Prediction of the severity of colorectal lesion by fecal hemoglobin concentration observed during previous test in the French screening program

Development and validation of colorectal cancer risk prediction tools: A comparison of models

BACKGROUND

Identification of individuals at elevated risk can improve cancer screening programmes by permitting risk-adjusted screening intensities. Previous work introduced a prognostic model using sex, age and two preceding faecal haemoglobin concentrations to predict the risk of colorectal cancer (CRC) in the next screening round. Using data of 3 screening rounds, this model attained an area under the receiver-operating-characteristic curve (AUC) of 0.78 for predicting advanced neoplasia (AN). We validated this existing logistic regression (LR) model and attempted to improve it by applying a more flexible machine-learning approach.

METHODS

We trained an existing LR and a newly developed random forest (RF) model using updated data from 219,257 third-round participants of the Dutch CRC screening programme until 2018. For both models, we performed two separate out-of-sample validations using 1,137,599 third-round participants after 2018 and 192,793 fourth-round participants from 2020 onwards. We evaluated the AUC and relative risks of the predicted high-risk groups for the outcomes AN and CRC.

RESULTS

For third-round participants after 2018, the AUC for predicting AN was 0.77 (95% CI: 0.76-0.77) using LR and 0.77 (95% CI: 0.77-0.77) using RF. For fourth-round participants, the AUCs were 0.73 (95% CI: 0.72-0.74) and 0.73 (95% CI: 0.72-0.74) for the LR and RF models, respectively. For both models, the 5% with the highest predicted risk had a 7-fold risk of AN compared to average, whereas the lowest 80% had a risk below the population average for third-round participants.

CONCLUSION

The LR is a valid risk prediction method in stool-based screening programmes. Although predictive performance declined marginally, the LR model still effectively predicted risk in subsequent screening rounds. An RF did not improve CRC risk prediction compared to an LR, probably due to the limited number of available explanatory variables. The LR remains the preferred prediction tool because of its interpretability.

[Évaluation de la gravité du cancer colorectal dépisté dans le contexte de la crise sanitaire liée au COVID19 en région Ile-de-France]

BACKGROUND

After the announcement in March 2020 of the COVID-19 pandemic, colorectal cancer (CRC) screening programs were suspended in several countries. Compared to the lesions detected during previous campaigns, this study aims to assess the severity of CRC detected during the 2020 screening campaign in Île-de-France, the French region most affected by the 1st wave of the pandemic.

METHODS

The descriptive and etiological study included all faecal immunochemical test (FIT) results carried out between January 2017 and December 2020 on people aged 50-74, living in Île-de-France. First, the proportion of colonoscopies performed within one month (One-month-colo) following FIT; the yield of colonoscopy (proportion of colonoscopies with a neoplasm lesion among those performed) and CRC severity (TNM Classification, Level-0: T0/N0/M0, Level-1: T1/T2/N0/M0, Level-2: T3/T4/N0/M0; Level-3: T3/T4/N1/M0; Level-4: M1) were described in 2020 compared to previous campaigns (2017, 2018, and 2019). Subsequently, the link between the level of CRC severity and the predictive factors, including campaign year and time to colonoscopy, was analysed using polytomous multivariate regression.

RESULTS

The one-month-colo (2017: 9.1% of 11,529 colonoscopies; 2018: 8.5% of 13,346; 2019: 5.7% of 7,881; 2020: 6.7% of 11,040; p < 0.001), the yield (65.2%, 64.1%, 62.4%, 60.8% respectively, p < 0.001) were significantly different between campaigns. The proportion of CRC level-4 (4.8% in 2017 (653 CRC); 7.6% in 2018 (674 CRC); 4.6% in 2019 (330 CRC) and 4.7% in 2020 (404 CRC); p < 0.29) was not significantly different between campaigns. The probability of having CRC with a high severity level was inversely related to the time to colonoscopy but not to the campaign year. Compared to patients having undergone colonoscopy within 30 days, the odds were significantly reduced by 60% in patients having undergone colonoscopy after 7 months (adjusted Odds-Ratio: 0.4 [0.3; 0.6]; p < 0.0001).

CONCLUSIONS

The French indicators were certainly degraded before the first wave of the COVID-19. The delay in access to colonoscopy as well as its extension induced by the COVID-19 crisis had no impact in terms of cancer severity, due to a discriminatory approach prioritizing patients with evident symptoms.

Faecal occult blood loss accurately predicts future detection of colorectal cancer. A prognostic model

OBJECTIVES

To examine the prognostic potential of repeated faecal haemoglobin (F-Hb) concentration measurements in faecal immunochemical test (FIT)-based screening for colorectal cancer (CRC).

DESIGN

Prognostic model.

SETTING

Dutch biennial FIT-based screening programme during 2014-2018.

PARTICIPANTS

265 881 participants completing three rounds of FIT, with negative test results (F-Hb <47 µg Hb/g faeces) in rounds 1 and 2.

INTERVENTIONS

Colonoscopy follow-up in participants with a positive FIT (F-Hb ≥47 µg Hb/g faeces).

MAIN OUTCOMES

We evaluated prognostic models for detecting advanced neoplasia (AN) and CRC in round 3, with as predictors, participant age, sex, F-Hb in rounds 1 and 2, and categories/combinations/non-linear transformations of F-Hb. Primary evaluation criteria included: risk prediction accuracy (calibration), discrimination of participants with versus without AN or CRC (optimism-adjusted C-statistics, range 0.5-1.0), the degree of risk stratification and C-statistics in external validation.

RESULTS

Among study participants, 8806 (3.3%) had a positive FIT result, 3254 (1.2%) had AN detected and 557 (0.2%) had cancer. F-Hb concentrations in rounds 1 and 2 were the strongest outcome predictors, with adjusted ORs of up to 9.4 (95% CI 7.5 to 11.7) for the highest F-Hb category. Risk predictions matched the observed risk for most participants (calibration intercept -0.008 to -0.099; slope 0.982-0.998), and discriminated participants with versus without AN or CRC with C-statistics of 0.78 (95% CI 0.77 to 0.79) and 0.73 (95% CI 0.71 to 0.75), respectively. The predicted risk ranged from 0.4% to 36.7% for AN and from 0.0% to 5.5% for CRC across participants. In external validation, the model retained similar discrimination accuracy for AN (C-statistic 0.77, 95% CI 0.66 to 0.87) and CRC (C-statistic 0.78, 95% CI 0.66 to 0.91).

CONCLUSION

Participants at lower versus higher risk of future AN or CRC can be accurately identified based on their age, sex and particularly, prior F-Hb concentrations. Risk stratification should be considered based on this information.