Validity of the hospital discharge diagnosis in epidemiologic studies of biliopancreatic pathology. PANKRAS II Study Group

BACKGROUND

The aim was to analyse the magnitude, direction and predictors of change in the main hospital discharge diagnosis (HDD) after a clinical expert review, among patients included in a multicentre molecular epidemiologic study of biliopancreatic diseases.

METHODS

A total of 602 patients with a suspicion diagnosis of pancreas cancer (PC), cancer of the extrahepatic biliary system (CEBS) or benign biliopancreatic pathologies (BPP) were prospectively recruited at five general hospitals. A structured form was used to collect information from medical records. A panel of experts revised all diagnostic information and established the main clinicopathological diagnosis (CPD) by consensus.

RESULTS

Of the 204 cases with a HDD of PC, 176 (86%) were deemed to have a CPD of PC, eight of CEBS, twelve a neoplasm of different origin, four BPP and four syndromic diagnoses. Thus, 28 cases (14%) were false positives. Of the 129 patients with a HDD of CEBS, 15 (12%) were false positives. Nine of the 396 cases with a HDD of non-PC (2%) had a CPD of PC (false negatives), whilst 14 of 471 patients with a HDD of non-CEBS (3%) were deemed to have CEBS. Overall, sensitivity and specificity of HDD for PC were, respectively, 95 and 93%, and for CEBS, 89 and 97%. Cytohistological confirmation and laparotomy were independent predictors of diagnostic change.

CONCLUSIONS

Validity of the HDD was high, but its association with some clinical variables suggests that sole reliance on HDD can significantly bias results, and highlights the need to review all HDDs. Alternatively, only patients at high risk of misdiagnosis could be reviewed: primarily, those lacking a cytohistological diagnosis or a laparotomy. No exclusions appear warranted solely on the basis of age, gender or tumour spread.

Learning from case reports: diagnostic issues in an epidemiologic study of pancreatic cancer

Epidemiologic studies on exocrine pancreatic cancer show a large heterogeneity in diagnostic criteria applied to define "caseness." Reanalyses conducted after review of diagnostic information have yielded substantially different results than those based on more crude classifications of disease. During a multicenter prospective study on mutations in the K-ras gene in pancreatic and biliary diseases, hospital diagnoses from 602 patients were reviewed by a panel of experts. There were two main motivations to do so: a generic interest for the quality of the diagnostic data, and the anticipation that a firm diagnosis could be needed to assess whether patients whose tumors did not harbor the mutation were true negatives or false negatives. In addition, the review of diagnoses was helpful to minimize tissue misclassification, and it had a high educational value for clinicians and epidemiologists. This article illustrates why and how this was so through a brief presentation of the 10 most significant cases. With respect to selection and classification of subjects, the main issues that studies on pancreatic cancer need to address are the differential diagnosis of exocrine pancreatic cancer and pancreatitis, the differential diagnosis of exocrine pancreatic cancer and other abdominal tumors, and the use of survival as a hallmark of pancreatic cancer. In epidemiologic studies of pancreatic cancer, it is warranted that a panel of experts centrally reviews all the existing diagnostic evidence (cytohistological and other) of all patients, regardless of whether they have cytohistological confirmation and of their hospital discharge diagnosis.