Quasi-empirical Bayes methodology for improving meta-analysis.

This article addresses the problem of heterogeneity among various studies to be combined in a meta-analysis. We adopt quasi-empirical Bayes methodology to predict the odds ratios for each study. As a result, the predicted odds ratios are pulled toward the estimated common odds ratio of the various studies under consideration. With strong heterogeneity among the studies, we jointly consider the display of the 95% CIs of the ORs and a Dixon's test (1950) for "outliers" to exclude the "extreme" estimated ORs. We demonstrate the effectiveness of our methodology based on the data analyzed by Thompson and Pocock (1987) demonstrating the power of the new approach to meta-analysis to find statistical agreement in what looks like great disagreement via a chi-squared test. We believe our technique (i.e., minimum mean-square sense) will go a long way toward increasing the trustworthiness of meta-analysis.

Prospective evaluation of the prevalence of gastric Helicobacter pylori infection in patients with GERD, Barrett's esophagus, Barrett's dysplasia, and Barrett's adenocarcinoma.

OBJECTIVE: This study was undertaken to prospectively determine the prevalence of gastric H. pylori infection in Barrett's esophagus and Barrett's complicated by dysplasia or adenocarcinoma. METHODS: The prevalence of H. pylori was determined in Barrett's esophagus patients compared to a control population of patients with gastroesophageal reflux disease (GERD) only. All patients had a minimum of 10 gastric surveillance biopsies obtained. H. pylori colonization was determined upon the basis of hematoxylin and eosin and use of a modified Giemsa and or Steiner's silver stain of all gastric biopsy specimens. RESULTS: Two hundred and eighty-nine Barrett's patients and 217 GERD control patients were included in the study. H. pylori was found in 95/289 (32.9%) of the Barrett's patients, compared with 96/217 (44.2%) of the GERD controls (NS). Forty-seven of the Barrett's patients had low-grade dysplasia/indefinite dysplasia, 14 high-grade dysplasia, and 20 Barrett's adenocarcinoma. When Barrett's was subgrouped according to absence of dysplasia, and presence of low-grade dysplasia, high-grade dysplasia, or adenocarcinoma, H. pylori prevalence was found to be significantly less for patients with Barrett's high-grade dysplasia (14.3%) and adenocarcinoma (15.0%) versus patients with GERD alone (44.2%), Barrett's alone (35.1%), or Barrett's with low-grade dysplasia (36.2%) (p = 0.016). This difference could not be explained by differences between Barrett's esophagus patients infected with H. pylori and those who were not with respect to gender, smoking history, alcohol consumption, use of proton pump inhibitor, or length of Barrett's mucosa. CONCLUSIONS: Barrett's high-grade dysplasia and adenocarcinoma are significantly more prevalent in patients who are not infected with H. pylori. H. pylori appears to have a protective effect against the development of Barrett's adenocarcinoma.

Prospective multivariate analysis of clinical, endoscopic, and histological factors predictive of the development of Barrett's multifocal high-grade dysplasia or adenocarcinoma.

OBJECTIVE: Our goal was a prospective follow-up of Barrett's esophagus to determine what clinical, endoscopic, and histological features at the time of initial diagnosis are predictive of the development of Barrett's adenocarcinoma or multifocal high-grade dysplasia (HGD). METHODS: Newly diagnosed Barrett's esophagus patients were prospectively followed with a standardized endoscopic and bioptic surveillance protocol. Features examined by chi2 and stepwise logistic regression analyses as potential predictors the development of multifocal HGD or adenocarcinoma included age, length of Barrett's segment, hiatal hernia size, severity of dysplasia at diagnosis, severity of dysplasia during surveillance, and type of long-term medical treatment. RESULTS: One hundred-eight Barrett's patients have had follow-up ranging from 12 months to 101 months (mean +/- SD, 39.9+/-20.8 months), for a total of 361.8 patient-years. Overall, five patients developed multifocal HGD and five developed adenocarcinoma. The incidence of adenocarcinoma as well as multifocal HGD was 1 per 71.9 patient-years. Chi2 analysis showed progression to Barrett's multifocal HGD/adenocarcinoma was associated with hiatal hernia (p = 0.02), the length of Barrett's (p = 0.001), the presence of dysplasia at diagnoses (p < 0.001) or anytime during surveillance (p < 0.001). Stepwise logistic regression analysis revealed progression to multifocal HGD or adenocarcinoma was significantly and independently associated with presence of dysplasia at diagnosis (p < 0.0001) or anytime during follow-up (p < 0.03), hiatal hernia size (p < 0.02, for hernia > or =3 cm), and length of Barrett's (p = 0.009, >2 cm). CONCLUSIONS: Endoscopic and histological features of Barrett's esophagus patients at initial diagnosis are predictive of risk of progression to cancer.

