A new study has been published^a tentatively suggesting that there are significant health effects for those living in close proximity to gas fracking sites. The study may make headlines despite the authors expressly stating that the results should be viewed as ‘hypothesis generating’. There are a number of problems with the survey which could indicate small sample size and biases in adjusting for other factors account for the difference. Alternatively there is also the possibility that reported health effects of living near the fracking sites is due to stress from the false perceptions of the risks of living near to a fracking site. Anti-fracking environmentalists may be damaging people’s health and happiness through misinformation.

The study is

Proximity to Natural Gas Wells and Reported Health Status: Results of a Household Survey in Washington County, Pennsylvania (Environ Health Perspect; DOI:10.1289/ehp.1307732)

Peter M. Rabinowitz, Ilya B. Slizovskiy, Vanessa Lamers, Sally J. Trufan, Theodore R. Holford, James D. Dziura, Peter N. Peduzzi, Michael J. Kane, John S. Reif, Theresa R. Weiss, and Meredith H. Stowe

The households were split into three groups based on distance from a gas well. <1km (62 households), 1-2km (57) & >2km (61). The major result was

The number of reported health symptoms per person was higher among residents living <1 km (mean 3.27 ± 3.72) compared with >2 km from the nearest gas well (mean 1.60 ± 2.14, p=0.02).

The study also found significantly higher incidences in two out of five health symptoms in the <1km group than in >2km group.

There are multiple reasons for expecting these tentative results will not be replicated.

• The small sample size for a very complex set of data.
• Perceived water quality is not related to fracking.
• Failure to control properly for obesity and smoking
• Failure to repeat the sampling process with the same model.
• Failure to corroborate the results by checks for actual contamination.
• Biases in answering the questions.

1. Small sample size
   

There is an obvious problem with the health status study. The sample size was reported as the sample size of 180 households with 472 people, too small to generate meaningful results when there are a number of inter-related factors involved.

Consider how this sample was selected. To select these households the researchers randomly selected 20 points on a map in each of 38 townships. On a map they located the nearest house to the spot. The researchers were concerned with the possible impact of fracking on ground fed water supplies, which only applied to a minority of households. This was the main reason for reducing the sample From 760 data points to 227 households. 47 refusals reduced this to down to the 180 households for which questionnaires were received. They then put the data through a model “that adjusted for age, gender, household education, smoking, awareness of environmental risk, work type, and animals in house.”

The results were based on comparing two sample groups – one with 62 households and 150 people, the other with 61 households with 192 people. The >2km households were 30% larger than the <1km group, and the average age was 7 years lower. Not only were the numbers small, but there were material differences in the sample groups. It was necessary to adjust for

1. Perceived water quality is not related to fracking.
   

Sixty-six percent reported using their ground-fed water (well or natural spring) for drinking water and 84% reported using it for other activities such as bathing.

If there were health effects from contaminated water due to fracking, then there should be a difference in distance between those who drank the water and those who did not. But although there were more households who said the water has an unnatural appearance near the in the <1km group, (13/62 for <1km v 6/61 for >2km), the position was reversed when for those who said taste/odour prevented water use (14/62 for <1km v 19/61 for >2km). If people believed there was a problem with the water due to fracking, then those living near the wells might be more likely to avoid drinking the water than those further away. It was not the case. The proportions drinking the water were the same. It would appear that water quality is generally considered poor in the area. This point can be demonstrated by water sampling.

1. Failure to control properly for obesity and smoking
   

Obesity and smoking have long-been accepted as having consequences for health. The questionnaire is in the Supplemental Material. For obesity it asks the respondent their height and weight, but not the height and weight of the other members of the household. For smoking the question is

Does anyone in this household smoke regularly inside the house?

Smoking causes health problems independent of whether someone smokes inside their home or not. Also the numbers of people smoking in a household matters, along with the number of years smoked and the quantity of cigarettes smoked.

1. Failure to repeat the sampling process with the same model.
   

The model that filtered out other elements could have had some very large biases within it. For instance, the model could have over-adjusted for smoking. Conducting a completely fresh survey with the same sampling method would have eliminated this possibility.

1. Failure to corroborate the results by checks for actual contamination.
   

If there were actual health issues water contamination or air contamination, then there should be some evidence in water and air samples. The authors did not consult any actual monitoring results to show contamination. In the case of water quality In the case of air quality they threw everything at the issue, including ‘operation of diesel equipment and vehicles‘. If there was something in the air and/or in the water that is causing real health problems, then it will be something that cannot be perceived.

1. Biases in answering the questions.
   

In the introduction the authors say

A convenience sample survey of 53 community members living near Marcellus Shale development found that respondents attributed a number of health impacts and stressors to the development. Stress was the symptom reported most frequently (Ferrar et al. 2013).

The study said

We found instead that the refusal rate, while less than 25% overall, was higher among households farther from gas wells, suggesting that such households may have been less interested in participating due to lesser awareness of hazards.

If participation was higher in people nearer to wells because of perceived hazards, and the people get stressed by this. It could be that this stress exacerbates the symptoms and/or people on hearing stories of possible health effects notice their own conditions more. That is, the results of reported health effects of living near fracking sites may be to some extent real, but caused by the stress of believing the scare stories. This could be coupled with the fears of resulting in people remembering minor health symptoms, as there might be a cause. This alone could explain why the number of reported symptoms was twice the level for people living near to the gas wells. Conducting a similar, but larger survey with both dwellings where water is mains supplied and from ground-fed wells. If there is “something in the water”, then those who are mains supplied would not suffer from health effects to the same degree.

1. Thanks to commentator “Entropic Man” at a Bishop-Hill discussion thread for alerting me to this study.
   

Kevin Marshall