Early detection versus primary prevention in the PLCO flexible sigmoidoscopy screening trial: Which has the greatest impact on mortality?

Abstract

BACKGROUND Screening for colorectal cancer (CRC) with flexible sigmoidoscopy (FSG) has been shown to reduce CRC mortality. This study examined whether the observed mortality reduction was primarily due to prevention of incident CRC via removal of adenomatous polyps or to early detection of cancer and improved survival. METHODS The Prostate, Lung, Colorectal, and Ovarian (PLCO) cancer screening trial randomized 154,900 men and women age 55–74. Individuals received FSG screening at baseline and 3 or 5 years versus usual-care. CRC-specific survival was analyzed using Kaplan-Meier curves and proportional hazards modeling. We estimated the proportion of CRC deaths averted by early detection versus prevention using a model that applied intervention arm survival rates to usual-care arm CRC cases and vice-versa. RESULTS 1008 intervention and 1291 usual-care arm CRC cases were observed. Through 13 years of follow-up, there was no significant difference between arms in CRC-specific survival for all CRC (68% intervention vs. 65% usual-care; p= 0·16) or proximal CRC (68% vs. 62%; p=0·11); however, distal CRC survival was higher in the intervention arm (77% vs. 66%; p<0·0001). Within each arm, symptom-detected cases had significantly worse survival than screen-detected cases. Overall, 29–35% of averted CRC deaths were estimated to be due to early detection and 65–71% to prevention. CONCLUSIONS CRC-specific survival was similar across arms in PLCO, suggesting a limited role of early detection in preventing CRC deaths. Modeling suggests two-thirds of avoided deaths were due to prevention. Future CRC screening guidelines should emphasize prevention via identification and removal of precursor lesions.

Early detection versus prevention in colorectal cancer screening: Methods estimates and public health implications

“…The ADP ranged from 76% to 84% in the intention-to-treat analysis, and from 75% to 92% in the per-protocol analysis. These estimates are consistent within the trials and with the results presented by Doroudi et al 6 Low rates of screen-detected cases were reported in the PLCO trial (31%) 6 and in the 2 RCTs (22% in the SCORE trial, 3 20% in the UK trial after 11 years of follow-up, 1 and 11% in the UK trial after 17 years of followup 2 ), suggesting limited early detection.…”

“…After 13 years of follow-up in both the intervention and usual-care arms, 1008 CRC cases in the intervention arm and 1291 CRC cases in the usual-care arm were observed.The 13-year survival rates in the intervention and usual-care arms were significantly different for cases of distal CRC (77% in the intervention arm vs 66% in the usual-care arm; P<.0001), whereas they were not for all CRC (68% vs 66%, respectively; P 5 .16) and for proximal CRC cases. Anatomic location and stage distribution by trial arm and CRCspecific survival by reason of detection also are reported by Doroudi et al 6 Moreover, whether mortality reduction in the intervention arm was due primarily to the early detection of CRC or to prevention via removal of adenomatous polyps was investigated. Overall, 29% to 35% of averted CRC deaths were estimated to be due to early detection and 65% to 71% were estimated to be due to prevention.We believe the idea in the study by Doroudi et al 6 to assess the percentage of mortality reduction due to prevention via the removal of adenomatous polyps or to early detection is brilliant.…”

“…As stated by the authors, "This assessment is of importance because it will enhance the scientific understanding of how FS screening prevents CRC deaths and may further inform future CRC screening recommendations." 6 Nevertheless, the authors have adopted an analytic approach in which the use of comparison of survivals in the intervention and usual-care arms is flawed by lead time, which inflates the estimate of survival in the intervention arm. Moreover, when comparing survivals in the 2 arms of the PLCO trial (in which intention-to-treat results are reported), other main factors can improve the point estimate of survival in the intervention arm, namely: 1) better access to treatment that would increase survival; it most likely is not the case because the intervention and usual-care arms of the PLCO trial were sampled within the same health maintenance organizations; 2) overdiagnosis, which is intrinsic to cancer screening; and 3) the sensitivity of the test and the compliance with screening, affecting the percentage of screen-detected cancers.The use of survival in the intervention arm should be avoided.…”

“…The ratio between the ADP and the overall averted CRC deaths in the intervention arm represents the percentage of ADP via the removal of adenomatous polyps (algorithm is shown in the footnotes in Table 1). [1][2][3]6 The use of FR in the control arm entails that the access to cancer care in the 2 arms is not differential and no further screening occurs in the intervention arm.…”

“…Following this approach (Table 1) in the PLCO trial, 6 the ADP in the intervention arm is 79%, and 21% are due to early diagnosis. These results are not of a different order of magnitude with respect to the number of CRC ADPs, as estimated in Figure 3A (blue box, 74) and with the number CRC deaths averted due to early detection in Figure 3B (red box, 24), as shown by Doroudi et al 6 We applied the same approach (Table 1) to the results of 2 RCTs. [1][2][3] We did not include the Norwegian Colorectal Cancer Prevention (NORCCAP) trial 5 for the sake of simplicity, given that the randomization ratio varies across age groups and study arms.…”