Prospective multivariate analysis of factors predictive of complete regression of Barrett's esophagus.

OBJECTIVE: Demographic, endoscopic, and histological features of Barrett's esophagus at initial diagnosis were examined for their ability to predict complete endoscopic and histological regression of Barrett's during long-term follow-up. METHODS: Barrett's patients who have been followed up for a minimum of 2 yr and who have had at least two follow-up surveillance examinations were included in the analysis. Complete regression of Barrett's was defined as total disappearance of all tongues and patches of Barrett's epithelium at endoscopy (confirmed with Lugol's iodine) in conjunction with only squamous epithelium on biopsy. Chi2 and stepwise logistic regression analyses were performed on the following clinical, endoscopic, and histological variables with regards to their ability to predict complete Barrett's regression: patient age in years (<65 or > or =65), length in cm of Barrett's (< or =2, >2<6, and > or =6), presence of a hiatal hernia (yes or no), presence of dysplasia at diagnosis (yes or no), and type of long-term medical treatment (histamine antagonists, proton pump inhibitor [PPI], or PPI and cisapride). RESULTS: Ninety-nine patients, all men with a mean age +/- SD in years of 61.6+/-13.1 have been followed prospectively for 24-106 months (mean +/- SD, 48.0+/-19.8). Seven patients have had complete regression of Barrett's. Univariate analysis showed that complete regression of Barrett's was associated only with absence of a hiatal hernia (p = 0.012). Stepwise logistic regression analysis revealed that complete regression was significantly and independently associated again only with absence of a hiatal hernia (p = 0.025). Stepwise logistic regression analysis utilizing only hiatal hernia (yes vs no) and length of Barrett's (<6 cm vs > or =6 cm) as variables revealed that absence of a hiatal hernia (p = 0.0447) and shorter lengths (<6 cm) of Barrett's (p = 0.0418) were significantly and independently predictive of complete Barrett's regression. CONCLUSIONS: The absence of a hiatal hernia was noted to be the most important factor associated with Barrett's regression. Complete regression of Barrett's esophagus occurs in a minority of patients, primarily in those with no hiatal hernia and shorter lengths of Barrett's epithelium.

Migraine in twins raised together and apart.

History of headache and migraine diagnosed by structured interview utilizing International Headache Society criteria was obtained in two samples of female twin pairs--154 raised together and 43 raised apart since infancy. Zygosity was determined by standard methods. Probandwise concordance rates were determined. Assuming that liability to migraine has a multifactorial etiology involving a number of genetic and environmental risk factors acting additively, tetrachoric correlations in the four groups of twins (monozygotic raised together, monozygotic raised apart, dizygotic raised together, and dizygotic raised apart) were then calculated. The genetic and environmental influences in the liability to migraine were estimated using biometrical model-fitting methods. Tetrachoric correlations of migraine were consistently higher in monozygotic than in dizygotic twins, for both reared together and reared apart samples, with a heritability estimate of 52%, Model fitting also suggested that approximately 50% of the variance in liability to migraine was attributable to genetic factors with nonshared environmental factors and measurement errors responsible for the remaining variance.

Prospective, multivariate evaluation of CLOtest performance.

