09.02.2014 Views

home edit2 whole TSD November 2002 PDF format - OEHHA

home edit2 whole TSD November 2002 PDF format - OEHHA

home edit2 whole TSD November 2002 PDF format - OEHHA

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

and can distort the measure of effect (e.g., relative risk or odds ratio) (Rothman, 1986). With respect<br />

to exposure assessment, statistically significant pooled estimates of elevated risk lacking evidence of<br />

heterogeneity were identified in several occupational subgroup analyses, both with and without<br />

additional stratification for smoking. Prior to stratifying by adjustment for smoking, the occupational<br />

subgroups involving trucking (pooled RR = 1.47, 95% C.I. = 1.33-1.63), the railroad industry<br />

(random-effects pooled RR = 1.45, 95% C.I. = 1.08-1.93), mechanics and garage workers (randomeffects<br />

pooled RR = 1.35 (95% C.I. = 1.03-1.78), general transportation and professional drivers<br />

(random-effects pooled RR = 1.45, 95% C.I. = 1.31-1.60) gave risk estimates greater than the overall<br />

pooled risk estimate. The pooled RR estimates for trucking and general transportation and professional<br />

drivers showed little to no evidence of heterogeneity; however, estimates for the railroad industry<br />

demonstrated considerable heterogeneity (Q statistic = 30.90, p < 0.001).<br />

Further stratification of the occupational subgroup analysis by adjustment for smoking produced a large<br />

impact on the pooled risk estimates, with all smoking-adjusted subgroup estimates displaying little<br />

evidence of heterogeneity and leading to increased risk estimates in all but one of the occupational<br />

categories. Pooled risk estimates by occupation in smoking-adjusted studies showed little evidence of<br />

heterogeneity for several occupations under both models, including truck drivers (random-effects<br />

pooled RR = 1.53, 95% C.I. = 1.20-1.94), railroad workers (random-effects pooled RR = 1.68, 95%<br />

C.I. = 1.28-2.19), and diesel mechanics and garage workers (random-effects pooled RR = 1.25, 95%<br />

C.I. = 0.87-1.80). The pooled estimates for the heavy equipment operators and dock workers and for<br />

the railroad industry studies adjusting for smoking displayed the most dramatic changes relative to the<br />

occupational analysis without smoking stratification. Among the former subgroup, the pooled risk<br />

estimate changed from 1.28 (random-effects model, 95% C.I. = 0.99-1.66) to 2.43 (95% C.I. = 1.21-<br />

4.88). Among the railroad industry studies, the pooled risk estimate also increased substantially (from<br />

1.45 to 1.68, 95% C.I. = 1.28-2.19). In both subgroups, the pooled smoking-adjusted estimates<br />

showed little evidence of heterogeneity, though these estimates were based on two studies in the former<br />

instance and three in the latter. However, the other two heavy equipment operator and dock worker<br />

studies and the other three railroad industry studies that were not adjusted for smoking still displayed<br />

evidence of heterogeneity (Q-statistics = 2.933, 1 d.f., p = 0.09, and 21.517, 2 d.f., p < 0.001,<br />

respectively).<br />

The meta-analysis also identified evidence of exposure-response relationships in the subgroup analyses<br />

based on duration of employment. However, as noted in <strong>OEHHA</strong> (1998), this analysis was hampered<br />

by the absence of duration-specific risk estimates in approximately one-half the studies. While the initial<br />

analysis conducted on all the included studies resulted in elevated pooled risk estimates for strata in<br />

which exposure durations were greater than 10 years relative to those with less than 10 years of<br />

exposure or for which the exposure durations were not clear from the published reports, there was still<br />

significant evidence of heterogeneity for several of the duration strata. In contrast, the analysis utilizing<br />

only estimates from the smoking-adjusted studies showed some evidence of an exposure-response<br />

gradient without evidence of statistical heterogeneity. The summary risks for all three exposure-duration<br />

strata were: RR = 1.39 (95% C.I. 1.19-1.63) for < 10 years (based on three estimates from two<br />

studies), RR = 1.64 (95% C.I. = 1.40-1.93) for 10 < to < 20 years (11 estimates from 6 studies), and<br />

RR = 1.64 (95% C.I. = 1.26-2.14) for > 20 years (four estimates from four studies). The pooled risk<br />

453

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!