Document emVMXO2woO5e9YerMJBdwe75q

-.j:4i4HiiSiM4 i; hU -4 f -i-W fc W Rt 336 Letters to the editors Registrar General for Scotland (1996-2000) Annual Reports. General Register Office for Scotland: Edinburgh. . Selikof U, Seidman H. Cancer risk of insulation workers in the United States, In: Pogovski P, Gilson JC, Timbrell V and Wagner JC, editors. Biological effects of asbestos. IARC Scientific Publication No. 8. Lyon: International Agency for Research on Cancer; 1973. p. 209-16. Sluis-Cremer GK, Liddell FDK, Logan WPD, Bezuidenhuil BN. The mortality of amphibole miners in South Africa, 1946-80. Br J Ind Med 1992;49:566-75. Crown Copyright 2001 ' "" Published by Elsevier Science Hd on behalf of British Occupational Hygiene Sorirry 1 AH rtglm reserved 3 . Printed in Greet Britain. - . OOttS4S7Sit)1tIIOjOO 7. PH: 50003-4878(01)00032-1 We thank Drs Liddell and Browne and Mr Howie for their interest in our paper (Hodgson and Darnton, 2000), and their generally supportive comments. Between them they make a number of points which we would like to respond to. We will first deal with the comments from Drs Liddell and Browne. CAROLINA COHORT Both correspondents challenge in one way or another our treatment of the Carolina cohort It is clear that there are difficulties in interpreting the observations from this cohort and a number of approaches are possible. As far as the assessment of mesothelioma risk is concerned, whether one counts the Carolina cohort along with the other chrysolite cohorts or not makes very little difference to the over all risk estimate. With only two mesothelioma cases, the Carolina cohort's statistical weight here is small. Our suggested risk is 0.001% per f/ml.yr, implying that the Carolina observations (0.013) do indeed over state the true risk (whether by chance or otherwise). Dr Browne thinks we should have made more of the possible role of amphibole in the Carolina cohort. It is true that this is a possible explanation for this cohort's "exceptional" status, though as we point out, the fact that significant levels of amphibole were found post mortem in the lungs of only one of ten lung cancer cases examined (Green et al., 1997) lim its the confidence one might otherwise have in this explanation. COHORT AVERAGES Another general criticism focuses on our use of what we have called "cohort average" data, rather than using (where available) more detailed exposure specific breakdowns within cohorts. Our reasons for this are discussed in the paper, and amount to a judge ment that more is gained by improving the stability of observations `and avoiding the probable errors of Received 6 March'2001. individual exposure estimation than is lost by pooling observations on individuals with very different exposures. Where both cohort average and internal analyses are available, we compare the two measures (figures 10 and 11 of our paper) and show a good degree of concordance between them. It is not true that "regression between cohorts is irrelevant to [the form of the individual dose response]'1. For a Unear relationship the two approaches will give identical estimates. It is of course true that non-linearity will distort the correspondence between individual and cohort level measures but the extent of this distortion for the degree of non-linearity we suggest will be small. `j ;1 .. .j j ! ! . j . . SMOKING Dr Liddell also challenges our treatment of smok ing effects. Clearly the reality of the joint effect of asbestos and smoking is more complicated than impUed by a simple multiplicative adjustment for "smokers" and "non-smokers". However in the con text of other substantial uncertainties such refine ments are arguably inappropriate and in any case our paper was challenging the limits of reasonable length already. . TREMOUTE Dr Browne refers to evidence that the tremolite contamination in Quebec chrysotile is a major factor in the observed mortality there, coupled with the claim that levels of asbestos related disease are very low or non-existent in cohorts exposed to chrysotile uncontaminated by tremolite, in South Africa, the for mer USSR and Brazil. The evidence from Quebec that tremolite plays an important role in the observed mortality there is reasonably convincing. Unfortu nately there is no good epidemiology on the low tremolite cohorts elsewhere. The reference cited in this connection by Dr Browne (Rees e/a/., 1999) does show that in a sixteen month case-ascertainment per iod covering most of South Africa no cases of meso thelioma associated with chrysotile mining were i ; _: . Letters to the editor 337 detected (out of a total of 123 observed). This cer- I tainly suggests (in agreement with our conclusions) that the risk from chrysotile is low. However the much larger chrysotile mining operation in Quebec has only generated about one death from mesotheli oma per year, so the absence of South African cases in a sixteen month period does not establish that the risk there is materially different from that in Quebec J (tbough it certainly may be). .......... THRESHOLDS AND FORM OF DOSE-RESPONSE Dr Browne also challenges our description of argu. merits for a threshold based on identifying the lowest i dose received by an observed case as "logical nonj sense". This description was perhaps a touch intem: perate, but it does seem to us that--particularly if it is accepted (as most observers do) that there are some "spontaneous" mesothelioma cases not associated with asbestos exposure--that this datum (the lowest ' dose received by an observed case) on its own carries . no information about the level of any possible "thres hold". Dr Browne suggests that the term threshold should be identified with a level of exposure beiow which any excess mortality would be epidemiologically 1 detectable. Such a concept would be inadequate in | any regulatory context, since we are concerned with | lowering risks well below any such level. | We agree with Dr Browne that the Sluis Cremer | evidence for adverse effects of amphibole exposure 1 below 5 f/ml.yr is not compelling. Nevertheless we ! believe these data do show that there are effects well below the levels at which a threshold for chrysotile toxicity is arguable. As we point out, the adoption of a slightly concave dose response curve for lung can cer does mean that we predict less than proportionate risk at low levels of exposure. We could have extended our discussion of the "threshold" argument, and both Liddell et at's (1998) breakdown of Quebec lung cancer mortality by inten sity groups and Vacek (1998) more extensive model ling of the relationship to mortality