Validity of self-reported data on pregnancies for childhood cancer survivors: a comparison with data from a nationwide population-based registry

Abstract

To what degree do records registered in the Netherlands Perinatal Registry (PRN) agree with self-report in a study questionnaire on pregnancy outcomes in childhood cancer survivors (CCSs)? This study suggests that self-reported pregnancy outcomes of CCSs agree well with registry data and that outcomes reported by CCSs agree better with registry data than do those of controls. Many studies have shown that childhood cancer treatment may affect fertility outcomes in female CCSs; however, these conclusions were often based on questionnaire data, and it remains unclear whether self-report agrees well with more objective sources of information. In an nationwide cohort study on fertility (inclusion period January 2008 and April 2011, trial number: NTR2922), 1420 CCSs and 354 sibling controls were invited to complete a questionnaire regarding socio-demographic characteristics and reproductive history. In total, 879 CCSs (62%) and 287 controls (81%) returned the questionnaire. The current validation study compared the agreement between pregnancy outcomes as registered in the PRN and self-reported outcomes in the study questionnaire. A total of 589 pregnancies were reported in CCSs, and 300 pregnancies in sibling controls, of which 524 could be linked to the PRN. A high intra-class correlation coefficient (ICC) was found for birthweight (BW) (0.94 and 0.87 for CCSs and controls, respectively). The self-reported BWs tended to be higher than reported in the PRN. For gestational age (GA), the ICC was high for CCSs (0.88), but moderate for controls (0.49). CCSs overestimated GA more often than controls. The Kappa values for method of conception and for method of delivery were moderate to good. Multilevel analyses on the mean difference with regard to BW and GA showed no differences associated with time since pregnancy or educational level. Not all pregnancies reported could be linked to the registry data. In addition, the completeness of the PRN could not be assessed precisely, because there is no information on the number of missing records. Finally, for some outcomes there were high proportions of missing values in the PRN registry. Our study suggests that questionnaires are a reliable method of data collection, and that for most variables, self-report agrees well with registry data. This work was supported by the Dutch Cancer Society (grant no. VU 2006-3622) and by Foundation Children Cancer Free. None of the authors report a conflict of interest. NTR2922 http://www.trialregister.nl/trialreg/admin/rctview.asp?TC=2922.