METHODS

There are 30 deaths in white men recorded from the Kansas City plant in the mortality surveillance system mortality registry for whom the underlying cause of death was NMRDxIP. These 30 deaths, all of whom were long term employees with 10 or more years employment, constitute the case series. The specific causes of death with the 9th revision of the international classification of diseases (ICD-9) codes for all 30 cases are shown in table 1.³ Twenty two of the 30 cases were concentrated in ICD code 496 (chronic airways obstruction: not elsewhere classified). The rest were: four cases in ICD 492 (emphysema), three cases in ICD 491 (chronic bronchitis), and one case in ICD 514 (pulmonary congestion and hypostasis).

Controls also were drawn from deaths at Kansas City in the mortality surveillance system mortality registry and were matched to cases on year of birth (±2 years) and survival to death (±2 years) or to the end of 1996 to allow controls the opportunity to survive as long as cases. Any death without NMRDxIP or malignant respiratory cancer as the underlying cause of death was eligible to be a control. All eligible controls who met the matching criteria were used resulting in a total of 133 matched controls. There were 18 cases with four controls each, seven cases with three controls each, and five cases with two controls each. Although not included in the matching criteria, the distributions by age when first joining the company and age at death were comparable for cases and controls.

All of the cases were employed for 10 or more years and nearly two thirds were employed for 20 or more years. All but four of the controls were employed for more than 10 years.

The procedure for enumerating exposures and measuring exposure concentrations has been described previously.^2,4 Briefly, an expert exposure assessment committee developed quantitative exposure estimates for workplace substances known or suspected to be respiratory irritants in each manufacturing and non-manufacturing process. Each process was assigned one of four estimated potential exposure concentrations (8 hour time weighed averages) specific for calendar time. This procedure resulted in estimates of the daily exposure to each of the substances over a worker's entire working lifetime. Cumulative exposure to each substance was developed for each employee as the product of the number of days in a process multiplied by the exposure concentration for that process and then summed over all processes and expressed as cumulative exposure. Table 2 gives the number of cases and controls by cumulative exposure-days (concentration multiplied by days of exposure) for each of the exposures in the analysis as well as the numbers for the sociodemographic variables obtained from interviews.

Because the number of cases is limited, we used an exact conditional logistic regression procedure with the LogXact for Windows software package.⁵ Matched, unadjusted odds ratios (ORs) were used to find whether there was an association between NMRDxIP and cumulative exposure history or sociodemographic factors one variable at a time (univariate analysis). Matched, adjusted ORs were used to estimate the effect of any one variable while controlling for the effect of all the others. To carry out the adjusted analysis, we used criteria suggested by Siemiatycki et al and entered both sociodemographic and exposure variables from the univariate analysis with ORs of 1.5 or greater or 0.67 or less without regard to significance.⁶ All tests of significance were two sided and carried out at the α=0.05 level.

RESULTS

Matched, unadjusted NMRDxIP OR and 95% confidence intervals (95% CIs) for individual sociodemographic and exposure variables are given in table 3. This table also gives the number of observations (cases and controls combined) along with the number of informative strata in producing each OR for the reader to evaluate the extent to which data were complete for any variable. Analyses are based on those cases and controls with available data (table 2). In this analysis, there was no significant relation with any of the exposure variables. There was an increased OR for the lower exposure concentration for respirable glass fibres (OR 2.24; 95% CI 0.54 to 10.97) but an OR less than one for the highest exposure (OR 0.79; 95% CI 0.20 to 3.52). For respirable silica there was no consistent relation: the OR increased through the first three exposure concentrations but decreased for the highest. The ORs although not significant, were greater than unity for all exposure concentrations of respirable silica. The OR for any exposure to talc was below unity but not significant (OR 0.66; 95% CI 0.24 to 1.64). The ORs for exposure to both asbestos and formaldehyde were not significant and close to unity for the exposure concentrations analyzed.

