Absence of Association between Organic Solvent Exposure and Risk of Chronic Renal Failure

A Nationwide Population-Based Case-Control Study

Abstract

ABSTRACT. Exposure to organic solvents has been suggested to cause or exacerbate renal disease, but methodologic concerns regarding previous studies preclude firm conclusions. We examined the role of organic solvents in a population-based case-control study of early-stage chronic renal failure (CRF). All native Swedish residents aged 18 to 74 yr, living in Sweden between May 1996 and May 1998, formed the source population. Incident cases of CRF in a pre-uremic stage (n = 926) and control subjects (n = 998), randomly selected from the study base, underwent personal interviews that included a detailed occupational history. Expert rating by a certified occupational hygienist was used to assess organic solvent exposure intensity and duration. Relative risks were estimated by odds ratios (OR) in logistic regression models, with adjustment for potentially important covariates. The overall risk for CRF among subjects ever exposed to organic solvents was virtually identical to that among never-exposed (OR, 1.01; 95% confidence interval [CI], 0.81 to 1.25). No dose-response relationships were observed for lifetime cumulative solvent exposure, average dose, or exposure frequency or duration. The absence of association pertained to all subgroups of CRF: glomerulonephritis (OR, 0.96; 95% CI, 0.68 to 1.34), diabetic nephropathy (OR, 1.02; 95% CI, 0.74 to 1.41), renal vascular disease (OR, 1.16; 95% CI, 0.76 to 1.75), and other renal CRF (OR, 0.92; 95% CI, 0.66 to 1.27). The results from a nationwide, population-based study do not support the hypothesis of an adverse effect of organic solvents on CRF development, in general. Detrimental effects from subclasses of solvents or on specific renal diseases cannot be ruled out. E-mail: [email protected]

Exposure to organic solvents has long been suggested to cause and exacerbate renal disease (¹), in particular glomerulonephritis (^2–12). Support for this association comes mainly from case-control studies that suffer from small sample size, imperfect exposure assessment, equivocal case definition, inappropriate control groups, and lack of information on important covariates (^13,14).

Most previous case-control studies were limited to various glomerulopathies, and the role of organic solvents in the development of other nephropathies remains unclear. In one study, solvent exposure was associated with the development of diabetic nephropathy (¹⁵), and in two others, solvent exposure was associated with either chronic renal failure (CRF) overall (¹¹) or specific diagnoses suspected to be linked to occupational nephrotoxins (⁸) albeit without dose-response trends. To evaluate whether organic solvent exposure is a significant factor in the development of CRF overall or in specific subgroups of CRF, we conducted a nationwide population-based case-control study in Sweden of incident CRF, while attempting to avoid the limitations of previous investigations.

Materials and Methods

Setting

The continuously updated Swedish National Population Register provided a well-defined study base of all 5.3 million native Swedes, 18 to 74 yr of age, who were resident in the country during the ascertainment period, May 20, 1996, through May 31, 1998.

Study Subjects

Details regarding inclusion of subjects and collection of data have been described previously (¹⁶). Eligible as cases were men and women whose serum creatinine level exceeded 300 μmol/L (3.4 mg/dl) and 250 μmol/L (2.8 mg/dl), respectively, for the first time and remained elevated above these levels. For ensuring complete identification of all eligible patients, medical laboratories covering practically all inpatient and outpatient care in Sweden provided monthly lists of all serum creatinine measurements performed. Physicians at each hospital treating patients with renal disease determined case eligibility by reviewing the medical records of all identified patients. A second creatinine measurement, 3 mo after the first, was taken when the chronicity of the renal failure was uncertain. For allowing for day-to-day variation, the thresholds for this second measurement were lower (250 μmol/L for men and 200 μmol/L [2.3 mg/dl] for women). Diagnoses of underlying renal disease were based on routine clinical evaluation. Patients with CRF of prerenal (e.g., severe heart failure) or postrenal (i.e., significant urinary outlet obstruction) origin and patients with kidney transplants were excluded.

The control subjects were randomly selected throughout the ascertainment period from the National Population Register and were frequency-matched to the case patients according to age (in 10-yr strata) and gender. The regional ethics committees and the Swedish Data Inspection Board approved the study protocol. All study subjects were included after informed consent.