OBJECTIVES: To determine prospectively the effect of aspirin, nonsteroidal anti-inflammatory drugs, H2-receptor antagonists, proton pump inhibitors, alcohol intake, race, age, history of Helicobacter pylori eradication attempts, and gastric biopsy location on CLOtest performance. METHODS: Biopsy specimens were obtained from the antrum, greater curve, and proximal stomach. One biopsy specimen from each site (except for the proximal stomach) was used for the CLOtest and two or more specimens were used for histopathology. Giemsa staining was used for the definitive determination of H. pylori status. RESULTS: One hundred seventeen patients were included in the study, and 50 of these were infected with H. pylori. The sensitivity and specificity of an antral CLOtest, based on the results of Giemsa-stained sections from the antrum alone, were 72.7 and 98.6%, respectively, whereas they were 66.0 and 100% when based on the results of Giemsa-stained sections from all three gastric biopsy sites. The sensitivity and specificity of a body CLOtest, based on the results of Giemsa-stained sections from the body alone, were 80.5% and 93.4%, whereas they were 76.0% and 100% when based on the results of Giemsa-stained sections from all three gastric biopsy sites. Combining CLOtest results from the antrum and body increased CLOtest sensitivity to 82.0%. Univariate analysis revealed only one factor significantly associated with CLOtest concordance: absence of alcohol consumption (p < 0.02). Stepwise logistic regression demonstrated that absence of alcohol use again was independently associated with the concordance between overall CLOtest and all Giemsa-stained biopsy specimen results (p < 0.03) as well as between body CLOtest and body Giemsa-stained biopsy specimen results (p < 0.03). Additional independent and significant associations were noted between antral CLOtest-antral Giemsa-stained biopsy specimen results and no history of H. pylori eradication attempts (p = 0.04) and between body CLOtest-body Giemsa-stained biopsy specimen results and race (African-American) (p < 0.03). CONCLUSIONS: Obtaining a gastric biopsy specimen from the antrum as well as from the body increased CLOtest detection of H. pylori. CLOtest performance was shown to be affected by several demographic and clinical factors.

Late hyponatremia in very-low-birth-weight infants: incidence and associated risk factors.

OBJECTIVE: To determine the incidence of late hyponatremia in very-low-birth-weight infants and to identify associated risk factors. Low serum sodium concentration in otherwise healthy premature infants beyond 2 weeks of life is referred to as the hyponatremia. DESIGN: Retrospective cohort review. SETTING/SUBJECTS: The intensive care nursery at St Luke's Hospital Perinatal Center, Kansas City, Mo. Criteria for subject selection were birth weight of 1,500 g or less; survival for more than 21 days; development of late hyponatremia or hospitalization for 42 days or more; and measurement of serum sodium concentration at least once between the 14th and 56th day of life. Of 515 infants admitted to the nursery for the 1992 calendar year, 124 had a birth weight of 1,500 g or less and survived for more than 21 days; however, 16 of these infants were discharged at 42 days of life or less, 11 did not have a serum sodium concentration measurement after the second week of life, and 1 did not have complete medical records. Thus, the final sample was 96 subjects. STATISTICAL ANALYSES: Percentages to determine incidence of late hyponatremia; t test of chi 2 test to determine differences between infants with and without late hyponatremia; multiple logistic regression to determine the strongest indicators of late hyponatremia. RESULTS: Incidence of late hyponatremia was 62.5%. Significant risk factors for late hyponatremia were birth weight of 1,000 g or less (P < .001), feedings of fortified human milk (P < .013), and occurrence of an intraventricular hemorrhage (P < .036). Fortified human milk feeding was a significant risk factor for late hyponatremia in both weight groups (ie, birth weight greater than or less than 1,000 g). CONCLUSION: Despite standard fortification, human milk may contain an insufficient quantity of sodium to met the needs of very-low-birth-weight infants.

High-risk children referred to an early-intervention developmental program.

Early-intervention programs for infants with developmental disabilities or with high-risk factors for such problems were first established in the United States more than 20 years ago. The benefits of such programs are generally recognized. This study describes the presenting problems of 698 children referred to an early-intervention program over a 15-year period (1975-1989). Medical condition groups rather than specific diagnoses are considered. The developmental progress of 464 children who attended the program for at least 6 months was determined by comparing their admission and discharge developmental quotients (DQ). Admission trends over time are noted and the value of intervention programs for young children with disabilities is discussed.

Efficacy of once daily extended-release theophylline in decreasing the use of inhaled beta 2-agonists in stable, mild-to-moderate asthma patients.

