This is a DRAFT (and far from exhaustive) review of the information provided by Health Canada (HC) pertaining to the HC Wind Turbine Noise Study and is designed to serve as a starting point for further discussion.
A review of the information provided by Health Canada with regards to the preliminary results published on the HC Wind Turbine Noise Study requires the consideration of a number of reports / articles / pamphlets. Specifically, the following found on the HC Wind Turbine Noise web-site:
- Summary of Results
- A Primer on Noise
- Frequently Asked Questions
- Results Pamphlet
- Additional Information
- Health Impacts and Exposure to Sound from Wind Turbines: Updated Research and Design and Sound Exposure Assessment
- Notice to Stakeholders – HC Wind Turbine Noise and Health Study
Additionally, the following article can be found in the trade news publication Noise News International.
Self-reported and Objectively Measured Health Indicators among a Sample of Canadians Living Within the Vicinity of Industrial Wind Turbines: Social Survey and Sound Level Modelling Methodology
For ease of review, sections copied from the various reports / articles listed above will be presented in the following discussions in blue typeface.
Although specified in the Michaud Noise News International publication, Self-reported and Objectively Measured Health Indicators among a Sample of Canadians Living Within the Vicinity of Industrial Wind Turbines: Social Survey and Sound Level Modelling Methodology, the Summary of Results section of the HC web-site does not state that this is an epidemiological cross sectional survey. Following is from the book, Epidemiology in Medicine (C. Hennekens, M.D., Dr.P.H. and Julie Buring, ScD).
For factors that remain unaltered over time, such as sex, race, or blood group, the cross-sectional survey can provide evidence of a valid statistical association. Such instances are rare, however, and for the vast majority of associations evaluated, the temporal relationship between exposure and disease cannot be clearly determined. Thus, cross-sectional studies are, in general, useful for raising the questions of the presence of an association rather than for testing a hypothesis.
Additionally, several published methods for ranking study designs with respect to their individual “weight of evidence” have been developed, one of the most widely accepted is listed below. While systematic reviews and meta-analyses are considered the gold standard of assessing a body of evidence, taken alone as an individual study, a cross sectional survey ranks as the least reliable method of study, marginally above individual case reports and expert opinion. The following is the hierarchical list in descending order of reliability:
- Systematic reviews and meta-analyses
- Randomized controlled trials (RCT) with definitive results (confidence intervals that do not overlap the threshold clinically significant effect)
- Randomized controlled trials with non-definitive results (a point estimate that suggests a clinically significant effect but with confidence intervals overlapping the threshold for this effect)
- Cohort studies
- Case-control studies
- Cross sectional surveys
- Case reports (individual patient evaluations)
The above hardly supports the claims listed in the HC Results Pamphlet:
The Wind Turbine Noise and Health study is a landmark study and the most comprehensive of its kind. Both the methodology used and the results are significant contributions to the global knowledge base and examples of innovate, leading edge research.
Updated Research Design and Sound Exposure Assessment Section: Statistical Power of Report
The following is from the Statistical Power section of the Updated Research Design and sound Exposure Assessment:
As sleep disturbance is a frequent health complaint associated with WTN in observational and case studies, one of the primary research objectives in the study is to quantify the magnitude of sleep disturbance due to WTN. There are currently no population-based normative data that are derived from actimetry from communities exposed to WTS. Therefore, statistical power in the study is based on reported sleep disturbance. Estimated sleep disturbance in the general adult population is approximately10% (Riemann et al., 2011; Tjepkema, 2005), with some estimates as high as 40% (National Sleep Foundation, 2005). In calculating the sample size needed for statistical power to detect a sleep disturbance in this study, the conservative estimate of 10% will be used. Based on a sample of 2000 dwellings and assuming that 20% of the sample live in close proximity to wind turbines (at 40 dBA or above) investigators will be able to detect at least a 7% difference in prevalence rates between the general population and the sample of individuals living in closest proximity to wind turbines, with a 5% false positive rate and a power of 80%.
This sample size calculation incorporates the following assumptions: a) there will be an 80% occupancy rate for dwellings in rural areas and b) there will be a 70% participation rate for sleep actimetry. A sample size of 1800 dwellings would be required, however given the possibility that not all assumptions may be met and that prevalence rates will be adjusted for other covariates in a logistic regression model (for example gender, age, receiving financial benefit, house construction type among others), the sample has been increased to 2000 dwellings. Based on the estimated sample size of 2000 dwellings, all other objective endpoints should be equally predicted with similar confidence. Other studies that have used actimetry to characterize aircraft sound impacts on sleep are based on far fewer subjects (Passchier-Vermeer et al., 2002; Fidell et al., 1995; Horne et al., 1994; Ollerhead et al., 1992).
Issue: The number of participants in the study is listed at 1,234, not the 1,800 dwellings (upped to 2,000 ) required to establish a statistical power of 80%. Therefore, the “reported sleep disturbance” analyses and “all other objective endpoints” in this study are not considered reliable as a result of the study failing to meet its own sample size and target power calculations, as stated in the Updated Research Design and Sound Exposure Assessment. Assuming the study followed the study design identified in the Updated Research Design and Sound Exposure Assessment, the study cannot identify the key outcomes at the chosen incidence / prevalence levels. In short, the study does not have the statistical power to discern the prevalence of the key outcomes as identified in the protocol.
Therefore all statements along the lines of, “No evidence was found to support a link between exposure to wind turbine noise and (X outcome variable)” are not supported by the study and are not based on the results of the study. Assuming the study followed the study design identified in the Updated Research Design and Sound Exposure Assessment this study cannot be used as the basis for the claims that are being made by Health Canada regarding the lack of association between wind turbines and health. The study has insufficient sensitivity and specificity.
From the Frequently Asked Questions (FAQ) – Section: Why did Health Canada take noise measurements in the summer when there is very little wind?
It is true that wind is, on average, lower during the summer months compared to other times of the year. In order to minimize awareness bias, noise measurements in any given community only began after Statistics Canada had completed the in-home interview. This meant that noise measurements began in June 2013 and continued through October 2013. Some of the field measurements needed to be carried out during the summer months simply due to the timing of the collection period. Taking measures during summer months simply meant that it sometimes took longer to acquire enough data (waiting for windy periods). It is important to note that field measurements were taken only to the extent that enough data was collected to validate the calculated A-weighted values used in the study. This objective was met.
Issue: The following need to clarified:
- Were noise measurements undertaken at all homes?
- How much time was spent at each home completing noise measurements?
A table detailing the number of homes within each sound / distance “bin” (group) at which noise measurements were undertaken would be expected in any peer-reviewed publication. This table would be expected to provide the amount of time spent at each home completing the noise measurements. The resulting table would provide data on “measured” versus “calculated” noise levels.
Of particular concern is the statement “Taking measures during summer months simply meant that it sometimes took longer to acquire enough data (waiting for windy periods). It is important to note that field measurements were taken only to the extent that enough data was collected to validate the calculated A-weighted values used in the study. This objective was met.” This seems to suggest that “validation” consisted of noise measurement equipment being installed in some homes during the relatively windless summer months. The equipment then remained at the home until the modeling-predicted average sound level was reached and then noise measurement ceased.
Data underwent several layers of modelling in order to 1) use a small sample of noise to predict overall means and 2) to work backwards from that data to enable cross-study comparisons. Each procedure introduces error and uncertainty. Importantly, the study fails to adequately explain why raw data at source could not be used.
Summary of Results – Section: Study Population and Participation
“All potential homes within approximately 600 m of a wind turbine were selected, as well as a random selection of homes between 600 m and 10 km.”
Read full review here….