Errors in TCE Analysis.

We read with interest the epidemiologic overview and evaluation of trichloroethylene (TCE) and cancer recently published by Wartenberg et al. (1). Unfortunately, there were a number of errors in Table 5 (1) concerning risks reported in our study of workers exposed to TCE during the manufacturing of aircraft (2). For example, the standardized mortality ratio (SMR) and the 95% confidence interval (CI) presented for multiple myeloma were actually those for Hodgkin's disease. There were also serious errors in the listed SMR and CI for stomach cancer as well as discrepancies in the presentation of CIs for 10 other sites (Table 1). These incorrect values were included in the meta-analyses to make inferences regarding the carcinogenic potential of TCE; thus, they should be revised.

Table 1. Discrepancies between TCE and cancer
risks as reported by Boice et al. (2) and as listed
in Table 5 of the recent review by Wartenberg et
al. (1).

                      Boice      Wartenberg
Cancer site         et al. (2)   et al. (1)

Multiple myeloma
  SMR (no. cases)    0.9 (6)      2.8 (4)
  95% CI            (0.3-2.0)    (1.1-7.1)
Stomach
  SMR (no. cases)    1.3 (17)     0.8 (7)
  95% CI            (0.8-2.1)    (0.4-1.7)
Cervix
  SMR (no. cases)    0.0 (0)         --
  95% CI            (0.0-5.5)        --
Leukemia
  SMR (no. cases)    1.1 (12)     1.0 (12)
  95% CI            (0.5-1.8)    (0.6-1.8)
Breast
  95% CI             0.5-2.7      0.6-2.7
Buccal
  95% CI             0.2-1.4      0.3-1.4
Colon
  95% CI             0.7-1.5      0.8-1.5
Esophagus
  95% CI             0.3-1.7      0.4-1.7
Hodgkin's disease
  95% CI             0.8-7.1      1.1-7.1
Kidney
  95% CI             0.4-2.0      0.5-2.0
Larynx
  95% CI             0.3-2.8      0.4-2.8
Rectum
  95% CI             0.6-2.5      0.7-2.5

We also found it peculiar that the reported CIs from our study were recalculated. We computed exact CIs, which are methodologically superior to, the recalculated ones. The method of recomputation assumed that our upper but not lower confidence limit was correct. These recomputations had the unusual property of exaggerating the lower CI, thus making the results appear more statistically significant (or closer to statistical significance) than they actual were.

John D. Boice, Jr.
Joseph K. McLaughlin
International Epidemiology Institute
Rockville, Maryland
E-mail: [e-mail]

REFERENCES AND NOTES

(1.) Wartenberg D, Reyner D, Scott CS. Trichloroethylene and cancer: epidemiologic evidence. Environ Health Perspect 108(suppl 2):161-176 (2000).

(2.) Boice JD Jr, Marano DE, Fryzek JP, Sadler CJ, McLaughlin JK. Mortality among aircraft manufacturing workers. Occup Environ Med 56:581-597 (1999).