Selection of study population, follow-up, and enrollment. The study population was selected from individuals enrolled in a cohort study on the effects of early-life exposure to PCE-contaminated drinking water. Participants provided written consent to undergo the vision examinations. This study was approved by the Institutional Review Boards of the Massachusetts Department of Public Health and Boston University Medical Center, and by the 24A/B/11B Review Committee at the Department of Public Health.
Participants were born from 1969 through 1983 (inclusive) to parents residing in one of eight towns in the Cape Cod region of Massachusetts (Barnstable, Bourne, Brewster, Chatham, Falmouth, Mashpee, Provincetown, and Sandwich). These towns had a proportion of their drinking-water distribution systems outfitted with asbestos-cement (AC) pipes whose vinyl liner (VL) was improperly cured and so it leached PCE into the water supply. These pipes were installed from 1968 through early 1980, according to the town’s replacement and expansion needs. Approximately 660 miles of VL/AC pipes were installed in Massachusetts, a large portion on Cape Cod. The irregular installation pattern led to a wide range of PCE levels in the drinking water; measurements taken in 1980 ranged from 1.5 µg/L to 7,750 µg/L (Demond 1982
). Reported levels of other water contaminants were generally low during this period. Because replacing the VL/AC pipes was prohibitively expensive, officials initiated a program of flushing and bleeding to reduce PCE levels to below 40 µg/L, the suggested action level for remediation when the contamination was discovered in 1980.
Selection and enrollment of the cohort have been described previously (Aschengrau et al. 2008
). “Exposed” individuals were identified as births to women who lived in affected homes by cross-matching the address on their birth certificates to geographic information systems data on VL/AC pipe locations. This initial designation was accomplished by visually inspecting water pipe maps in the immediate vicinity of the birth residence. “Unexposed” individuals were selected from births to women living in unaffected homes and were frequency matched on the month and year of birth to exposed individuals. This process provided a tentative exposure designation until more extensive assessments (described below) were completed.
Recruitment letters were mailed to successfully located individuals along with questionnaires to gather information on demographic characteristics, sources of solvent exposure, medical histories, behavioral factors, and a detailed residential history. Additional information including the subject’s date of birth, parents’ demographic characteristics, and maternal solvent exposure was obtained from birth records and questionnaires completed by mothers in 2002–2003 for a study of developmental outcomes in this population.
Among participants available for testing from the initial cohort, only singletons currently residing within the testing area with maternal questionnaire data were eligible for the vision examinations (). Individuals who reported severe hearing or vision problems, excessive drug or alcohol use, history of neurological disease, or occupational or environmental exposure to solvents were ineligible (). Of 112 exposed participants and 107 unexposed participants who met the eligibility criteria, 56% and 61%, respectively, never responded to any of our contact attempts (i.e., three letters and several telephone calls). Another 15% of exposed participants and 8% of unexposed participants refused to participate. Ultimately, 65 participants underwent vision testing.
Selection and enrollment of study population (n).
PCE exposure assessment.
A leaching and transport algorithm developed by Webler and Brown (1993)
for our prior epidemiological studies (e.g., Aschengrau et al. 2008
) was used to estimate the amount of PCE delivered to each reported residence during the subject’s gestation and through five years of age.
The components of the algorithm included the initial amount of PCE in the liner, the age of the pipe, the leaching rate of PCE from the liner into the water, and estimates of the direction and rate of flow of water derived from EPANET, water distribution modeling software developed by the U.S. EPA (Rossman 1994
), which accounts for the pipe configuration and number of users in a water system. The initial amount of PCE in the liner was determined based on the pipe’s diameter and length. Laboratory experiments suggested that the leaching rate of PCE from the vinyl liner into the water declined as the pipe aged. The leaching rate followed a simple first-order exponential decay relationship with a diffusion rate constant of 2.25 years and a half-life of 1.56 years (Demond 1982
; Gallagher et al. 2011
). Because the study area was predominantly residential, we assumed that each residence on the distribution network used the same amount of water.
Using these data we estimated the mass of PCE delivered to subject residences for each year of the study period, based on residential move-in and pipe installation years. We calculated cumulative exposure during gestation and early childhood as the sum of nine-twelfths of the estimated mass of PCE delivered to the residence during the birth year (representing an average 9-month gestation) and the estimated mass of PCE from the month and year following birth to the month and year of the fifth birthday. Exposure assessments beyond the fifth birthday could not be conducted with confidence for participants born at the end of the study period because of limitations in available water system records. We used simple percentages to account for partial years. Individuals with no PCE exposure using the algorithm were considered unexposed. Following this exposure assessment, four individuals initially thought to be unexposed were reclassified as exposed ().
We classified exposed individuals into “low” and “high” exposure groups using a cut point of 78.4 g that corresponded to being exposed to an average drinking-water PCE concentration of 40 µg/L with an average household use of 90,000 gallons/year during gestation and early childhood. Because few families moved from exposed to unexposed residences immediately following the birth of the subject, all participants with prenatal exposure also had some childhood exposure. Thus, we were unable to examine the independent impact of prenatal exposure alone.
