Hospital-Based Clusters and the Epidemiology of Gastroschisis

Recently, Elliott et al¹ examined a gastroschisis case series from 2 Reno, Nevada, hospitals, comparing patients with controls who had maternal birth dates within 1 year and who received prenatal care at the same clinics. Notable findings focused on patients with gastroschisis who had greater exposure to vasoconstrictive recreational drugs, including methamphetamine, prior to pregnancy. We raise several methodological concerns.

While the authors refer to the case series as a cluster in the title and throughout the article, they provide no statistical results of analyses to determine whether the case series meets formal criteria for a statistical cluster, nor does their methodology follow established guidelines for cluster investigations.^{2 - 3} The authors also fail to indicate whether prenatally diagnosed and electively terminated cases were included in the case series. Additionally, the authors do not indicate whether the series consisted exclusively of cases of isolated gastroschisis, yet multiple defects indicate a more complex etiology.⁴ For example, were cases of limb-body wall defect excluded? Third, epidemiologists typically define clusters by place of residence; of special interest here would be residence prior to conception or during the first trimester of the index pregnancy. While the reported birth prevalence estimate of 22.6 of 10 000 live births (95% confidence interval, 12.8-37.0) is higher than many recently published state-level estimates, this estimate potentially falls within normal limits if the catchment area for the denominator is broader than that used for this calculation.⁵

The case-control methodology the authors used raises additional concerns. Among the 389 controls initially identified, only 57 participated. While we presume that all patients with gastroschisis participated in the study, selection bias in favor of the vasoconstrictive recreational drug use result obtained may have occurred, as women who were recreational drug users would be less likely to participate in a study of this nature. Additionally, while 2 of 14 case subjects spoke Spanish, 26 of 57 controls required an interpreter, raising the possibility of cultural bias in their responses to the apparently unblinded interviewers. Also, time interval to interview averaged twice as long for controls (187 days) as for patients (94 days), which is indicative of potentially greater recall bias among controls. For these reasons, we suggest that readers view Elliott and colleagues' findings with even more circumspection than would normally be warranted in a small case-control study and await validation of these findings in larger, population-based, and ideally multicenter studies.

Correspondence: Dr Kirby, Department of Community and Family Health, College of Public Health, University of South Florida, 13201 Bruce B. Downs Blvd, MDC56, Tampa FL 33612 (rkirby@health.usf.edu).

Author Contributions:Study concept and design: Kirby. Analysis and interpretation of data: Kirby and Marshall. Drafting of the manuscript: Kirby. Critical revision of the manuscript for important intellectual content: Marshall. Statistical analysis: Kirby. Administrative, technical, and material support: Marshall.

Financial Disclosure: None reported.