Accuracy of family history data on Parkinson's disease

Background: Genetic studies of PD frequently rely on family history interviews (FHI), yet the accuracy of data obtained in this way is unclear. Objective: To assess the interinformant reliability and validity of family history information on PD in first-degree relatives of PD cases and controls. Methods: A structured FHI was administered to nondemented PD cases and controls and to a second informant (self-report, sibling or child of the subject) for each relative. Interinformant agreement was assessed on four algorithm-derived diagnostic categories of PD: definite, definite or probable, definite, probable or possible ("conservative diagnosis"); or definite, probable, possible, or uncertain ("liberal diagnosis"). The sensitivity and specificity of each diagnostic category were assessed, using as the gold standard diagnoses based on either in-person examination or medical record review. Results: Five hundred thirty-six families containing 2,225 first-degree relatives were included in the interinformant reliability study. Agreement between informants was excellent for definite or probable PD for all three pair-wise comparisons: proband vs self-report (kappa = 0.92), proband vs sibling of subject (kappa = 0.80), and proband vs child of subject (kappa = 0.87). Agreement was also good to excellent for the conservative diagnosis (kappa = 0.66, 0.49, and 0.79). In the validity analysis (141 individuals in 96 families), the conservative diagnosis provided the best combination of sensitivity (95.5%) and specificity (96.2%) for the proband's family history report. No difference was apparent across categories defined by case or control status, relationship to the proband, or gender or age at onset of the proband. However, specificity was lower for deceased relatives than for living relatives. Conclusion: The FHI can be used to obtain reliable and valid family history information on PD in first-degree relatives when a conservative diagnostic algorithm is applied.