This evaluation of the population-based screening program in Stockholm-Gotland, Sweden, demonstrated a similar overall CRC incidence among those invited to the screening as compared to the non-invited, and a decreased incidence after screening cessation. This change in post screening incidence was significant for early-staged CRC and CRCs in women and differed by colorectal localization. However, we did not find a significant change in incidence in subgroups during the gFOBT program, nor a decrease in late-stage CRC from the screening program.
Several previous studies have shown an increased CRC incidence when introducing population-based FIT screening, due to both incidence and prevalence screening in the first round, and thereafter a return to or decrease below pre-screening levels in subsequent rounds (12, 22, 23). At long-term follow-up of the randomized gFOBT screening trial in Minnesota there was a marked decline in incidence with multiple screening rounds as compared to controls (24).
In the present study, we did not find a significant change in the incidence among those invited to the gFOBT program. However, the invited age cohorts consisted of both prevalent and incident screening rounds for each attained age (except the 60-years-olds), hence the net effect could be an unchanged incidence. Apart from this, approximately 40% of the gFOBT invitees did not participate in screening and could therefore not contribute to an increased CRC incidence (25). Furthermore, gFOBT, as compared to FIT, has a lower sensitivity for advanced adenoma; 10–15% vs 25–30%, hence it is not expected a large decline in CRC incidence post screening because of polypectomy in a gFOBT screened population (26–29). In the Minnesota trial the cumulative gFOBT positivity was approximately 30%, so the large decrease in incidence could partly be explained by the large number of colonoscopies (and polypectomies) performed in the intervention group. The positivity rate in the Stockholm-Gotland gFOBT program was approximately 2% (16).
The gFOBT sensitivity for CRC is stage-dependent, with reported sensitivity rates of Dukes A 89%, Dukes B 79%, Dukes C 72%, and Dukes D 48% (15). A meta-analysis of FIT studies (cut-off-levels between 10–20 µg/g) estimated FIT sensitivity to 73% for stage I CRC, 80% for stage II, 82% for stage III, 79% for stage IV (30). The overall CRC sensitivity in the Stockholm-Gotland FIT program with cut-off levels 40 µg/g and 80 µg/g in women and men was estimated to 65%, as compared to 40% in the gFOBT program (16, 31). The shift to FIT in the program generated an increased participation rate (69%), and a higher sensitivity for advanced adenomas and CRC, which contributed to the net increase in CRC incidence in subgroups of early-staged and proximal CRCs during screening and a decrease post screening due to the preventive effect of polypectomy. The decreased incidence post screening after an increase during screening as compared to the non-invited was only seen in stage I CRCs, probably due to the earlier detection of CRC with screening leading to a compensatory drop after screening cessation.
The 14% reduction in CRC mortality recently reported from the Stockholm-Gotland screening program would most likely be explained by a shift from late to early-stage CRC (17). However, in the present study we did not a see a decreased incidence of stage III-IV CRC among the invited. Since the mortality is low in stage I CRC, the reduced stage I incidence post screening could not explain the reduced mortality (32). However, the invited included both the non-participants as well as prevalence screening rounds which could balance out a favorable stage shift in the participants. In the Dutch screening program, the incidence of late-stage CRC increased at screening implementation followed by a decrease when compared to the incidence rates prior to screening (8). Moreover, given the low number of CRCs stage III and IV in the present study they were combined and categorized into late-stage (III-IV). There might be a stage shift among the invited from IV to III or in subclassification of T stages that we were not able to detect albeit important for the prognosis (33). In the previously cited Danish study, there was an increased incidence of stage I-III CRC and no significant difference in stage IV CRC in the invited as compared to non-invited, but this study included only the prevalence round (12).
On the other hand, it is not certain that a favorable stage shift due to screening directly translates into a disease specific mortality reduction, because this is dependent on the difference in survival between the early and late stages and the proportion of cancers that the screening is able to shift to earlier stages, and it also assumes that the late stage cancers among the invited have the same survival as the late staged cancers among the non-invited (34). Invitation to screening could raise the awareness of the disease and make the invited individuals more prone to seek health care and to life style changes, e.g., regarding smoking, than the non-invited, potentially affecting the disease mortality other than as a shift in stage (35).
During the FIT screening, the incidence of proximal CRC increased. Proximal CRCs are more common in women, and the Stockholm-Gotland program applies lower cut-off-levels in women than in men, which might explain the increased incidence with FIT screening and the decreased incidence in women post screening, although gFOBT and FIT performs worse in proximal CRCs as compared to distal in gender-uniform screening (14, 31). The increased incidence of early-stage proximal CRCs in women could have contributed to the decreased mortality of the program, since proximal CRCs confers a worse prognosis and are diagnosed clinically at a late stage (36). The overall incidence of proximal CRC was not decreased post screening. However, only birth cohorts 1948-51 (and part of 1947 cohort) were screened with FIT and reached post screening age at end of follow-up, hence gFOBT screening was over-represented in this group. Post screening, a significant decrease was seen only for distal localization reflecting the higher sensitivity for distal CRC of both gFOBT and FIT and the higher rate of distal CRC in men (14).
The strengths of this study were the evaluation of a large population-based CRC screening program shifting from gFOBT to FIT screening, the linkage with individual data to validated cancer registers with low number of missing data enabling analyzation of CRC subgroups, and the assessment of CRC incidence of invited and non-invited individuals within the same time-period and in the same region. Moreover, evaluation of the invitation to screening precluded the self-selection bias from healthy participants.
Nevertheless, this study has several limitations. The screening program implementation was made by randomization of birth cohorts into early, late or no screening invitation rather than individual randomization, making the comparisons between invited and non-invited biased by age. This problem was overcome when comparing age-adjusted incidence ratios, but we were unable to assess the incidence change over time when initiating screening and by subsequent screening rounds. Furthermore, we did not analyze stage III and IV CRCs separately due to a low number of cases, hence not capture any change in incidence in between the late-stage CRCs as discussed above. Moreover, the post screening comparisons of incidence were, due to the recent shift to FIT screening, dominated by gFOBT screened individuals, and additional studies are needed to fully address the effect of FIT in 70-75-year-olds, especially since the shift to FIT increased the participation rate by 12% (37).
In conclusion, the shift to FIT in the population-based screening program of Stockholm-Gotland, Sweden, significantly increased the CRC incidence for early-staged and proximal CRCs, and the overall decrease post screening was mainly seen in distal, early staged CRCs in women as compared to the non-invited. The full effect of gender-based FIT screening on the incidence post screening needs further evaluation.