Tobacco Documents Online

-6- is invuitively Implauslble and c~nt~adicCs a11 previous a~l~rlenee (except at smch hJLsbar dose levels). Table 9 sho~ tbm Cumori4eu/c =sties sad their 95Z =ouf£dmnce 1isles, mad from tlt~8 table Lt san be seen that the ij9[3 to n9/$ tumoriSenic =argo from the two ezpa=~msts CKstt~Ile a~d W~~ are uot s£ip~££csnt~7 d~fere~t. Also, if the Bg/S, 56.6 uS point is £Snored, Yi4~uz~ 1 sussescs that the ImSmral ~evel of response ob~ at giclcham 4s I~LKher thms that obr~ulxu~ at I~ttelle, and in £8ct the 4U£e:eace beCmmu the two laboratories 4s ~usC s~Lcamt at the 95Z 1Level of confidence. Zt is morCh nott~4~ that the hLKhm= :espouse at W~ckham ross obts~:ed vith the lows: level of ~ ptecre~cmsnt (120 ~ ac W£ckhmm; 180 ~ at Battelle). This tends to cou~Lrm the result repo~te~ ia (2) whecetn 4c was noted that in another p=~mot~ study at Bsttelle the 8zoup =ec~vtn4~ a 180 ~S pretreaCmant of IRfB~ gave • lomex response than chose receivLu4 120 ~g. h(£i) Comparison of l~It~ varLsncs. Using comm~ values of k a~d v and sepazate values of b for each group the ~e~bull d£str4but4ou ~unct4on m f£t~ed to the data of T~le 3 for groups 3 to 6 and 11 Co 26. The mthads used are described 4. (~)| the results axe shoun in Table 10 and plocte4 in Figure 2. Treble 11 shows the result of • Sootiness--of-fit test of these parameter est~unstes from which it can been seen that an adequate fit has been obtained (X2 - 7.60 with 6 deKrmm8 o£fza~dms~ 818n~£cance level 59.6Z). Table 12 shows the lge4b~11 scat4sttcs S and V derived from thsse est4-~tes of the pa=mttrs. BAT Co LTD - MINNESOTA TOBACCO LITIGATION O t.J-I

INDEX HAS INDICATED GAP IN BATES RANGE HERE

-7- J~om FiKurt Z it uould appe~ that the tobacco control tends to be mote s£t£vo than the PEt vszl~ntSo Also uot£ce~ble £s t]~ aeSat£ve dooe-tmspmu~ tolations]~p fc~ the £s~tor combLzLst~ou OH, L, H). As noted eazlior, such neg&t£ve slopes are extremely $,~plausiblo, but tram tba £tj~uto £t £s not poss£ble to deto~ which of tba t~o points £s lilly to be ~ ez~or. Table 13 shows the results of an 8nalysis of vs~ianca of tha d~ta o£ T&bls 12 from wh£eh it can be seen that, sitar c~ndausate dose ~vol e££octs h~m been xllou~d £ot, tharo rema~s • hi4~h~y si83~icant betweon-r~n~lensate e.~feat, as miKht be expected from FiSu~e 2. Also, Table 13 shows that there are no s~Kni£1cent ~teract~ons! thus it m~y be ~J~ertod that the noB&tire dose-response I~nA £or (R, L, H) in P4Kute 2 is not stat£sticall7 si~£c4mt. Table l& shows the t~morisenlc tatlos betws~ the tobacco control and the atria ~]~ variants. As noted £tom F~4~u~o 2, the ratio £s loss than 1.0 £ot ~]L the ¢ompat~sous, but £s sis;n~ic~mt £ot o~ly two, those fo~ (K, L, L) sad (R, R, W). TILLs c~nfi~as the conclusion dtsva ft~n FiKure 2, that the ~R~ vs:~ants are Sea•rally 18so tmwr~Ke~ic than the control tobecco. 4(£££) R~lat£~s e~£eccs of F~t varLables. ~na re~&t£o~sl~Lps amonsst the £8~tors o£ the PR~ varis~ts ~nro ~u=vtstijtted bY two mtlbodo. (•) Fixed E£~.ects Model Using tills approach~ each o£ the Chloe vaz-£ables (bass v~iKlst, &.S.V., 8olubles), a~d abe condensate dose level, uere coded u baths at either BAT Co LTD - MINNESOTA TOBACCO LITIGATION

INDEX HAS INDICATED GAP IN BATES RANGE HERE

-8- the high or the %or level. The median po~t was d~sre$&tded. (The ned~n point contt£b~tes uo Lu~ormst4on tosa~L~n| • test £or the um£n eJ~£ect of the variablt8. It only provides • test for non-lineaz£t7 of response, and in the context of • fJ.3ued-e~£ectJ model tk~s vest: yes ~p~oz~d.) The tmmlttjzK analys~s of variance i8 8hewn 4~ Table 15, £tom ~b~ch it c~m be seen that the X2 for the ma4~ effects (21.307 irlCh 4 desrea• of freedom) is hishly significant. PartitLoni~K this X2 8hove that the doe• e££eet m~d the base ve4Kht e~fect m biKhly siKntfL~t, and a.s.v, seems to have •1moat no effect at a11. TtLte latter coucZus£on 48 mod4££ed by the result that slthouKh the ove~aX1 X2 of 8.&34 with 6 de|tees of £:•edom for the 1st oz~le: £nteract£cms is not 81lp~£icsmt, pa~£tiou~ this X2 reveaLs that the dose • •.s.v. intetact~on £8 s£es~£1cant. Znsp~:tlon of FiKure 2 8ugKests thAC the results for the (R,~. L, H) combinat£on of the £actove, 8Lv~ns • nesat£ve dose/response line, may be respous£ble for th£s tnteract£on, and the hiKhly s£Kn/~£cant ef£ect of base weiSht £nd£cates the Oi, L. H) 40 ms Kroup as hay/n• pz~dueed an unszpectedly h£$h uumber o£ tumour-beatLuS animals, since the high base ~£8J~ Stoups have lenera~ly 8~Lven the lower response. • fuz~her pa:titJ.on£ng of the X2 o£ 5.019 Or£t~ 4 degrees of £reedom) £or the 2rid orde: £n~e:act£ou effect £8 due to base ~e£1iht • •.8.v. z solublas (X2 u &.876 with i degree o4v f::eedom. SI4p~Lt[L~t: at 97.28Z). The th£rd order ~Lute~aet4on 4s •18o sLezt~fLcant. Th~s althouKh a.n.v. ~ould seem to have a msgl£1;tble mL{n ef£ect, the £a~tor a.s.v, sppears In al1 the e~l~Lf~csnt £nteract£ons. A re-s~slys~s o£ tbm dsta omit~$ the ~esult for the rE, L, H) 40 q IWoup t8 shoun in Table 16. Note that the om£ss~on of tk4s data BAT Co LTD - MINNESOTA TOBACCO LITIGATION

