Jump to comment:

• Meta-analysis on benzene exposure and non Hodgkin lymphoma
  Gerard M H Swaen
  Published on: 27 April 2016
• Response to Swaen et al.
  Craig Steinmaus
  Published on: 27 April 2016

• Published on: 27 April 2016

  Meta-analysis on benzene exposure and non Hodgkin lymphoma

  • Gerard M H Swaen
  • Other Contributors:
    • Shan P. Tsai and Carol Burns

  Meta-analysis on benzene exposure and non Hodgkin lymphoma

  June 30 2009

  Gerard M H Swaen1 Shan P. Tsai2 Carol Burns1

  1 Department of Epidemiology, The Dow Chemical Company, Midland, Michigan USA. 2 Shell Health, Shell Oil Company, Houston Texas, USA.

  Corresponding Author: Gerard M H Swaen, The Dow Chemical Company, P.O. Box 444, 4530 AK Terneuzen, The Netherlands 31-(0)43-3626042. E-...

  Show More

  Meta-analysis on benzene exposure and non Hodgkin lymphoma

  June 30 2009

  Gerard M H Swaen1 Shan P. Tsai2 Carol Burns1

  1 Department of Epidemiology, The Dow Chemical Company, Midland, Michigan USA. 2 Shell Health, Shell Oil Company, Houston Texas, USA.

  Corresponding Author: Gerard M H Swaen, The Dow Chemical Company, P.O. Box 444, 4530 AK Terneuzen, The Netherlands 31-(0)43-3626042. E- mail: gswaen@dow.com

  Key words: benzene, non Hodgkin lymphoma, epidemiology, meta- analysis, occupation, risk

  Sir, A recent meta-analysis on benzene exposure and non Hodgkin lymphoma (NHL) concluded that the reviewed epidemiological studies provide “new evidence that benzene causes NHL.” 1 The meta-analysis conducted by Steinmaus et al differed from the others in several aspects. Firstly, it selected subgroups with the highest putative exposure to avoid dilution. Secondly, cohort studies were adjusted for the Healthy Worker Effect (HWE) by considering the NHL deaths as cases and all other deaths as controls. A HWE for cancer endpoints is small if any, compared to non cancer endpoints.2. The results from a meta-analysis of 150 occupational cohort studies suggest that HWE adjustment may artificially inflate the true RR for cancer endpoints. 3 Thirdly, outdated cohort study results were used instead of more recent updates. Fourthly, Steinmaus et al did not consistently apply their own selection criteria. We have re-analyzed the Steinmaus et al meta-analysis by precisely applying their inclusion and exclusion criteria. These include selecting the highest exposure category relative risk (RR), selecting cancer incidence versus mortality data and we used the most recent results from cohort updates whenever available. To maintain consistency, we adjusted the RR estimates for the HWE by using the all cancer SMR of the selected sub cohort instead of the all mortality SMR. Herein we present only a summary table of results. Additional comparative tables are available upon request.

  There were 16 case-control studies included in the Steinmaus et al. review. We excluded the study by Dryver since it provided no odds ratio (OR) for benzene exclusively. For the Fabbro-Peray et al. study, Steinmaus et al selected the OR of those with over 810 days of exposure. However, we used the subcategory with exposure duration of over 15 years since it represents the group with longer exposure. For the study by Schnatter et al the OR for the highest intensity was 0 and Steinmaus et al selected the OR for the second highest exposure intensity group. We combined the highest and the second highest exposure intensity groups (OR=0.49 95% CI: 0.10-2.32). The revised meta-OR for the 15 case-control studies was 1.17 (95% CI 0.96 – 1.44) (see Table 1).

  For the six non refinery cohort studies, we observed inconsistencies in the authors’ use of the selection sequence. For the Bloemen study, the longest duration RR was used by Steinmaus et al. although RRs by cumulative exposure were given. We combined the two cumulative exposure categories into one category of 28.3+ ppm-years (SMR=0.63 95% CI: 0.08- 2.28). For the Rinsky study, we selected the RR for men (1.00), assuming that the men were higher exposed, since the jobs with benzene exposure were generally done by men. Lastly, for Wong et al. the RR for highest cumulative exposure was replaced with that for highest intensity exposure. The revised meta-RR for these non-refinery studies was 1.13 (95% CI: 0.80- 1.61) and a meta-RR of 1.05 (95% CI: 0.75-1.47) after adjusting for the HWE.

