We appreciate the opportunity to respond to the letter by Goldstein et al.¹ The letter mentions a few issues that Goldstein et al consider to be major flaws of our study. Unfortunately, it appears that Goldstein et al did not fully understand the databases we used or our analytical approach. We welcome the opportunity to further clarify these issues.

Our analysis focused on all wells drilled in Pennsylvania between 1990 and 2009. The Pennsylvania Department of Environmental Protection (PDEP) database does not specifically identify hydraulic fractured (HF) wells. The variables, “Marcellus Shale (yes/no),” well type (oil, gas, or combined oil and gas), and “Horizontal Well (yes/no),” were available in the PDEP database at the time we downloaded data. In our analyses, we considered horizontal gas wells to most likely be HF wells. Goldstein et al state that they do not find the same number of wells in the PDEP database as we did. They fail to recognize that the PDEP database is a “living” database, and that it is constantly being updated and corrected (Roger Dietz, IT specialist in the Pennsylvania DEP Office of Oil & Gas Management, personal communication, August 13, 2013). In addition, the database was redesigned in April 2012 as a result of Pennsylvania House Bill 1950.² Act 13 required the PDEP to keep track of “unconventional gas development.” Therefore, the categories of wells had to be changed. The variable “Marcellus Shale” was replaced with “Unconventional well.” The Marcellus Shale is an unconventional formation, but so are 9 other formations in Pennsylvania. Furthermore, an unconventional well is one that is drilled on an unconventional formation, not necessarily how it is drilled (vertical or horizontal). If one chooses to select unconventional wells for a report from the updated PDEP database, all horizontal and vertical wells drilled on these 10 formations will be selected. Goldstein et al chose all unconventional wells by year, which is why they got different numbers of wells than we did.

Goldstein et al were concerned that analyses by lag period were not conducted. We have completed analyses for 1-, 2-, and 5-year lag periods. These analyses do not change our conclusions.

Goldstein et al state that the time period that we examined “misrepresents the timeline of the issue of concern” and that any effect we found “would be highly diluted by their apparent use of perhaps a decade of cancer data in each county to typify the ‘after drilling’ period.” We apologize if our statistical methods were not clear. We examined cancer risks in two time periods. The first time period was from 1990 to the year before the first well of the specific type of wells examined for the analyses was drilled in a county and the second time period was from the year the first well of the specific type of wells examined for the analyses was drilled in a county to 2009. For example, the before drilling time period for the analyses of horizontal gas wells included cancer risk between 1990 and the date before the first horizontal gas well was drilled in a county.

Goldstein et al also express concerns that time at risk for exposure was not considered in our analyses. We clearly state that the design of this type of study does not allow us to determine individual-level exposures for any individual in the study. However, these types of study have been extremely useful in assessing risks assumed to be associated with environmental exposures and are used by the National Cancer Institute³ and state health departments.⁴

We are disappointed by the comment in Goldstein et al that our study obfuscates the issue of health effects associated with HF. Our analyses are based on publically available data that are free for any person who wishes to download it from the Internet and conduct further analyses. There is no evidence to suggest that the Pennsylvania Cancer Registry data or the well drilling data from the PDEP are biased. In contrast, in their comments about our study, Goldstein et al present their recent findings on health impacts and stressors perceived to result from unconventional shale gas development in the Marcellus Shale region⁵ from a convenience sample of admittedly biased individuals whose participation rate and exposure to unconventional gas development were unknown. The results of these analyses⁵ do not provide reliable evidence that unconventional gas development is related to any stressor or health impact.

Our study addressed a concern raised by a previous investigation that there was a high risk of childhood leukemia in Pennsylvania counties that have large amounts of hydraulic fracturing.⁶ We believe that replication of findings is a fundamental goal of science, and we sought to do so. Using the same data sets as the previous investigators, we found no association. We do not believe that following accepted scientific methods in an effort to replicate a previous finding fits the definition of obfuscation. Rather, it comports with good science.