Document 9LqyM7Y4nNv6Kg6ZknyOegnDL

RE: Seeping of benzene pooled epidemiological analysis From: "De Jong, Geert G SI-SHS" <geert.dejong@shell.com> To: "Swaen, Gerard (G)" <gswaen@dow.com>, "Tsai, Shan P SHLOIL-SHS" <shan.tsai@shell.com>, "Roythorne, Christopher" <chris.roythorne@uk.bp.com>, "Satin, Ken (KSAT)" <ksat@chevrontexaco.com>, Urbanus Jan <jan.urbanus@concawe.org>, chris.money@exxonmobil.com, rolf.ahlberg@nesteoil.com, jeanphilippe.gennart@total.com Date: Fri, 03 Jun 2005 10:43:01 +0100 AIL We have a range of opinions on possible way forvvard. Should we set up a telecorn to discuss? I! you agree, Jan could you action? Kind regards, Geert de Jong Senior Health Adviser Shell International B.V. PO Box 162, 2501 AN The Hague, The Netherlands Tel: +31 70 377 2516 Fax: +31 70 377 2840 Other Tel: +31 6 55 124 123 (mobile) Email: geert.dejong@shell.com Internet: http://vvww.shell.corn -----Orig ina I Message----From: Swaen, Gerard (G) [mailto:GSwaen@dow.com] Sent: Friday, June 03, 2005 9:16 AM To: Tsai, Shan P SHLOIL-SHS; Roythorne, Christopher; Satin, Ken (KSAT); Urbanus Jan; chris.money@exxonmobil.com; DeJong, Geert G SI-SHS; rolf.ahlberg@nesteoil.com; jeanphilippe.gennart@total.com Subject: RE: Scoping of benzene pooled epidemiological analysis Shan, I was aware of the possibility that I would propose something against an earlier decision. Still I would encourage some more thoughts on including other datasets: 1. 1. Other datasets are available and getting permission would not delay the project very much. 2. 2. The highest exposure to benzene in the petroleum industry is still very low. Some top loading and splash loading of drums may have occurred, but the TWA's would not be higher than a few ppms, I estimate. Now if a dataset with this small dose range will be used to find a threshold, it can never come up with a higher threshold than the highest observed exposure. The point is: Maybe the whole study population is below the threshold and the pooled study can never provide evidence for a threshold. 3. 3. Including datasets with higher exposures, in which some small effects have been reported in historical exposure situations, would allow us to evaluate a much wider range of the exposure spectrum. This would provide a much clearer dose-response curve. CGU BEN0000029 Gerard Gerard Swaen Tht~ Oow Cht~mk~~tl Company Epidemiology Health Savirt~s tel: 31-43-3626042 -----Orig inaI Message----From: Tsai, Shan P SHLOIL-SHS [mailto:Shan.Tsai@shell.com] Sent: donderdag 2 juni 2005 16:40 To: Swaen, Gerard (G); Roythorne, Christopher; Satin, Ken (KSAT); Urbanus Jan; chris.money@exxonmobil.com; DeJong, Geert G SI-SHS; rolf.ahlberg@nesteoil.com; jeanphilippe.gennart@total.com Subject: RE: Scoping of benzene pooled epidemiological analysis Gents, The objectives of the proposed pool study was 1) to improve the analyses and interpretation of the three existing individual studies, 2) to further clarify the potential risk of overall and cell-type specific leukemia !rom low-level exposures to benzene, and 2) to extablish a threshold for benzene exposures To achieve the above objectives in a "timely" schedule and minimize the potential complications of exposure assessment, it is important that we should not expand beyond the existing three stuidies, i.e., IOL, IP, and HW. To e.>..'tend the follow-up of tl1e IOL and IP woulcl increase the statistical power, however, the time delay (additional 2-3 years) should be a factor to consider. Although the original IOL case-control study includes cases thorugh 1983, the cohort of this nested case-control study has been updated through 1994 (Lewis et al. Occup Environ Med. 2000 Sep57(9j595-60J.). One alternative lo consider is io add only new cases horn the IOL that were identified from the upclatecliOL study ancl not include the additional update ofthe IP. Regards, Shan P. Tsai, Ph.D. Manager, Epidemiology Shell Health Services Shell Oil Company P 0 Box 2463, Houston, TX 77252-2463, United States of America Tel: +1-713-241-6078 Fax: +1-713-241-5875 Email: shan.tsai@shell.com Internet: htto://wll-IW. sfleff. com This e.~rnan rna~:/: ('1) be subject to the .i\ttorney~-c:nent pdvUeqel (2) contain C~()NF"![)ENTl~\L ~,.1E[)!(~-:6..L !NF.C)~~tiV-\T!C)N, or (3) be strictly conndeniiaL H you n.::c.e;ve it in t::rror, you are on notice of its status. F)!ease not!fy us irnrrH.~djah:~!~/ t.1~/ reply err~an Hnd thfH) ch:~!E.:te tr:is rne~~sa~1e frorn your cornputer ~~ysten:. F~!e.:~~:r:! jo not copy it or use it f)r any purpo~:r:!. or ::.J~sc!ose its contents to anv per:~on to do so couki t.1e a breach of conHrJence. 