Data Collection

Each subject first received a mailed self-administered questionnaire inquiring about education, marital status, anthropometric measures, and personal lifestyle factors such as alcohol intake, tobacco use, and diet. In a subsequent face-to-face interview, information was obtained on occupational history, physical activity, and medical history including a detailed history of lifetime use of nonnarcotic analgesics, as described elsewhere (¹⁶). At the personal interviews, the interviewers from Statistics Sweden examined the mailed questionnaire to ensure completeness and helped the subject to complete missing answers. The interviewers could not be kept blinded to the case-control status of the subjects. They were, however, unaware of the study hypotheses and were trained to interview subjects in a standardized manner.

Organic Solvent Exposure

The occupational history included company name, type of industry, occupational title, work tasks, and duration for each work period of at least 1 yr during the subject’s lifetime. Occupations were coded according to the Nordic Standard Classification of Occupations (¹⁷) and the type of industry according to the Swedish Standard Industrial Classification (¹⁸). The interviewers also explicitly inquired about occupational and leisure-time exposure to organic solvents, including descriptions of implicated activities, frequency, and duration. To evaluate exposure status and the necessity for additional exposure information, a senior occupational hygienist (G.N.), blinded to case/control status, reviewed the occupational histories and the information on solvent exposure. Every subject who was presumed to be exposed underwent an additional standardized telephone interview, whereupon detailed information was sought regarding work tasks, equipment, duration and frequency of solvent use, the general work environment including ventilation and use of personal protective equipment, and changes in the conditions over time. We intended to keep the occupational hygienist assistant who performed the additional telephone interviews blinded to the case/control status of the subjects, but we could not avoid that some case patients revealed their renal disease during the interview.

For every work period, the senior occupational hygienist estimated the intensity of exposure to organic solvents (e.g., aliphatic, alicyclic, aromatic hydrocarbons, aldehydes, ketones, alcohols, glycols, glycol ethers, or mixtures of these compounds) for each job held, using the expert rating method as described by Siemiatycki et al. (¹⁹). The exposure was classified on a five-level scale in terms of approximate additive hygienic effect (HE). Exposure to an HE of 1.0 for a single solvent corresponds to an average exposure level during an 8-h working day equal to the occupational exposure limit (OEL) prescribed by the Swedish Work Environment Authority in 1996 (²⁰). The arithmetic mean exposure levels for each work period were calculated as the sum of fractional contributions from different organic solvents in relation to their respective OEL. The classes were (1) “unexposed,” corresponding to no more than background level exposure (<3% of the OEL); (2) “low,” ≥3% and <10% of the OEL; (3) “intermediate,” ≥10 and <30% of the OEL; (4) “high,” from 30% of the OEL to the OEL; and (5) “very high,” exposure above the OEL. The reliability of exposure estimation and classification was tested through reassessment of the information given by 53 (19%) randomly selected exposed patients and 52 (18%) exposed control subjects. The agreement between initial assessment and reassessment was excellent (κ = 0.87).

Cumulative lifetime exposure for organic solvents was calculated as the product of the intensity (HE), exposure frequency (days per month), and the duration (years) of the exposure, summed over all work periods in the subject’s occupational history. Dose estimations were categorized according to quartiles among exposed control subjects.

