We used data from a case-control study examining several risk factors for CKD 17-19
. The study population consisted of hospital patients and community controls, age 30-79 years residing in North Carolina between 1980 and 1982.
Cases were patients from one of four North Carolina medical centers (Duke University Medical Center, North Carolina Memorial Hospital, Charlotte Memorial Hospital and North Carolina Baptist Hospital) with newly diagnosed CKD, identified by review of kidney-related ICD-9 discharge diagnoses. Potential cases had at least one of the following ICD-9 codes given as part of their discharge diagnosis: 250.4 (diabetes with renal manifestations), 403 (hypertensive renal disease), 404 (hypertensive heart and renal disease), 582 (chronic glomerulonephritis, including interstitial nephritis), 583 (nephritis and nephropathy, including diabetic nephropathy), 585 (chronic renal failure), 586 (renal failure, unspecified), 587 (renal sclerosis, unspecified), 590.0 (chronic pyelonephritis), 590.8 (other pyelonephritis, not specified as chronic or acute), and 593.9 (unspecified disorder of kidney and ureter). Inclusion criteria required that patients, in addition to having newly diagnosed CKD, also had 2 or more measurements of serum creatinine, all of which must have been greater than 1.5 mg/dL. Patients were excluded on the basis of age under 30 years, residence outside North Carolina, preexisting kidney disease or evidence of prior creatinine measurements greater than 1.5 mg/dL (ascertained from medical histories and chart review), and evidence of normal kidney function (despite the diagnosis of CKD). In addition, patients with systemic lupus erythematosus, polycystic kidney disease, or missing silica exposure data were excluded.
Controls were selected using random digit dialing and Medicare recipient listings and were frequency matched to case patients by age (within 5 years), gender, race and proximity to the hospital (residence in or adjacent to counties containing study hospitals). Medical histories and exposure data were obtained for both cases and controls with telephone interviews conducted by trained study personnel.
Six hundred and seven of 709 case patients could be contacted for interview, 554 of whom (91%) participated in the study (78% overall response rate for cases). Among the control participants, 608 of 717 could be contacted and 520 (86%) were interviewed (73% overall response rate for controls). Interviews revealed 4 controls with a history of probable kidney disease and these participants were excluded from the study. Job histories were provided by proxy for 302 cases. (Proxies included, for example, a spouse or caregiver). After also excluding those with missing silica exposure data, a total of 504 cases and 457 controls were included in the analysis. The sample demographics were similar before and after exclusions.
A detailed account of participants’ work history was collected. In addition to general questions on exposure to sand or silica and employment in industries or occupations with potential for exposure (e.g. stone, clay, glass manufacturing), data were collected on all jobs that were held for two or more consecutive years. Along with open-ended descriptions of job-specific activities and what the employer specialized in, participants were also asked (1) if, for each reported job, they were exposed to sand or silica at least 5 times; or (2) if they were exposed to other dusty conditions at least 5 times.
Based on complete information from lifetime job histories, silica exposure was assessed independently (and blinded to case-status) by authors (CGP and LNF) from specific occupational data. For each job held, qualitative dose estimates were assessed (high, moderate, and low) and differences resolved by consensus review. Examples of exposure ratings are available in the Appendix
. A certainty rating was applied to each exposure estimate based on the extent and quality of data provided by respondents. Jobs for which exposure was possible, but unlikely, were also identified to permit sensitivity analysis. For each individual, lifetime exposure ratings were calculated based on the assigned exposure dose and the length of time at each job. Exposure scores were weighted by a factor of 1.0 for high certainty, 0.75 for moderate certainty, and 0.25 for low certainty. Weighting for exposure intensity ranged from 0 (unexposed) to 3 (high), for increasing levels of exposure. The product of intensity by certainty was summed across years spent at each job with silica exposure for analysis of cumulative exposure. In addition to the job-based exposure estimates, self-reported work in specific industries or with specific materials collected by checklist was grouped as probable (i.e., ever have a job in stone, clay or glass manufacture; and work with sand or silica, clay, ceramics or pottery products, and ceramic glazes) and possible (i.e., ever have a job in: chemical manufacture, auto mechanics or repair, plumbing, heating or air conditioning, smelting, lead or other metal industry, paint manufacture, commercial painting or spray painting; work with paint such as restoring homes, removing house paint, ship repair or repainting).
During the interview, data on demographics, clinical measures, medications and medical conditions (including hypertension, diabetes, gout, urinary tract infections, pyelonephritis and kidney stones) were collected. Annual income, years of education, and height and weight were also ascertained. Body mass index (BMI) was calculated using weight in kilograms divided by height in meters squared.
Means and standard deviations for continuous variables and percents for categorical variables were calculated for various traits of the study population by case and control status. The distribution of baseline characteristics among cases and controls were compared using Student’s t-tests and chi-square tests.
In order to examine the association between occupational silica exposure and CKD, odds ratios (OR) were calculated using conditional logistic regression models. The models were conditioned on race, gender, and proximity to hospital. Since cases were matched within five years of age to controls (as opposed to matched in five-year age groups), age was included in the models as continuous variable. Initial regression models adjusted for age and education. Potential confounders also controlled for in fully adjusted models included age, respondent status (proxy vs. self), education level, body mass index, and history of hypertension and diabetes. We added individual covariates to a base model that included age and education and demonstrated only slight differences in ORs. The OR was unchanged after adjusting for all covariates in a single model compared to the base model. Thus, to improve the precision of estimates for analyses of CKD subtype we report estimates from the reduced models. Multiplicative interaction terms of silica exposure with diabetes and with hypertension status were evaluated using product terms in fully adjusted regression models. All statistical analyses were conducted using SAS 8.0 (SAS Institute, Cary, NC).