INDEX HAS INDICATED GAP IN BATES RANGE HERE

~o point :sans that there is no test £or the third order intetsct£~, s~uce • destae 0£ treadle has l~sn lost. ¢~mpat~son 0£ Tables 15 and 16 s~nm Ct~c there is l£ctle chase ~ the results of r~a SilP~T~£cance r~sts as far u che msJ~ e~ects ate conce~d, but that in Table 16 ~e o£ t~ tnte:sctlcms ~s e~Lf£cant. Zt ~s a/so of ince=est to z~-calculete the tumorigenic ratios with (R, L, B) 40 s~ c~tta~I; t~ zlsu]Lts are shown in Table I7. Comparison of Tables I7 and 14 shows that in ram fetter the conf/dence I/m~ts are t/~hter, and • third ret£o, that £or ~obacca v. (a, L. B) is nov sil~L£1cant. (b) ms=:essiou Hodel UslnS this approach the average values £or base ~iEht, a.s.v., 8olubles, sad dose l~rel Rre used ~mJtaad o£ tlm 0/1 codtns used to denote the hish/lov levels of the var/~ables in the fixed e~£ects :o4el. The values of base wa£sht, a.s.~, a~d soLubles are shown in Table 1. The dose levels were 40 ~li and 56.6 ~1~ t:broufhout. ~ ~z~ss£on ~_els ~=e used:- 1. D in(b) - ~ + a~ (base =e£Sht) + a~ (base ~isb~Z) + a~ (a.s.~.) + as (solubLes) + =7 (does) + + + ] + 1-, these smdels, b £s the Ve/.b~11 parameter b £o~ m~ch o£ the 18 Irroups, sad I:he 8~L ~ the b~ 8re p~z'~ste=s to be detexmizmcl. ]rm: mc~lel I the scale4 4at;a shmm i,~ Table 1 were used. The sc~lia8 was a~opted to m~ BAT Co LTD - MINNESOTA TOBACCO LITIGATION Oo r',o ~O

INDEX HAS INDICATED GAP IN BATES RANGE HERE

-10- ¢ompu~ttioLtI probIemJ vi~h Iarge n~mbe:s. 7or model 2 ~he (n~zural) los~£cbms of cbe un~.sled da~t ~re used. Us~a~ methods described 4~ (4) Chose models ~tre ££Z:ed b~ maximum l~dkellh~od; e~ L'a~l~illE mmmlySe8 Of vaz"J.snce 41rm S~ ~J1 T&bles 18 aud 19. mlot:t that Cbe i.uclus£on of the me4£an po:[,nt in these reilreas:Lou. ammlysea doe8 not provLde • eeoc £or ~o~-ltJso~c£cy of dome response since onl.y Cue dose levels were used ClnrouShout. Theme two tables show Chat on~y the dose effect mnd the effect of base vm4KhC ere siKn~£cmnt, and chat the effect of bus ~8LSbt £s 8dequmcely represented by • s~npl• liJzezc model. The 4-~luence of the intermct4on~ becvoen She PIT £•ctors can be see~ in Tables 20 and 21 ~bere the coef£icieucs of the vexious ~eKceesiou models that mare E4tte~ to the dJca ~ce shmm. Znspoctiou of chose rye Cables 8ho~ that ~ter the dose and linear base ve£sht te~ms ~e t~cluded there is l£ttle chanSe in ~be l~tl£bood, but the tnclusLon of fuz~e~ terms in the model cm~es larks fluctuations in ~he estimates of the coaf££cienc8. ~J.tJ~oush the e£~ect of soluble• d~es not: reach siKn/~£canr.e, th, tre £s some ev£dence in. Tab~e 18 chat: £cs effect may not: be ne81~Lble; the 1~ses~ te~n ~s slK~L~mst at: 8.So&3•. T~ • model u8~ dose, broom ma~ht l~emr and 8o~uble 1~ £e f£tted to the data, the equ~t£on becomes ~n(b) - -9.9657 - 0.3562 (bsoe ~m£ShC) ÷ 0.2834 (soluble•) ÷ 1.26S0 (do~) [Tn:ILs equa, t£cm :ts for scaled data']. The l£k~l.Lhood o£ t:his u~lLeZ £8 L -- -107.136 aad ~ oo~ubles temm £s 8~sn~icant et 85.88~. BAT Co LTD - MINNESOTA TOBACCO LITIGATION m r',,o C~

INDEX HAS INDICATED GAP IN BATES RANGE HERE