Statistical Analyses

Odds ratios (OR) with 95% confidence intervals (CI) estimated relative risk of CRF in different strata of exposure to organic solvents using the risk among nonexposed as reference. We used unconditional logistic regression to model this relationship while controlling for potentially confounding covariates. The final model was based on the scientific literature and the statistical significance of explanatory variables in the model, as assessed by the likelihood-ratio test (²¹). It contained terms for gender, age (in 10-yr strata), alcohol consumption (grams per week), smoking (cumulative cigarette pack-years), and lifetime cumulative dose of acetaminophen and aspirin (grams). Smoking, alcohol consumption, and acetaminophen and aspirin doses were categorized in quartiles according to the distribution among exposed control subjects. We examined the associations of organic solvent exposure with overall CRF and with disease-specific CRF such as glomerulonephritis, diabetic nephropathy, and renal vascular disease. To address possible confounding by socioeconomic status, we analyzed the associations between solvents and CRF separately for manual and nonmanual workers as defined by the official Swedish socioeconomic classification scheme. Adjustment for hypertension was not done, because hypertension may be the result of CRF rather than its cause and adjustment for this condition in multivariate models will not yield interpretable results. To minimize the possible impact of reverse causation, i.e., that employability and organic solvent exposure were influenced by progressing CRF, we performed analyses in which exposure during the 10 yr before the interview was disregarded. Furthermore, to explore the possibility that recent exposure hastened CRF progression, exposure before the most recent 10-yr period was disregarded. The mean dose of solvent exposure was compared by the t test. To assess the dose-response relation, we performed tests for trend on groups of ordinal variables by assigning each group a score equal to the mean value, fitting the resulting scores into the model, and assessing statistical significance by the likelihood-ratio test.

Results

We identified 1189 eligible patients. Sixty-nine died before they could be contacted, 83 were too ill to participate, and 111 declined participation. Thus 926 (78%) of the case patients participated. Of 1330 eligible control subjects, 998 (75%) participated; 221 declined, 56 could not be reached, and 55 had diseases that precluded participation. Ninety-nine percent of the participating patients (n = 913) and control subjects (n = 991) provided information on occupational history.

There were approximately twice as many men (n = 1,250) as women (n = 674) among our study subjects. The mean age was similar among men (58 yr) and women (57 yr). The clinical diagnoses among case patients are shown in Table 1. Diagnoses were based on renal biopsies in 30% of the patients (n = 277). Diabetic nephropathy was the most common underlying renal disease diagnosis (31%), whereas nearly one fourth (24%) of the patients had glomerulonephritis. In 61% (n = 135) of the patients with glomerulonephritis, a renal biopsy was performed. Of the 926 patients in this study, 120 (13%) reported not having a diagnosis of a renal disease before inclusion. The mean disease duration reported by the 806 patients who previously had a diagnosis of renal disease was 10 yr (median, 4 yr; interquartile range, 2 to 13 yr). The median value of the estimated creatinine clearance (²²) was 21 ml/min (interquartile range, 17 to 26 ml/min). Seventy percent of the patients (n = 646) had a clearance between 15 and 29 ml/min, and the clearance was <15 ml/min among 19% (n = 174) of the patients.

Table 1: Clinical diagnoses and measures of renal function among 926 patients with CRF^a

Case patients and control subjects did not differ with regard to number of reported occupations (median, 3 in both groups). The industry in which most solvent exposure of 3% of the OEL or more occurred was manufacturing. Seventy percent of the case patients’ and 62% of the control subjects’ reported solvent exposure was during manufacturing work. The largest group was metal workers (24% [n = 67] among case patients and 21% [n = 60] among control subjects), whereas 6% (n = 16) of the case patients and 7% (n = 21) of the control subjects were painters and floor layers (Table 2).

Table 2: Distribution of solvent-exposed and unexposed subjects by occupational group^a

Overall, 30% of the case patients (n = 276) and 29% of the control subjects (n = 288) were classified as exposed to organic solvents (Table 3). Among male patients and control subjects, 41% (n = 242) and 38% (n = 245), respectively, reported organic solvent exposure ≥3% of the OEL, whereas the corresponding numbers among women were only 10% (n = 34) and 13% (n = 43). Twenty-one percent (n = 59) of the exposed case patients reported solvent exposure above the OEL for at least one work period. The corresponding proportion among the control subjects is 15% (n = 42). Among 83 case patients with lifetime cumulative organic solvent dose in the highest quartile, 19% (n = 16) worked in the metal industry and 18% (n = 15) were painters and floor layers. The corresponding proportions among 72 control subjects were 22% (n = 16) and 17% (n = 12), respectively. Lifetime cumulative doses, exposure durations, and frequencies were not statistically different between exposed patients and exposed control subjects. The difference between the mean average yearly exposure dose, however, was borderline significant (P = 0.15; Table 3).