of different func tional forms of exposure and exposure timing are valuable contributions to the literature. However the possible conclusions are twofold: either cumulative exposure is so misleading a metric that no quantitat. ive conclusions can be drawn, and we must wait for better analyses (which are very unlikely ever to be available); or we can cautiously see how far this mea sure can take us. The coherence of the picture < revealed by the cumulative dose approach I (particularly for mesothelioma) encourages us that it is adequate as a first approximation. CONSISTENCY WITH OTHER OBSERVATIONS : Dr Browne suggests that we should have "sub jected our calculations and guesses to the test of reality", and refers us to two authors Liddell (1993) and Camus et at. (1998) who have done so. Although we did not make these comparisons in the paper, we do not believe our conclusions are contradicted by the realities recorded, in either of these papers. The main conclusions of Camus et al. is that the EPA risk assessment overstates the lung cancer risk from chrysotile by at least a factor of ten. Our suggested risk coefficient is one tenth of the EPA's. It is clearly consistent with the observations of Camus et al.--as indeed a very wide range of estimates (down to zero) would be: we are at the limits of what can be resolved by epidemiological observation. Liddell (1993) argues for a "sub-linear" (concave) response for mesothelioma on the basis that the expected number of cases from person years gener ated by subjects with exposure durations below the minimum exposure duration for an observed case (in each cohort) is significantly greater than the (by definition) zero cases observed. This is an ingenious argument, but it is not clear that the simple observed to expected comparison it is based on can be tested statistically in the usual way, since the upper duration of exposure considered is defined by the minimum exposure duration seen among cases. Some sort of extreme value test is needed here. Furthermore, this analysis is based only on duration of employment, not on dose. LOW EXPOSURE RISK ESTIMATES Browne and Liddell take different views of our suggested risk estimates at low exposures- Dr Liddell proposes modifications to the risk estimates in our table 11, using linear models for lung cancer and for mesothelioma at high (above 50 f/ml.yrs) exposures, but retaining (on precautionary grounds) our sug gested convex (r = 0.75) non-linear relationship for mesothelioma at lower exposures. Dr Browne is much more critical, suggesting that our risk estimates "have no sound basis". Certainly these estimates are much less soundly based than one would wish. Some view does however need to be taken, and in presenting both range and point estimates we have sought to convey both a rea soned best estimate and the degree of uncertainty (which, of course, increases as doses fall below epide miologically observed levels). It is difficult to convey the balance between what is reasonably arguable in the way of risk estimation, and the associated uncer tainties. That is why we chose in table 11 to give verbal summaries and use heavily rounded figures, rather than displaying estimates graphically, or other wise implying inappropriate precision. Mr Howie's main concern is that the low overall mortality in some cohorts may have led to under esti mation of the true risk. It is true that overall mortality in several of the underlying cohorts is low, and conse quently that a substantial amount of asbestos related 338 Letters to the editors mortality will occur in these cohorts that does not agrees) substantially more potent both for mesotheli form part of the present observations. We were very oma and lung cancer then the higher mortality levels aware of this aspect of the data, and the problems it in manufacturing can (with die sole exception of the poses are discussed in detail in our Appendix A. Carolina lung cancer observations) readily be Briefly, we argue that the pattern of mortality over explained by differences of fibre mix. time revealed by those cohorts which' do have extended follow-up justifies the assumption that after JOHN HODGSON allowing for a 10 or 20 year latency period the rela and ANDREW DARNTON tive mortality excesses (for lung cancer relative to expected lung cancer, for mcsbthclioma relative to total expected mortality) remain broadly the same Epidemiology and Medical Statistics Unit, Health and Safety Executive, Magdalen House, Stanley Precinct,' Bootle 120 ZQZ, UK over time (at least for 20 or 30 years--beyond this : the probability is that relative risks decline, which is REFERENCES part of our justification for truncating "lifetime" risks at age 80). if relative risk is broadly constant over time then provided one does have some observations from the relevant time period, a valid estimation of the long term risk can be made. Mr Howie also asks for an indication of how one might estimate risks starting at ages below 20. Here the possibility of later decline in relative risk cannot reasonably be avoided. We tentatively suggest in our paper that risk might then be truncated at 60 years follow-up. The effect of this would be to predict approximately the same absolute risk for any age of exposure from age 20 downwards. _ It is not clear why "the inclusion of data for mining severely skews the risk estimates". It is true that the chrysotile mines produce lower per fibre risk esti mates that most manufacturing cohorts, but as our data shows, if the amphiboie fibres are (as Mr Howie Camus M, Siemiatycki J, Meek B. Nonoccupational exposure to chrysotile asbestos and the risk of lung cancer. N Engl J Med 1998;338:1565-71. Green FHY, Harley R, Vallyathan V, Althouse R, Fick G, Dement D, Mitha R, Fooley F. Exposure and mineralogical correlates of pulmonary fibrosis in chrysotile asbestos work ers. Occupational . and environmental medicine 1997;54:549-59. Hodgson J. Damton, A. (2000) The quantitative risks of meso thelioma and lung cancer in relation to asbestos exposure. Ann occup Hyg 2000;44(8):565-601. Liddell FDK. Exposure-response: asbestos and mesothelioma. European respiratoiy review 1993;3:98-9. Liddell FDK, McDonald AD, McDonald JC. Dust exposure and lung cancer in Quebec chrysotile miners and millers. Ann Occup Hyg 1998;42:7-20. Rees D, Myers J, Goodman E, Blignaut C, Chapman R. Bachmann M. Case-control study of mesothelioma in South Africa. Am J Ind Med 1999;35:213-22. Vacek FM. Effects of intensity and timing of asbestos exposure on lung cancer risk at- two mining areas in Quebec. J Occup Environ Med !99S;Sep;40(9):821-8.