We thank Ms Harris for her insightful comments, and wholeheartedly agree that much more work needs to be done, in both rural and urban settings, on this important topic and its relationship to maternal and child health.

In sum, we agree that studies on rural food environment and perinatal outcomes are lacking and should be advanced at a higher resolution than our initial study allowed. We acknowledged in our article the following limitations, which Ms Harris appropriately emphasizes. First, our study setting is, by design, an urban study in the third most populous center in the United States and therefore does not aim to analyze the prevalence of gestational diabetes mellitus (GDM) in a rural setting. Moreover, subject population draws entirely from Harris County, which exhibits the second highest population density within Texas (population density per square mile in 2010: 2402.41 ). We acknowledged in our study that our findings may not be applicable to different geographic areas (ie, rural areas). Second, we benefitted from our large cohort (N = 8912 subjects), but did so with the inherent limitation of lack of subject-specific high-resolution geographic information systems data. In line with our institutional review board protocol and best practice in the conduct of research, we limit collection of our patients’ residency information to the ZIP code level, rendering individual geographic information systems coding unfeasible. We concur that it would be of great interest to gain deeper insight at a higher resolution in both urban and rural settings, and we are very much looking forward to seeing Ms Harris’ future contributions to this important field of research.

We do wish to clarify two key statements in Ms Harris’ letter to the editor. First, we did include convenience stores in our analysis and concur that they are potential sources for food of lower quality. However, our analysis revealed that convenience store density in ZIP code areas and neighborhoods was not associated with GDM occurrence ( P = .207 and P = .458). We speculate that this may be due to fewer subjects relying on convenience stores for regular food intake, and the variation in both quality and quantity of food. However, this remains mere speculation and we have no data supporting nor refuting such a premise. We further concur that this may vary in rural settings, and further research is needed. Second, we do wish to emphasize that the plausibility of our own approach toward analyzing associations of GDM occurrence with food environment is, despite the named limitations, strongly supported by the correlation of hemoglobin A1c levels with fast food restaurant density (Spearman rank, r = 0.290, P = .009).

We again wish to thank Ms Harris for taking the time to provide insightful feedback to our work, and look forward to hers and others important advances in this area of investigation in the years to come.