As with exposure variables, there were no significant increases or deficits in the univariate analysis for any of the sociodemographic variables including smoking. However, the smoking OR was substantially increased (OR 5.09; 95% CI 0.65 to undetermined). The OR for drinking was increased for regular drinkers (OR 1.42; 95% CI 0.36 to 5.63). The ORs were inversely related to level of income and directly related to education.

Table 4 shows the results of a matched analysis using a conditional logistic regression procedure which was carried out to estimate the effect of any one variable while controlling for the effect of all the others. As already described, variables giving an OR ≥1.5 or ≤0.67 for sociodemographic or exposure variables in the unadjusted (univariate) analysis were included in the model. Six variables met the inclusion criteria (income was excluded because of the scarcity of data). Note that the number of observations was only 68 and the number of informative strata was reduced to 19. As a result, these results should be interpreted with caution.

There were no significant variables in this analysis when all variables were simultaneously adjusted. ORs for respirable glass fibres are below unity at all exposures. The ORs for silica were not significant. These ORs are above unity at all exposure concentrations but there was no evidence of a dose-response relation. Adjusted ORs from conditional logistic regression models which included only smoking and exposure variables were also examined and yielded very similar results.

DISCUSSION AND SUMMARY

A search of the Medline database suggests that publications on the relation of NMRDxIP to occupational exposures is sparse and inconclusive especially for chronic airways obstruction, the predominant cause of death for cases in this case control study. Publications on the relation between NMRDxIP and occupational exposures in the fibreglass manufacturing industry is also sparse, especially when compared with that of malignant respiratory disease, and there is no evidence suggesting an association between NMRDxIP and respirable fibres. Marsh et al in an industrywide historical cohort mortality study which included the Kansas City plant state that “analyses by duration of employment, small diameter fiber exposure and cumulative and average intensity of fiber exposure revealed no evidence of a relationship between MMMF exposure and nonmalignant respiratory disease.”⁷ In the paper by Marsh et al, the standardised mortality ratio (SMR) for NMRDxIP for the Kansas City plant was only slightly increased (SMR 114) and not significant.⁷ No excess mortality from non-malignant respiratory diseases was found in the International Agency for Research on Cancer historical cohort study of European manmade mineral fibre production workers.⁸ A study of Owens Corning's Newark plant found that smoking was the only significant factor for NMRD although the ORs for respirable fibres and silica were greater than unity for the highest exposure categories.²

This study of the Kansas City plant did not show any relation between respirable glass fibres and NMRDxIP. The ORs for respirable glass fibres were below unity when adjusting for smoking and other sociodemographic variables as well as other exposure variables. Even in the univariate analysis, the OR for the highest exposure concentration for respirable fibres was less than unity. On the other hand, the ORs for exposures to silica were all above unity in the adjusted analysis and in the univariate analysis. However, none of the ORs were significant and there was no clear dose-response relation.

It should be pointed out that the exposure concentrations for all substances considered were low. Further, given the number of cases and controls, the statistical power to detect relatively small increases in risk, if any increase truly existed, was relatively low. The study would have required three times as many cases to have an 80% chance of detecting a true doubling in risk in the lowest exposure category. Overall, the findings for Kansas City did not show an association between respirable glass fibres and NMRDxIP. The adjusted ORs for all exposures of respirable fibres were less than unity. The OR for the lower exposures in the unadjusted analysis was increased although not significantly. On the other hand, the OR for the higher exposures was below unity. Obviously a small study such as this cannot provide any definitive answers on the question about any relation between respirable glass fibres and NMRDxIP. On the other hand, given the paucity of literature on this subject, the current study adds some reassurance that no overwhelming risk is present.

The ORs for exposure to silica were all above unity in both the unadjusted and adjusted analyses although there was no clear dose-response relation and none of the ORs were significant. However, these raised ORs for silica suggest that continued surveillance would be prudent.