'rhHnk. vou for your cooperation. CGU BEN0000030 -----Orig inaI Message----From: Swaen, Gerard (G) [mailto:GSwaen@dow.com] Sent: Thursday, June 02, 2005 2:49AM To: 'Roythorne, Christopher'; Satin, Ken (KSAT); Urbanus Jan; Tsai, Shan P SHLOIL-SHS; chris.money@exxonmobil.com; DeJong, Geert G SI-SHS; rolf.ahlberg@nesteoil.com; jeanphilippe.gennart@total.com Subject: RE: Scoping of benzene pooled epidemiological analysis All, In my perspective Ken and Christopher have both raised some very crucial points and I believe the first step should definitely be to avoid comparing apples with oranges, with respect to the exposure characterization as well as with respect to the outcome parameter. If possible, phase 2 should make sure that all contributing participating institutes have done their best to collect specific data on diagnosis. In addition to the points made by Ken and Christopher I would like to make the following points: 1. Why is the pooling effort restricted to the three petroleum industry studies? Pooling of only three studies will be less informative than if more studies would be included. There are several other studies on benzene and leukemia with reliable exposure information that also contain relevant information and would increase the exposure range under investigation (indispensable for threshold analyses) and would significantly increase the power, Some other interesting studies that could be included are the Monsanto study, the recent Dow study and the recent DSM study. The advantage of including the Dow study and the Monsanto study is that they report an excess in the high dose group and thus contain information on a possible threshold. The IOL IP and HW do not contain such clear doseresponse curves, if any. Even including the pliofilm cohort would have my preference, but I realize we run into trouble because of the different exposure estimates that are available. An extended pooled analysis would also have a better chance of evaluating the relevance of peak exposures. One of the most probable mechanisms for leukemogenesis is thought to work through bone marrow toxicity. Including bone marrow toxicity in the study opens the opportunity to study the relevance of bone marrow toxicity as the mechanism. If we include more studies we can test the hypothesis that increased leukemia rates have been restricted to cohorts or sub cohorts in which bone marrow toxicity has occurred or at least has been reported. 2. If point 1 is no option, I very much support the option to extend the follow-up of the three studies. If no new cases are added to the IOL, IP, and HW current databases a pooled analysis will only have a limited added value. 3. Including cell type information is very important. Given the very small excesses, if any, it is very important to have the outcome parameter as specific as possible. If not, the doseresponse analysis will become diluted. In addition, the results can be criticized for not having information on cell type. 4. I do find the estimated costs on the high side. The total is over one million Euro. For such a budget we could have a complete new study done. With respect to cell types: Have you seen the recent publication on Benzene workers in the UK by Sorahan, Kinlen and Doll in Occupational and Environmental Medicine? He concludes that effects are limited to ANL. Gerard Swaen Gerard Sw aen CGU BEN0000031 The Do\<Y Chemical Company Epidemiology Health Services tel: 31-43-3626042 -----Orig inaI Message----From: Roythorne, Christopher [mailto:chris.roythorne@uk.bp.com] Sent: woensdag 1 juni 2005 09:02 To: Satin, Ken (KSAT); Urbanus Jan; shan.tsai@shell.com; chris.money@exxonmobil.com; Swaen, Gerard (G); geert.dejong@shell.com; rolf.ahlberg@nesteoil.com; jeanphilippe.gennart@total.com Subject: RE: Scoping of benzene pooled epidemiological analysis All, I think Ken has raised some interesting points and highlighted sorne significant challen~p:;s with respect to exposure and cell type diagnoses which are the two critical factors" Leavin9 the exposure issue to one side I see the task of obtainin9 specific cell type results as a huge task and one that may not be achievable to Hw extent that would be of value across all ofthe studies. Tllere is also another aspect to consicler, that of original cell type diagnoses" Having recently been in Shanghai my understanding is that the study there is developin9 vvhat I would describe as very interesting data regarding cell types \Nhich will have, in my vievv, significant implications !or previously published studies. As a consequence, not withstanding the huge practical difficulties of any work, I would be reluctant to embark on any further vvork until we understand what the implications of the China Study are Re!Jards Chris From: Satin, Ken (KSAT) [mailto:KSAT@chevrontexaco.com] Sent: 31 May 2005 23:36 To: Urbanus Jan; shan.tsai@shell.com; chris.money@exxonmobil.com; gswaen@dow.com; geert.dejong@shell.com; rolf.ahlberg@nesteoil.com; jean-philippe.gennart@total.com; Roythorne, Christopher Subject: RE: Scoping of benzene pooled epidemiological analysis Gentlemen: ,Jan's scoping outline pretty much covers the full range of options" I think most are irnpmtant to consider, but ~~iven rnonetary and tirne constraints (althou~!h at this point in lirne I don't knovv what those are), it is likely the entire scope would not be agreed to up front. This suggests a phased approach might be better. with each subsequent phase contingent on the successful completion of the previous. Of course, this could make the whole project take lon~jer and cost rnore than be~jinning all aspects at the sarne time. Nevertheless, assurning a phased approach is most practical, I vvould suggest the following phases: 1. Rationalize exposure across the studies. \JVe are trying to quantify a relationship so unless we can achieve consistency across studies, there is little basis for combining data and thin kin~! \Ne will obtain a rneanin~jful estimate of effect. This should be the first step. If il can't be done, we need lo rethink what outcome we want frorn the project. 