Applicability of Number Needed to Screen or Invite as Measures of Cancer Screening Efficiency Beyond Clinical Trials
Menée dans le cadre de l'essai "The Nordic-European Initiative on Colorectal Cancer" (NordICC) et à partir de données portant sur 84 958 personnes ayant reçu ou non une invitation pour participer à un programme de dépistage (âge : de 55 à 64 ans ; absence d'antécédents de cancer colorectal), cette étude estime les délais d'enregistrement des cancers colorectaux par les registres nationaux et évalue l'impact de ces délais sur les effets du dépistage rapportés par le NordICC
Colorectal cancer (CRC) is the second leading cause of cancer deaths worldwide, and improving delivery of screening across all populations is critical for further decreasing incidence and mortality. An invitation to screen is a commonly used population-based strategy to improve the reach of CRC screening. The number needed to invite (NNI) or screen (NNS) to prevent 1 CRC diagnosis or death, the reciprocal of the absolute risk difference (RD), is used to assess the efficiency of screening strategies or tests. This study by Brenner and colleagues reanalyzed data from a randomized clinical trial (RCT) of invitation to screening colonoscopy to show how delayed reporting of cancer diagnosis and mortality by registries used for outcome ascertainment could lead to spuriously high NNI and NNS. In the original RCT, Bretthauer et al examined the invitation strategy in 84 585 presumed-eligible individuals aged 55 to 64 years in Poland, Norway, and Sweden. The RCT randomly assigned men and women to invitation (28 395 individuals) or usual practice (56 785 individuals) and ascertained CRC diagnosis and death over 10 to 15 years of follow-up using cancer registries. The analysis excluded people who were found to have CRC (221 individuals) or had died (373 individuals) prior to randomization and only confirmed after inclusion due to delays in registration. In the analysis, 10-year CRC cumulative incidences were 0.98% in the invitation group and 1.20% in the usual practice group (RD, 0.22%) in intention-to-treat (ITT) analysis. In per-protocol analysis, 10-year CRC cumulative incidences were 0.84% in the invitation group vs 1.22% in the usual practice group (RD, 0.38%). Based on the 221 people excluded due to registry reporting lags, Brenner et al estimated a 2- to 3-year cancer registration delay. Accounting for reporting delay with an assumption of nondifferential ascertainment bias, they derived RDs of 0.44% for a 2-year delay and 0.88% for a 3-year delay in ITT analysis and 0.76% for a 2-year delay and 1.52% for a 3-year delay in per-protocol analysis. The corresponding NNI and NNS for a potential 2-year reporting delay were 227 and 132, respectively, in contrast to 455 and 263 from the RCT results.
JAMA Network Open , éditorial, 2023