Tobacco Documents Online

IEP01~r NO. ID. 1404 RE~C'IZD 24.8.1976 AUTHOR: R.B. W£1kas LSSI~.~ lJ~: B.E. W£11£s PROJECT JOB HO= 11.05.02 DZb'TRZ~L'ZON: Dr. S.J. Grean Copy Ho. 1 DE. D.G. ¥elCon " " 2, 3 Library I, " 4, 5 COI~N0: ii i i BAT Co LTD - MINN~TA TOBACCO LITIGATION C~

INDEX HAS INDICATED GAP IN BATES RANGE HERE

EBWlLX/46Z>'-2 Group Research & Development Centre, British-American Tobacco Co. Ltd., SO~HA~PTQ~. 2ACh August, 1976 SOb~ FURTI~R ANALYSIS OF AN EXPERIMENT TO com~J~J[ T~ zz~ou~-zqux~r~c Ac'rzvrrzzs oF c(x~zz~sx_~zs FZa~"CX~Z'nZ~ zzg/z ~.V' Z9/6 (Report No. RD. Z40A lestrlcted) SUHHAR¥ This report desc=ibes a stmtlsClc•l amalysls of the data arisln8 from the 3ANUS promoClon experiment cmzpmrlns condensaCes B9/1 end B9/6. The data were adjusted to simulate r~he situation as if the experLmmnt had been terminated •t week 32 or 36 or 40 or 44 o= 48 or 52 as well as the actual teWnLiZmlCiOU time of 57 weeks. Ic is concluded that doses of 120 US and 180 .1 o£ DH3A as Luitiltor mrs too h~sh, but ~he opC/=mm dose level cannot be determ£ned. The 60 U8 level of Dt4~A did however yield uae~uI date. lc is •leo concluded ~hac early tenatnation o£ the experiment may have been possible uelnS the 60 US DHBA groups, the C~oriSenlc raclo belng signlflcanc from week ~A onwards. It is recoemended Chat in future iniCieCor-promotor experimmmts of t.h:l.s kind at least 80 animals per group should be used, and as a rule of thumb the experiment should be terminated when the group ehowlng the lomwC response has • turnout--raCe o£ about 1OZ. BAT Co LTD - MINNESOTA TOBACCO LITIGATION o O4 o~ ".--4

2- L BAT Co LTD - MINNESOTA TOBACCO LITIGATION r~ OO

-2- INTR~DUCTION An earlier report (I) describes the desiKn of cheexper/ment and the analysis of the dace when the experiment was terminated. The objective of this further analysis is to estimate the effect on the s~alysis of terminating the experiment sooner. The exper~nent actually ended at week 56. all surviving animals being sacrificed du~iag week 57. Since the we@ks of occurremca of turnout-bearing animals and else wmek~ of death of animals dying tumourless are known the data can be adjusted to simulate the termination of the exper4~mut at any giv@n week, all surviv/ng mnjJnals be£nK resarded as having been sacrificed dur4~s that week. THE DATA The whole a~jpar~mJnc used 34 L~roups o£ mai~als, of .which 24 Kroups are used in this analysls. These Stoups were subjected to the 24 combinations arising from 2 condensates (Bg/L, B9/6), 4 levels of application (0, 60, 120, 180 ~S) and 3 condensate dose levels (20, 28.3, hO ag three t/~ms weekly). :In the oris/~al report Zhe data arm shown livi~s the incidence of turnout-bearing an/male and the incidence o£ tumourless deaths on a 6-weekly basis. SiJxce durlns the early weeks of the exper~nmnt the numbers of cumour-baarinK mnJJmls mad the numbers of cumourlass duths are small it was felt that the er=o=s introduced by grouping the data into &-week periods would be unacceptably hiKh and so, for this analysis, the data are used on a 1-~s@k bu£s. (Times of deaths and t/ross of turnouts are recorded on a week-b~-wmek bmsis.) Tables 1A r~ 2~A sho~ the data. METSOD OF ..ANALYSIS The data were analysed usinK the We£bull distribution function. Comnon values of the parameters k and w, and separate values 0£ b, were BAT Co LTD - MINNESOTA TOBACCO LITIGATION