BACKGROUND/OBJECTIVE: The purpose of this study was to determine whether the addition of extended-release theophylline to the daily treatment regimen of inhaled beta 2-agonist users would result in decreased use of beta 2-agonist while maintaining similar efficacy for treatment of asthma. METHODS: This was a single-blind, multicenter (six sites) study. Sixty-one patients with a history of mild-to-moderate asthma treated with inhaled beta 2-agonist were randomized to treatment with Theo-24 (anhydrous extended-release capsules) plus inhaled beta 2-agonist or placebo plus beta 2-agonist. Patients kept daily symptom diaries, measured peak flow rates, recorded puffs of inhaled beta 2-agonist, and adverse events during a 4-week treatment period. RESULTS: Fifty-five patients were included in the efficacy analysis. The primary efficacy variable in this study was the mean number of puffs (adjusted for baseline differences) of beta 2-agonist inhaled per day. In this study, the addition of theophylline to the daily regimen of inhaled beta 2-agonist for 4 weeks significantly reduced the total daily dose of inhaled beta 2-agonist at weeks 3 and 4 of treatment compared with placebo. The differences were significant at the P < .05 level. For patients in the theophylline group, the number of puffs decreased from an unadjusted mean of 9.81 at baseline to an adjusted mean of 6.78 after 4 weeks of treatment compared with 9.91 at baseline and 8.17 for the placebo group. There were no unexpected or serious adverse events. CONCLUSIONS: In this study, the addition of once daily, extended-release theophylline to the daily regimen of inhaled beta 2-agonist for 4 weeks significantly reduced the total daily dose of inhaled beta 2-agonist at weeks 3 and 4 of treatment compared with placebo, while maintaining acceptable asthma symptom scores.

Transmission of Toxoplasma gondii in Panama City, Panama: a five-year prospective cohort study of children, cats, rodents, birds, and soil.

A cohort of more than 500 children from Panama City, Panama was studied prospectively over five years for acquisition of antibody to Toxoplasma gondii. The direct agglutination test showed that 72 of 571 children seroconverted between one and six years of age, for a cumulative incidence of 12.6%. Children were examined by pediatricians quarterly, and illnesses that had occurred in the interval and their activities were noted on questionnaires. Thirty-eight variables were examined for their role as risk factors for seroconversion. There was a higher correlation between children's seroconversion and contact with dogs than with cats. Combinations of significant predictors without dogs explained only 67% of the seroconversions, but the same factors with dogs explained 90%. On the other hand, ingestion of raw or rare meat or eggs appeared to play no role in transmission. Cats were examined and 110 (45.6%) of 241 had Toxoplasma antibody on the first bleeding. Only two (0.5%) of 383 cat fecal specimens, when tested in mice, resulted in seroconversion. Ten (1.1%) of 924 soil samples resulted in seroconversion in mice that had been injected. Antibody to Toxoplasma was found in 52 (23.3%) of 226 rats (Rattus norvegicus) and two (0.035%) of 571 mice (Mus musculus). Two hundred sixteen birds of 16 different species were bled. Antibody to Toxoplasma was found in 13.4% of these birds, mostly in grackles, blue-gray tanagers, and doves. The rate of isolation of Toxoplasma was low: one of 23 in rats and three of 201 in birds. High relative risks (RRs) of transmission to children were predicted by contact histories with nursing dogs (RR = 5.8), weaned dogs (RR = 4.7), many flies (RR = 3.6), 6-12-month-old dogs (RR = 3.4), weaned cats (RR = 3.0), 6-12-month-old cats (RR = 2.7), nursing cats (RR = 2.5), much garbage (RR = 2.4), and many roaches (RR = 2.2). The high statistical correlation of dog contact with seroconversion in children suggests the possibility that dogs, by eating and rolling in cat feces, are instrumental in mechanically transmitting Toxoplasma infection. In addition, flies, and to a lesser extent, cockroaches, may have practically important roles in transmission.

Iliac crest bone graft harvest donor site morbidity. A statistical evaluation.

STUDY DESIGN: This study analyzed the cause, rate, and risk factors of iliac crest bone graft donor site morbidity. OBJECTIVES: All complications or problems, no matter how small, were sought to develop strategies of prevention. SUMMARY OF BACKGROUND DATA: A wide range of major, 0.76% (Keller et al) to 25% (Summers et al) and minor complications, 9.4% (Keller et al) to 24% (Summers et al) has been reported. METHODS: A consecutive series of 261 patients, whose bone graft harvest was done by one surgeon, was studied by chart review and a mail survey that was not conducted by the operating surgeon. The survey presented specific open-ended questions designed to uncover any complication/problem, no matter how small. Complications then were categorized as major or minor and subcategorized as acute or chronic. Statistical analysis was done using chi-squared and multiple logistical regression. RESULTS: None of the 261 patients had a severe perioperative complication--e.g., superior gluteal artery injury, sciatic nerve injury, or deep wound infection. None of the 225 patients with long term follow-up (average, 66 months; range, 32-105 months) had a severe late complication--e.g., donor site herniation, meralgia paresthetica, pelvic instability, or fracture. Of the 180 patients meeting the qualifications for statistical analysis, major complications occurred in 18 (10%), only three of which affected function (pain). Minor complications occurred in 70 (39%). CONCLUSIONS: The results indicated that severe complications from iliac crest bone graft harvest can be avoided and major complications affecting functioning are uncommon, but minor complications are common. The findings suggest that procedural refinements of limiting subcutaneous dissection and providing layered tension-free incision closure may improve results.

