
留言 (0)