INDEX HAS INDICATED GAP IN BATES RANGE HERE

-3- fitted Co all 2~ groups by the method of maximum llke~thood. The values of b, k and w are 8hewn ~n Table 1; these values were obCalned ug£ng ~e dace as an week 57. Strictly spaaklng, 4f we ~ro to simulate Ehe analysis of the exper~nc terminated at var£ous times, then the maximum likelihood values of b, k and w should be re-asclmated for each elmg-polnt chosen for taz~inmclon. This procedure would however imvolve a large mmountlnK of co~putlng and iC was decided co fix the values of k and w arbitrarily and only re-compute the values of b. 7rou Table I it would seam Chat k - 3.0 azd w - O would probably fit the data quite well. Table 2 shows the values of b corresponding to chase values of k and w, ~ Table 3 showB the ~odnema of fie £emt for the parameters shown in Table 2. From Table 3 it can be seen that the parameters of Table 2 do indeed fit the data fairly wet1, the value of X2 - 12.088 wlCh 6 degrees of freedom be~ juaC noC significant (iC is siEnlflcanC aC the 94Z level of confidence). The mjor contributor Co chls X2 is Che zroup of animals receiv~ag 39/1, ~0 mE, 120 .8; as treatment. In this group there were more tw~mar-bearinE animals than expected in the ~eaks ~ to 47, trich fiver Chan expected 4n Che muccaedlng 10 weeks. In an expar~JnenC eont:a:tu~uK 24 frroups of an/reals, with relaCivaly few tumour-bemr~ an;,nals in each ~roup, it i8 onlT to be e~rpeeCed Chat one or Cuo groups may show unuemal patterns of inc£dem:e of fiamour-boar~ng animals. Thus the degree of lack of fit Lnd£cated by the above value of X2 is considered Co be acceptable mad the mmalTsis was carried cmC umtJaK k - 3.0 and w - 0 for all time points. The dace were analysed aC weeks 32, 36, 40, ~, 48, 52 and 57. At each of chase c~nm poises values of S mad V for each of the 24 groups BAT Co LTD - MINNESOTA TOBACCO LITIGATION

BAT Co LTD - MINNESOTA TOBACCO LITIGATION

-4- were co,~.tad where S - I Dll ~d V - ~ CDIi + ~21> _Cai - V>k. ~ ~ese expressions DIi ks the number o£ turnout-bear!he animals occurring during week el, and D2i is the number of tumourleas deaths occurring during reek t.. Since we have chosen k = 3 and w = 0 we have V- ~ (DIi + D2i) ts. X Zt can be shown that the maximum likelihood estimate of the WelbuI1 paramtter b is the ratio S/V. Thus, at each o£ the time po/Jxts 32, 36, 60, ~, 48, 52 and 57 weeks ~e have available the max/~Jm likelihood values of b, and the values of S and V, for each of the 24 groups. RZSULTS OF ANALYSIS Using the methods dmscribed in (2) and (3), an analysis of variance was £ormed for each of the time po4ucs. The results are shown in Tables 1B to 21B. The four levels o£ DHBA pte-tre~taeat (0, 60, 120, 180 ,S) were treated separately, and at math of the time points, £or each of the DMBA levels, the Bg/1 to B9/6 tumorigenic ratio (sad its 95Z coufide~ce limits) were calculated. The data from the 0 D~MA Kroups was /~8ufgiclent to allow the calculation of the ~umor/~enic ratio (only 2 t~mour-bearing an/male occurred in these 4 8Taupe); the results o£ this analygi8 of the data £rou the ocher three l~vel8 of ~(BA pre-tteatmant is shown in Fisures 1 to 3. These Krephs show plots of the (natural) Iosarithm of the tumor£Kenic ratio of B9/1 to B9/6 with its 95Z eou£1dence I/wits, the slope o£ the dose/response curve, and the numbers of tUmOttT--bedLT~ anJJnals, versus time. In order to compute the tumoriSen£c rat£o, a parallel line regression of In Cb) versus In (condensate dose level) ks used, the two lines representing the two condensates. Than the In (tumorigenic ratio) i8 the distlmce between the lines divided by the slope of the lines; it i8 the slope of these parallel dose-response BAT Co LTD - MINNESOTA TOBACCO LITIGATION o c.,.rl %.0

INDEX HAS INDICATED GAP IN BATES RANGE HERE