2. We would like leukemia cell-type specific results so the next question is whether we can get this from the data. I would think rationalizin9 diagnoses across studies would entail obtaining historic medical records. Depending on the availability and access issues related to sucl1 data, the ~1oal of this phase may not be acllievable" Therefore. I propose a feasibility study to see what additional data coulcl be obtained from a sample of each studies' cases. I would run this pilot concurrently with the exposure rationalization. In the CGU BEN0000032 event this task cannot be completed successfully. it is not necessarily a show stopper to performing a poolecl analysis. For example, if cell~specific information was not available, the analysis of the case~control results could be run (as a series of simulations) based on rnodeling the cell-specific outcomes based on the "usual" cell~type distribution for a given time and study location and seeing how sensitive the results are to the assumed cell~type distribution and number of imprecise or inconsistent leukemia diagnoses sucl1 estimatin!J must be applied to. 3. Assuming phases 1 and 2 are successful, the analysis could be run now. This approach \NOuld be the least tirne and money intensive. However, the nurnber of cases in the kNJ exposure range might still be too small to yield the effect resolution we're looking for. Therefore, this is the time to consider further updating the parent cohorts to identify more cases. Note for example, that in the IOL cohort that the case-control study was based on only 11% of the cohort had died (3909/34597). This is a lovv percentage and may not be representative of the cohort's exposure experience so updating Hw vital status of cohort members is probably a goocl idea. To save time. this phase coulcl be started as soon as it appeared that the exposure rationalization was going to be successfuL 4. Conduct the pooled case-control analysis using "rnodern statistical rnethods". Since the issue here was lhe elimination o! arbitrary exposure categorizations, I suspect the goal is to use a regression technique. My caveat here is that we need to be careful that the analysis is not limited to, or for that matter, even includes, (log) linear models. Instead, a general additive model rnight be the default used. Note that there is another irnplication of using nonlinear models~- namely, average or unvveighted cumulative exposure rnetrics will not be particularly meanin!Jful. Consequently, some special modeling may be needed to pursue the application of a nonlinear exposure~response curve. Note that, instead of using rewession techniques, the arbitrariness of exposure categorization can be looked at by testing different category cutpoints and seeing lhe impact on the leukemia risk. I'm not sure how one would interpret the results if we found the effect was very sensitive to the cutpoint, but it rni~!ht tell us something about our data. 5. Sensitivity rnodeling of exposure-response curve. While the essence of this task mi9ht contained in the analysis as I've described it in Phase 4, I don't think doing a sensitivity analysis for its own sake is wortl1 it. If tile modeling results are similar, we've not learned very much other than may be our data is lacking in its ability to achieve the resolution we are interested in. On the other hand, if the results of different exposure-response rnodels are di!ferent, we vvill have interpretability problems since we lack the golden ruler to tell us whicl1 is correct. I would pursue sensitivity modeling only if we had leftover funding. Ken Satin, DrPH ksot@chevron.corn T: 510 242-1610 F: 510 242-7022 -----Orig ina I Message----From: Urbanus Jan [mailto:jan.urbanus@concawe.org] Sent: Tuesday, May 24, 2005 6:21 AM To: Satin, Ken (KSAT); shan.tsai@shell.com; chris.money@exxonmobil.com; gswaen@dow.com; geert.dejong@shell.com; rolf.ahlberg@nesteoil.com; jeanphi li ppe.gen na rt@tota l.com; chris. roythorne@ uk. bp.com Subject: Scoping of benzene pooled epidemiological analysis Dear Colleagues, As requested at our recent Health Management Group meeting, I have drafted a first attempt at scoping out a pooled analysis project of the 3 nested case-control studies of CGU BEN0000033 benzene-exposed workers (attached). I would like to encourage you to comment on this via the 'reply all' function, so that we can develop this idea further into a more realistic scope. Please note I have agreed with Louis Bloemen that he would contribute to the Benzene DCR project as a consultant (after leaving Concawe member company Dow) up to the finalisation of the Phase 1 IOM/IRAS report, and hence he is not part of this next stage. I have included the remaining STF-30 members (Ken and Shan), supplemented with some HMG members- I trust you don't mind, but please advise if you feel we need others to contribute as well. Jan Urbanus Technical Coordinator, Health Issues CONCAWE (the oil companies' European organisation for environment, health & safety) Boulevard du Souverain, 165 B-1160 Brussels, Belgium T: +32-2-566 9163 F: +32-2-566 9181 E: 4an.urbanus@concawe.ora W: www.concawe.oro CGU BEN0000034