Prevalence of aneuploidy, overexpressed ER, and overexpressed EGFR in random breast aspirates of women at high and low risk for breast cancer.

Breast tissue biomarkers which accurately predict breast cancer development within a 10 year period in high risk women are needed but currently not available. We initiated this study to determine 1) the prevalence of one or more breast tissue abnormalities in a group of women at high risk for breast cancer, and 2) if the prevalence of biomarker abnormalities is greater in high risk than in low risk women. Eligible high risk women were those with a first degree relative with breast cancer, prior breast cancer, or precancerous mastopathy. Low risk women were those without these or other major identifiable risk factors. Ductal cells were obtained via random fine needle aspirations and cytologically classified. Biomarkers included DNA ploidy, estrogen receptor (ER), and epidermal growth factor receptor (EGFR). The prevalence of DNA aneuploidy was 30%, overexpression of ER 10%, and overexpression of EGFR 35%, in the 206 high risk women whose median 10 year Gail risk (projected probability) of developing breast cancer was 4.5%. The prevalence of aneuploidy and overexpressed EGFR was significantly higher in the high risk women than in the 25 low risk controls (p < 0.002), whose median 10 year Gail risk was 0.7%. The difference in the prevalence of ER overexpression between high and low risk groups was not statistically significant (p = 0.095). This may be due to the low prevalence of overexpressed ER and the small number of controls. A significant difference was noted in the prevalence of one or more abnormal biomarkers between the high risk and low risk women (p < 0.001). A large prospective trial is needed to determine if one or more of these biomarkers, is predictive of breast cancer development.

In vitro growth modulation by L-ascorbic acid of colony-forming cells from bone marrow of patients with myelodysplastic syndromes.

In vitro colony growth was studied on bone marrow cells from 51 patients with myelodysplastic syndromes (MDS), using a cell culture method with the unique feature of daily feeding, in an effort to gain insight into the pathophysiology of MDS and to assess the clinical utility of this cell culture assay. The colony growth pattern of MDS marrow cells is remarkably similar to that of acute myeloid leukemia but quite dissimilar from that of normal marrow, in support of a common pathophysiological mechanism for these two disorders. In particular, L-ascorbic acid (LAA) enhanced colony growth in 30% and suppressed growth in 16% of cases, a finding also similar to that in acute myeloid leukemia, indicating a unique growth requirement which may be explored for therapeutic purposes. Further, these LAA effects have prognostic value, with LAA-sensitive (both LAA-enhanced and LAA-suppressed) cases displaying shorter survivals than LAA-insensitive cases (median survival of 5 months versus 18 months; P = 0.011). This prognostic value is independent of, and more powerful than, bone marrow blasts; the median survival was 18 months for less than 5% bone marrow blasts and 8 months for greater than 5% bone marrow blasts (P = 0.044). These two risk factors can be used together to identify patients with an extremely good or an extremely poor prognosis. This study establishes the clinical usefulness of the LAA effect in MDS as a prognostic factor and provides a new lead to explore in understanding differential biochemical/molecular events and, possibly, a new therapeutic approach to the management of MDS.

The effect of stimulus duration using the Brookler-Grams closed-loop caloric irrigator.

Nystagmic responses to traditional 40-sec, 30 degrees C water calorics were recorded and compared to those obtained with the Brookler-Grams closed-loop irrigator using 30 degrees C, 40-, 50-, and 60-sec irrigations. Significant differences were noted between the responses to the water irrigator and the 40- and 50-sec closed-loop irrigations. The 60-sec closed-loop irrigation produced responses that were equivalent to the responses obtained with the water irrigation in slow component velocity, amplitude, frequency, latency, and duration. The 30 degrees C 60-sec closed-loop irrigation is an acceptable stimulus in electronystagmographic caloric testing.