Table 3: Statistics of organic solvent exposure variables among 276 exposed patients with CRF and 288 exposed control subjects^a

Exposure to organic solvents was not associated with an increased risk of CRF after adjustment for potential confounders (OR, 1.01; 95% CI, 0.81 to 1.25). There also was no evidence of a dose-response trend with increasing lifetime cumulative solvent exposure or with increasing average solvent exposure dose (P = 0.26 and 0.65 for trend, respectively; Table 4). The negative results pertained to both men and women (data not shown). Only 14% of the solvent-exposed subjects were women, however, so the precision of female-specific data was limited. Furthermore, neither duration (years) nor exposure frequency (days per month) showed any dose-dependent relationships with risk of CRF (data not shown).

Table 4: OR and 95% CI for CRF associated with solvent exposure dose^a

No significant association between solvent exposure and CRF was seen when subjects who reported exposure above the OEL only were regarded as exposed (OR, 1.29; 95% CI, 0.83 to 1.99). Increasing cumulative exposure above the OEL was not associated with an increasing risk of CRF (P = 0.13, for trend).

Table 5 presents the OR for subgroup-specific types of CRF associated with lifetime cumulative solvent exposure. The OR for “ever exposed” were close to unity for CRF associated with glomerulonephritis, diabetic nephropathy, renal vascular disease, and other renal diagnoses. Also, there was no clear trend in risk across dose levels of solvent exposure in any of the diagnostic groups. Solvent exposure was evenly distributed between our control subjects who reported diabetes or hypertension and control subjects who did not report these conditions (data not shown).

Table 5: Adjusted OR and 95% CI for various classes of CRF associated with lifetime cumulative solvent exposure dose^a

Analyses that disregarded the most recent 10 yr revealed a similar pattern of no association between solvent exposure and CRF. The number of solvent-exposed subjects in these analyses decreased by 8% among both case patients and control subjects (data not shown). Similarly, in analyses that took into account only the most recent 10 yr of exposure, we found no increased risk of CRF associated with any measure of solvent exposure. Only 44% and 47% of the solvent-exposed case patients and control subjects, respectively, were included in these analyses. Stratified analyses confined to the socioeconomic classes “manual workers” and “nonmanual workers” also showed the same absence of association between solvent exposure and risk of CRF (data not shown).

Discussion

In this large, population-based, case-control study, exposure to organic solvents was unrelated to risk of early-stage CRF of all diagnostic types. Our results are at odds with the findings in several previous case-control studies (^{2–6,8–11,15,23}) but consistent with the absence of association between solvent exposure and idiopathic chronic glomerulonephritis reported from a population-based study by Asal et al. (²⁴). Moreover, in a carefully designed study by Stengel et al. (¹²), no overall relationship was observed between solvent exposure and histologically well-defined primary chronic glomerulonephritis, although heavily exposed male patients admittedly had an excess risk.

An important strength of our study is the well-defined study base, which permitted strict random sampling of population control subjects and the comprehensive organization for case finding, which ensured identification of all new case patients occurring in the same study base. In contrast, several of the previous studies of organic solvents and renal disease were hospital based (^{2,3,6,9,10,12,25}) or had control selection procedures that did not guard against influence of solvent exposure status or could not ensure that cases and control subjects represented the same study base (²⁶). Moreover, as opposed to most previous studies, which concerned patients with ESRD (^2,4,8,10), we defined incident cases as those who permanently passed a predefined serum creatinine level—a level sufficiently high to avoid the possible bias linked to detection of clinically silent disease (comorbidity or other circumstances leading to a higher probability for serum creatinine testing) yet sufficiently low to ensure that most case patients had not reached ESRD. Not only did we thereby avoid a long disease period that could potentially have affected the case patients’ self-reports of previous occupational exposures, but we also minimized selective loss of patients with rapid disease progression and early death, in comparison with the assessment in studies on ESRD patients. Moreover, we avoided some of the ambiguities associated with the distinction between incident and prevalent cases with ESRD.

Nonparticipation might conceivably be related to solvent exposure. This may introduce selection bias, particularly if participation rates differ among cases and control subjects. Our participation rates, however, were relatively high, and the difference between patients and control subjects was not large enough to explain our findings.