Separate tests were administered to assess acuity, contrast sensitivity, and color discrimination by a trained examiner who was blinded to the exposure status of each subject. During the vision examinations, participants were required to wear their best corrective lenses for near viewing. Because acquired visual dysfunction can be unilateral, all tests were administered separately for each eye, except for the Farnsworth D-15 (Bowman 1982
), which was administered binocularly. All tests were conducted under standardized conditions consisting of an examination room illuminated by a daylight fluorescent lamp providing luminescence of 70 foot-Lamberts (corrected color temperature of 6,500° K; color rendering index > 90; intensity = 1,150 lux).
Near acuity test. The Rosenbaum Pocket Vision Screener (Grass Instruments Co., Quincy, MA) was used to assess acuity. A perfect acuity score is considered 20:20, and higher scores indicate poorer acuity. Participants placed their chin on the head support of the test card holder that held the card 14 inches from their eyes. With their left eye covered with a handheld occluder, participants read each number on the card progressing from top left to bottom right, beginning with the third row. Acuity score was obtained from the last row for which all numbers were correctly identified. The test was repeated for the right eye. Acuity scores for each eye were converted to LogMAR units.
Near contrast sensitivity test. The Functional Acuity Contrast Test (FACT; Stereo Optical Co., Chicago, IL) was used to assess contrast sensitivity. The FACT examination chart consists of five rows of eight sine-wave grating patches arranged in order of decreasing contrast in 0.15 log unit steps. Patches are arranged in order of increasing spatial frequency from low frequency [1.5 cycles per degree (cpd)] to intermediate frequencies (3 and 6 cpd) to high frequencies (12 and 18 cpd). Grating bars on each patch are either vertical or tilted 15 degrees to the left or right. The calibrated holder was used to hold the FACT chart 18 inches from the participants’ eyes. With their right eye covered with the occluder, for each row of grating patches participants were asked to state the orientation of the bars for the patch furthest to the right that they could see clearly. If they were correct, they continued to the right stating the orientation of grating bars for each patch in the row. If they were incorrect, they were asked to look back at each preceding patch until they gave a correct response. The raw contrast score was translated from the last correctly identified grating orientation at each spatial frequency. The test was repeated for the left eye.
Color discrimination. Both the Farnsworth D-15 and Lanthony Desaturated 15 Hue (D-15d) tests (Geller 2001
) were administered to participants to assess color discrimination according to recommended protocols. For both tests, participants were shown a rectangular box containing 16 colored magnetic caps arranged in chromatic order. Participants practiced manipulating the caps with the magnetic wand before the test. The examiner then scrambled the caps in front of the subject and correctly positioned the first cap. Participants were then asked to order the remaining caps in a regular color series. They were permitted to reorder caps at any time. When they finished, the box was flipped, and the cap order was documented. Numbers of transpositions of adjacent caps (minor errors) and cap reversals across two or more cap positions (major errors) were recorded. Because the distance in color and perceived space between each successive cap is not equivalent, not all errors represent the same level of deficiency in color discrimination. Bowman (1982)
and Geller (2001)
published estimates of the perceptual distances between each pair of caps for the Farnsworth and Lanthony tests, respectively. A Total Color Distance Score (TCDS) was calculated as the sum of the published perceptual distances between each pair of caps in the order placed by participants. A Color Confusion Index (CCI) was calculated for each subject as the ratio of their TCDS to the TCDS associated with a perfect performance (116.9 and 56.4 for the Farnsworth and Lanthony tests, respectively) such that a CCI of 1.0 indicates a perfect score for either test. The Farnsworth test, which consists of color caps that are more vivid than those of the Lanthony test, was completed binocularly first. Then the Lanthony test was completed separately for each eye. Participants chose which eye to test first. An adhesive patch covered the other eye during this test.
Statistical analysis. To avoid confounding by excessive optical blur, individuals with acuity scores worse than 20:70 in either eye were excluded from analyses (n = 1) (). To focus our assessment on exposure to PCE-contaminated drinking water, we also excluded individuals whose mother reported occupational exposure to solvents before or during their gestation (n = 10). Descriptive statistics were calculated for remaining participants.
The unit of analysis for each test was the average score of the left and right eyes, except for the Farnsworth test where raw CCI scores were analyzed. Linear regression models were used to estimate the mean differences [95% confidence intervals (CIs)] in acuity, contrast sensitivity at each spatial frequency, and CCI between exposed and unexposed participants. Differences in contrast sensitivity and CCI between the high- and low-exposure groups were also calculated. We also performed a repeated-measures profile analysis with an interaction term between PCE exposure group and spatial frequency assuming unstructured covariance to determine whether exposed and unexposed participants had different patterns of contrast sensitivity across the five spatial frequencies.
Confounders were identified as characteristics that met each of the following conditions: They preceded the exposure window, differed by > 10% between exposed and unexposed participants, and was an independent predictor of the outcome of interest (< 0.05), and inclusion in a multivariate model resulted in a > 10% change in the effect estimate for PCE exposure. No confounders were identified in our analyses of acuity or color confusion, so crude results are presented. Sex was identified as a possible confounder of the relationship between PCE exposure and contrast sensitivity, but only at spatial frequency 1.5 cpd. Therefore, crude contrast sensitivities are presented.
Study population characteristics by exposure status.
Effect measure modification by smoking status was assessed by stratification. Participants were classified as ever regular smokers or never regular smokers based on their answers to specific questions about smoking on “a regular basis.” All analyses were performed using PC-SAS (version 9.1; SAS Institute Inc., Cary, NC). An alpha level of 0.05 was used as the cut point for statistical significance.