Identifying the needs of home caregivers of patients with cancer.

Increasing numbers of patients with cancer are being cared for by home caregivers. The primary purpose of this methodologic, correlational study was to identify, categorize, and assess the importance of needs expressed by 492 home caregivers and to determine how well these needs were satisfied. Caregivers surveyed were selected from the records of two nonprofit community cancer agencies and two hospital outpatient oncology clinics in the Midwest. These individuals were identified by at-home patients with cancer as unpaid people who helped with physical care or coping with the disease process. Caregiver characteristics and patient activity were examined to determine their relationships to caregiver needs, and needs were examined over time. The 90-item Home Caregiver Need Survey used in this study was developed by the author in 1989 and demonstrated internal consistency, reliability, and construct validity. Using factor analysis, six need categories were identified: psychological, informational, patient care, personal, spiritual, and household. Caregivers' greatest needs were informational and psychological. Significant correlations between certain caregiver characteristics and caregiver needs and between caregiver needs and patients' activity levels were found. Both the importance and satisfying of needs changed over time. Findings indicate the urgent need for nurses, who usually provide support for caregivers, to establish specific programs and services to meet the identified and unmet informational and psychological needs of caregivers of at-home patients with cancer. Frequent reassessment of caregiver needs seems to be indicated.

Classification of essential tremor.

Classification of essential tremor was attempted using tremor frequency; tremor duration; family history of tremor; responsiveness to alcohol, propranolol, and primidone; muscle contraction pattern; and long-latency reflexes. Sixty-one patients were evaluated. The majority of patients had a tremor frequency less than 7.0 Hz, a positive family history, and a positive response to alcohol. Approximately 46% of patients had a beneficial response with propranolol and 71% with primidone. Tremor frequency was inversely correlated with age and directly correlated with an antagonist pattern of muscle contraction. Enhanced long-latency reflexes were not found. Other characteristics of essential tremor were not significantly correlated. It is concluded that essential tremor can not be classified into subtypes.

Efficacy and efficiency: self-designed versus instructor-designed study tools.

During the course of their education, occupational therapy students learn to administer complex structured assessments. For easier administration of these assessments, students design note cards, which then replace cumbersome test manuals during administration. This study considered whether students could learn test administration with equal efficiency and efficacy if given test administration note cards rather than having to design their own. The results showed that the subjects using instructor-designed cards earned written test and practical examination scores similar to those of the subjects using self-designed cards. The subjects using instructor-designed cards spent significantly less (p = .003) total time in study than did the subjects using self-designed cards. The difference in time between the two groups was attributable to the time spent designing note cards. Therefore, distribution of instructor-designed note cards appears to offer equally effective and significantly more efficient learning when compared with that produced when students design their own cards. The differences in efficacy and efficiency were similar for students of different learning styles (as classified by Witkin's field-dependence/field-independence continuum) [corrected].

Comparison of responses to air, water, and closed-loop caloric irrigators.

Three types of commercially available caloric irrigation units--air, water, and closed-loop--were evaluated for nystagmus response patterns and comfort. Electronystagmographic recordings were analyzed for slow-phase velocity, response duration, amplitude, frequency, and test-retest reliability. Significant differences were found among the three types of irrigators, between warm and cool irrigators, and between male and female subjects. Test-retest reliability was adequate for all three units. Subjects clearly favored the closed-loop system over the water and air irrigators, but the water irrigator yielded the most robust caloric response.

Environmental risk factors in siblings with Parkinson's disease.

To investigate possible risk factors in Parkinson's disease, we conducted a case-controlled study of 19 families having two or more siblings with Parkinson's disease. Demographic data were collected, including lifetime histories of places of residence; sources of drinking water; occupations, such as farming; and exposure to herbicides and pesticides. Rural living and drinking well water, but not farming and herbicide exposure, were significantly increased in 38 parkinsonians compared with 38 normal control subjects. A comparison of parkinsonian siblings with siblings with essential tremor revealed no differences in any risk factors for the years of shared environment. These data suggest that living in a rural environment and drinking well water are risk factors for Parkinson's disease and that the total life exposure to an environmental toxin may be more important than exposure in early life.