Retrospective exposure assessment is a potential weakness of case-control studies. Severe nondifferential misclassification, i.e., poor recollection of previous occupational exposures among cases and control subjects, could theoretically turn a true association into a null result (²⁷). In our study, detailed data about occupations, specific tasks associated with exposure, and exposure information obtained in an initial interview and a follow-up interview were evaluated by an expert occupational hygienist. This “expert rating method” is considered to be a valid approach in the context of a population-based study, and excellent reliability was noted in an evaluation (¹⁹). Our own test of intraobserver reliability also indicated excellent agreement. We believe that the thorough and individual assessment of solvent exposure reduced the potential for nondifferential exposure misclassification. Exposure assessment has varied in the previous studies. In the very first investigation published, the exposure scores used were poorly defined (²). The exposure score developed by Ravnskov et al. (³) and later used in several studies (^{6,9,10,15,23,25}) represents an improvement, but the assessment scale with exposure intensity factors according to 17 predefined occupational activities was crude and may have led to varying degrees of misclassification.

Recall bias is a very unlikely explanation for our negative findings. To negate a true positive association between solvent exposure and CRF risk, our case patients would have had to systematically underreport their exposure, or our occupational hygienist would have had to systematically underestimate the exposure of the cases. Although this seems remote, it is conceivable that recall bias may have contributed to spurious associations in previous positive case-control studies.

Likewise, negative confounding cannot readily explain our null results. It is possible, though, that individuals with conditions known to predispose for CRF were advised to refrain from organic solvent exposure or to choose different occupations, following the reports from earlier studies. Such negative confounding by underlying disease, however, could not explain the absence of association with CRF as a result of glomerulonephritis, because this disorder is not preceded by any known predisposing disease.

Previous case-control studies have primarily been conducted among patients with glomerulonephritis despite that most animal experiments have shown that solvent exposure leads to renal tubular damage (^28,29). The only previous study of patients with CRF regardless of underlying disease (¹¹) showed no more than a weak and statistically nonsignificant association with solvents overall. Exposure to the solvent subclass oxygenated hydrocarbons was admittedly associated with a fivefold increase in risk, but the CI was wide and no dose-response relation could be demonstrated. Yaqoob et al. (¹⁵) showed a significantly greater exposure to organic solvents among patients with insulin-dependent diabetes and diabetic nephropathy as compared with those without diabetic renal disease.

Most organic solvents consist of a mixture of several different compounds, and organic solvent exposure often entails exposure to several different mixtures and compounds. We chose to assess organic solvent exposure overall to avoid the uncertainties of retrospective assessment of the different subclasses of organic solvents. Accordingly, any associations between possibly more nephrotoxic subclasses of solvents and CRF may have remained undetected. Although our patients are possibly not as heavily exposed to solvents as patients in previous studies, our analysis restricted to the subjects with heaviest exposure also showed no associations. A cohort study with heavily exposed subjects and prospective exposure assessment admittedly would be a more suitable design to investigate the possible effect of high-level exposure of certain subclasses of solvents on renal disease. A population-based case-control study, however, is the preferred design of a study of risk factors for a rare disease and will provide a better picture of the impact of exposure on the overall incidence of CRF in the population.

Organic solvents conceivably could act as initiator or promoter of chronic renal disease, or both. This study was designed to investigate the association between solvent exposure and CRF in general. An initiating effect on rare and specific subtypes of renal disease may have remained undetected because of possible diagnostic misclassification or inadequate statistical power. Our results, however, argue against a promoting effect of any importance, and the etiologic fraction of solvent exposure on CRF, if at all, would be small. Furthermore, any major links between solvent exposure and CRF among patients with a clinical diagnosis of either glomerulonephritis or diabetic nephropathy likely would have been detected, considering the strong associations reported in several previous studies (^3,9,10,15).

In conclusion, the results from our study, designed to overcome many of the weaknesses noted in previous investigations, from a public health perspective, do not support an adverse effect of organic solvents on the development or progression of CRF in general. Detrimental renal effects from high exposure to certain subclasses of solvents or on specific renal diseases cannot be ruled out.

This study was financed by a grant from the Swedish Council for Work Life Research and by the International Epidemiology Institute, Rockville, Maryland.

We are indebted to Birgit Rousk, the occupational hygienist assistant who performed the additional telephone interviews with precision and care.