  In the meta-analysis of the refinery studies, as per the recommendations of Steinmaus et al., cancer incidence data were selected if both incidence and mortality were reported. There were two refineries included in the 1993 study published by Tsai, et al, and the combined results were used instead of selecting only one of the refineries. Additionally, we used the combined results from maintenance and process workers for the study conducted at the Beaumont refinery by Wong since it is unclear that process workers were more exposed. We selected results based on the longer duration of work for the study by Pukkala et al, following the Steinmaus et al. hierarchy. We used the updated data for the following refinery studies (Sorahan, Thomas and Tsai et al, 1996) 4 5 6. In the update by Tsai and colleagues NHL was listed as one of the specific disease categories, replacing reticulosarcoma used in the 1996 paper. We selected the updated RR although the longest duration of employment was shorter. 7 The revised meta-RR for the 21 refinery studies was 1.00 (95% CI: 0.89-1.12) and 1.06 (95% CI: 0.95-1.18) after adjusting for the HWE. All the models gave similar results and we only present those of the fixed model.

  All of the meta-RRs calculated in the re-analysis were less than those reported by Steinmaus et al and most were close to unity. The only meta-RR that remained statistically significant was that for the high exposure non refinery studies, consisting of four cohort studies and nine case-control studies (meta-RR of 1.33, 95% CI: 1.02-1.74), for which exposure was largely based on self-reported data. The statistical significance of this meta-RR was lost after adjustment for the heterogeneity effect (95% CI: 0.99-1.80).

  The re-analysis shows that the differences of the applied analytical methods and relative risks selected can vastly alter the overall meta- relative risk. The re-analyzed evidence provides little if any support for an association between benzene exposure and NHL.

  Gerard M H Swaen1, Shan P. Tsai2 Carol Burns1. 1 Department of Epidemiology, The Dow Chemical Company, Midland, Michigan USA.

  2 Shell Health, Shell Oil Company, Houston Texas, USA.

  Table 1 Summary results of the meta-analysis on benzene exposure and NHL and for refinery workers and NHL, comparing the results by Steinmaus et al to our findings (fixed models, other models gave similar results)).

  N RR as in Steinmaus RR in the re-analysis Benzene and NHL

  All studies 22 1.22 (1.03-1.46) 1.16 (0.97-1.38)

  Case control studies 16 1.23 (1.00-1.50) 1.17 (0.96-1.44)

  Cohort studies 6 1.21 (0.86-1.71) 1.13 (0.80-1.61)

  All high exposure studies 13 1.49 (1.15-1.92) 1.33 (1.02-1.74)1

  HWE adjusted cohort studies 6 1.22 (0.89-1.67 1.05 (0.75-1.47) Refinery work and NHL

  All studies 21 1.21 (1.06-1.38) 1.00 (0.89-1.12)

  High exposure studies 14 1.30 (1.04-1.62) 1.07 (0.87-1.31)

  HWE adjusted all studies 21 1.42 (1.25-1.62) 1.06 (0.95-1.18)

  HWE adjusted high exposure studies 14 1.51 (1.22-1.88) 1.20 (0.99- 1.47) 1 statistical significance was lost after adjustment for heterogeneity (95% CI: 0.99-1.80)

  References

  1. Steinmaus C, Smith AH, Jones RM, Smith MT. Meta-analysis of benzene exposure and non-Hodgkin lymphoma: biases could mask an important association. Occup Environ Med 2008;65(6):371-8.

  2. Li CY, Sung FC. A review of the healthy worker effect in occupational epidemiology. Occup Med (Lond) 1999;49(4):225-9.

  3. Greenberg RS, Mandel JS, Pastides H, Britton NL, Rudenko L, Starr TB. A meta-analysis of cohort studies describing mortality and cancer incidence among chemical workers in the United States and western Europe. Epidemiology 2001;12(6):727-40.

  4. Sorahan T. Mortality of UK oil refinery and petroleum distribution workers, 1951-2003. Occup Med (Lond) 2007;57(3):177-85.

  5. Dement JM, Hensley L, Kieding S, Lipscomb H. Proportionate mortality among union members employed at three Texas refineries. Am J Ind Med 1998;33(4):327-40.>

  6. Tsai SP, Ahmed FS, Wendt JK, Foster DE, Donnelly RP, Strawmyer TR. A 56-year mortality follow-up of Texas petroleum refinery and chemical employees, 1948-2003. J Occup Environ Med 2007;49(5):557-67.

