Work

Reply to limitations in the imputation strategy to handle missing nativity data in the Surveillance, Epidemiology, and End Results program

Year: 2014

Type: letter

Abstract: We appreciate the comments from Pineda et al regarding our article,1 and agree with their statement that the distribution of missing birthplace is not at random and is related to vital status. Although they are correct in noting a large discrepancy in the proportion of US-born and foreign-born individuals when comparing the imputed data to cases whose birthplace was unknown in 2006 but became known in 2012, we would like to comment on 2 issues: 1) when applying the exclusion criteria specified in our original analysis,1 the proportion of cases with updated nativity is small, representing ≤11% of cases with missing birthplace and <5% of total cases; and 2) when applying the exclusion criteria, the nativity distribution of cases with updated nativity is quite different from what they present. We requested a custom data linkage of Hispanic cases diagnosed from 1995 through 2004 who were reported in 2006 and 2012. Among cases of cervical cancer, there were 98 cases for whom nativity was unknown in 2006 but became known in 2012, representing 6% of 1636 cases with missing nativity and <2% of the total cases. Among these, 53% were found to be US-born and 47% were foreign-born, a distribution overall not different from the 1:1 ratio expected if death certificate reporting is independent of birthplace. For prostate cancer cases, nativity became known for 1229 cases, representing <10% of cases with missing data and <5% of the total cases. Similarly, the observed distribution reflected independence between reporting and birthplace (53% US-born). Only for colorectal cancer was the observed distribution different from what may be expected by chance, with US-born cases accounting for 66% of cases with updated data. However, these 671 cases still represent 11% of the cases with missing data and only 4% of the total cases. We did not observe significant changes in the survival curves presented in our original analyses1 with the addition of the updated data (affecting <5% of cases). Although birthplace will eventually (and inevitably, due to death certificate reporting) become known for all cases, there is no reason to suspect that the data will follow the same distribution as the cases for whom birthplace became known (≤11%). Given the need for timely analyses of the public Surveillance, Epidemiology, and End Results (SEER) data set, we argue that multiple imputation of missing birthplace is an improved approach over listwise deletion, the usual method for handling missing birthplace data in SEER, which would exclude >30% of the original sample. Supported in part by a grant from the National Institutes of Health (P01CA082710; PI: M. Follen). Dr. Montealegre and Ms. Zhou are supported by an Innovation for Cancer Prevention Research Postdoctoral (Dr. Montealegre)/Predoctoral (Zhou) Fellowship (The University of Texas School of Public Health-Cancer Prevention and Research Institute of Texas grant RP101503). The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health or the Cancer Prevention and Research Institute of Texas. The authors made no disclosures. Renke Zhou, MS Division of Biostatistics The University of Texas School of Public Health Dan L. Duncan Cancer Center Baylor College of Medicine Houston, Texas Jane R. Montealegre, PhD Division of Epidemiology, Human Genetics, and Environmental Sciences The University of Texas School of Public Health Dan L. Duncan Cancer Center Baylor College of Medicine Houston, Texas E. Susan Amirian, PhD Michael E. Scheurer, PhD Dan L. Duncan Cancer Center Baylor College of Medicine Department of Pediatrics Baylor College of Medicine Houston, Texas

Source: Cancer

DOI: 10.1002/cncr.28868

Authors: Renke Zhou, Jane R. Montealegre, E. Susan Amirian, Michael E. Scheurer

Institutions: Baylor College of Medicine, Dan L Duncan Comprehensive Cancer Center

Language: English

Topic: Global Cancer Incidence and Screening

Subfield: Oncology

Field: Medicine

Domain: Health Sciences

SDGs: Good health and well-being, Reduced inequalities

Open Access status: green

Funder: National Cancer Institute