  7. Tsai SP, Chen VW, Fox EE, Wendt JK, Cheng Wu X, Foster DE, et al. Cancer incidence among refinery and petrochemical employees in Louisiana, 1983-1999. Ann Epidemiol 2004;14(9):722-30.

  Licence Statement: The Corresponding author has the right to grant on behalf of all authors and does grant on behalf of all authors, an exclusive licence on a worldwide basis to the BMJ Publishing Group Ltd to permit this article (if accepted) to be published in OEM and any other BMJPGL products and sublicences such use and exploit all subsidiary rights, as set out in our licence

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top
• Published on: 27 April 2016

  Response to Swaen et al.

  • Craig Steinmaus
  • Other Contributors:
    • Allan H. Smith and Martyn T. Smith

  Response to Swaen et al.

  Authors Craig Steinmaus1, 2 Allan H Smith1 Martyn T Smith1

  Authors affiliations 1. School of Public Health, University of California, Berkeley, University Hall, Berkeley, CA 94720-7360, USA 2. Office of Environmental Health Hazard Assessment, California Environmental Protection Agency, 1515 Clay St., Oakland, CA 94612

  Corresponding author Craig Steinmaus, School of Publ...

  Show More

  Response to Swaen et al.

  Authors Craig Steinmaus1, 2 Allan H Smith1 Martyn T Smith1

  Authors affiliations 1. School of Public Health, University of California, Berkeley, University Hall, Berkeley, CA 94720-7360, USA 2. Office of Environmental Health Hazard Assessment, California Environmental Protection Agency, 1515 Clay St., Oakland, CA 94612

  Corresponding author Craig Steinmaus, School of Public Health, University of California, Berkeley, University Hall, Berkeley, CA 94720-7360, USA e-mail: craigs@berkeley.edu

  Sir, The differences in methods between our meta-analysis and that of Swaen et al. did not “vastly alter” our most important finding or change our overall conclusion.1,2 Our major conclusion was that evidence of an association between benzene and non-Hodgkin lymphoma (NHL) could be found when analyses were focused on results from highly exposed workers and were appropriately adjusted for the healthy worker effect. In their analysis of high exposure benzene studies, Swaen et al. implemented a series of differences in the data used from the studies compared to our own. But despite the fact that every single one of their major changes resulted in a lowering of the summary relative risk, Swaen et al. also found a statistically significant elevated relative risk in the high exposure studies (RR = 1.33 (95% CI, 1.02-1.74)), just as we did. Even after they adjusted for heterogeneity, the relative risk remained statistically significant if correctly estimated. The one-sided p-value after this adjustment can be estimated from their confidence interval to be 0.03. Since our a priori hypothesis was that benzene increases, not decreases, the risk of NHL, a one-tailed p-value is clearly most appropriate here.

  The Swaen et al. use of the all-cancer standardized mortality ratio (SMR) to adjust for the healthy worker effect assumes that benzene or any of the other agents that the benzene-exposed workers are exposed to will not cause any other cancer type. However, benzene is a well known cause of leukemia and risks of this cancer were elevated in many of the studies used in our meta-analyses. Relative risks of other cancers such as lung cancer and mesothelioma were also elevated in some of the studies, probably due to co-exposures like asbestos that can commonly occur with benzene. Because of these increases in other cancers, the use of the all- cancer SMR probably biased the Swaen et al. summary relative risks towards the null. Without this bias, the Swaen et al. summary relative risks would have been higher than they reported and even more similar to our results than they already are. In any case, as we noted above, using different data and different methods, Swaen et al. have produced additional evidence that benzene increases the overall risk of non-Hodgkin lymphoma in the available studies to date in a manner which is unlikely to be due to chance.

  References 1. Steinmaus C, Smith AH, Jones RM, Smith MT. Meta-analysis of benzene exposure and non-Hodgkin lymphoma: biases could mask an important association. Occup Environ Med 2008; 65:371-8. 2. Swaen GMH, Tsai SP, Burns C. Meta-analysis on benzene exposure and non Hodgkin lymphoma (Letter to Editor). Occup Environ Med:TBD

  Copyright The Corresponding Author has the right to grant on behalf of all authors and does grant on behalf of all authors, an exclusive license (or non- exclusive for government employees) on a worldwide basis to the BMJ Publishing Group Ltd and its Licensees to permit this article to be published in Occupational and Environmental Medicine editions and any other BMJPGL products to exploit all subsidiary rights, as set out in our license.

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top