Tobacco Documents Online

866 It is unclear whether the result in the youngest cohort can be espl>lined by' aggressive tttmouy growth, by too low a sensitivity' of the mammographic test, bytoo short a follow-up, or by chance. The absence ofany effect so far in the youngest age-group eorrespondc with the early results ofthe age•group40-49 at~entry inrtheHealth Insurance Plan (IiIP) trial. 5 years after the HIP screening programme had begun, the number ofbrast cancer deaths was 19 in the study group and 20 in the control group (RR-0• 95)a !0 yars after the start the RR was 42154 (0 ~ 78), and 14 years after the surt it was 4b/61 1(0• 75):4 Nevertheless the HIP investigators hesitated to accept this finding as evidence of the effectiveneas of screening ttnder, age 50. In Nijmegen a disease stage classification aystem according to mammographial and/or histopathologial tumour size was used: ••Advanced~ stage" means that the uilliry' lytnph nodes were histologically involved or that the lesion consisted of infiltrative carcinoma and was at least 2 cm in size. In the age-group 35-49 at diagnosis 38°10 of 4D sereen-detecned cases had advanced disease stage„s opposed to 4 out of the 6 cases in women who did not participate in the screening programme. According to thesrfigures a subscqumt tnortality reduction can be expected in the youngest age-group. Finally,, attemion should be paid tothe.pak effect ofaaeeaing on breast cancer mortality in the oldest age-group. It is assumed'that breast cancer grows rather slowly in this group.s•6 As a consequence, the lead-time should be very long, and'a strong effect could be arpected afier a longer period offollow-up. The odds ratio for the birth cohort born before 1910 is now only 0~ 81. Maybe this RR estimate is weak because of differentatnderlying mortality risks (independent of any'sereening effect),in the participating and non• participating groups. Maybe difTerences in patient't delay explain that the effect was less favourable than expected. And maybe selective miscl9ssifiation of the dtath certificates is another explanation. Further studies will focus on these potential biases. Dtpcrment, of SecW Medrarre, A. , L. M. VExBEEx Aadiokot,r; nd. PaLoloay, Um.enrtpofNiymcre.n,. ]. H.,C. L HENDRIAS 6525 El.l:nmKsn, HetEcrlaodl R. Hot.IAND Depeemrm e(PY[EWoay... CmuurFAMl^~ Ho.pinl,' Thjme6rn M. MM'VUHAC Depomxa, ofEpdemrWoe, Um.aairy d l.,meur6 F. SJtTRMArPS I. Vrrbe.Y AtJd. HenYrlb 1HCt., HdLndA, M».ue.c M, Semm.o. F„DYy PrE. ttsdunwn of.Errst mvcv avortalu7 Wrou6Em.a .oemas ietN eodrrnm,mmup.pby: Fue reulu ortae Nymqw Psopct, 1975-t9al:lr.m t(U4, ~ . 1Zt2-24 . 2. tlra/ar NE. Dry NE. TL. myw dorrcaavol irWra. Is D..'s W, d 6mmd mahod. u oocaraurch. Lyoo. IARC.Samurie PuW+oriaa, 19a0: 12. 3. SA.pvo S, Srnt P4. Vcnn L. Penodu boot oaeu .noraiea m vamaYi moruliq hnm besu aocm. )AMA .107 1, aU: .,T77-65. 4. Sl.pno S, VrneiR',Sern PE, Vmrt,L, Roeer R. Tm-m favtttoyw drect of .nocnmtoa brerec ona: morW ny, j JJ.rl.C.G.o. l+n 1%2;' q: 349-55. s. rrournkr D.on• W sber E,.Hocnlcn Q, B.uer M, Rubli F, Rt.n4 V: Gre.tD rm a(I47. m-m.ry. oroaomo.. fi.nr. 1 aa0; Ya: 2146-2Q7, , Y. Mlysr 1S, McDnut 7ltr, Siooc Rll, Prn MU, D.utt pC. hatid t.vu raomm. otIl lieettm Itr.it..nd rpd.te )rru. Cs.a. Ra Tr.n1YNi 4 7f-a6. LIFEfIME PASSI1rE SMOSi1VG AND CANCER RISit SlR,-Dr Sandler, and'colleagues (Feb 8, p 312); in a prcliminary report, describe the relative cancer risk for persons living in households with 0, 1, 2, and 3 or more members who smoke. The risk incrtasedi both for active and for oon-unokers, with the number of household members who smoked, and Sandler et al suggest that exposure to ambient smoke in the household might be responsible. In table 1 they normalise the odds ratio for cancer risk to unity for households with no (other) smokers and disregard exposure to cigarette smoke outside the home. Calculating to two decimal places, the odds ratios for houscholds with 1(other), smoking member are 1• 42 for active smokers and 1• 45 for non•smokers; for households with 2(other) smoken the corrtsponding ratios ara 2• 25 and 2• 32; and'for 3 or more, the ratios are 2.42 md 2.7 5. The risk ntios for active smokers are therefore, within the error limits, the same as for non,smokers; to simplify the argumenti I shall treat them as identical THE uwcEr, ArR1L 13,1985 Suppose the average risk of cancer is N from all auses unconnected with smoking;, A from active smoking, and P from passive smoking. An active smoker is alw a passive smokeri of his own ambient cigarette smoke, ao the total cancer risk for active smokers in households with no other smoker is N+A+P; in housetioldt with I other amok'er the risk is N+A+:2P. For non- amokers, thecrorraponding risks are N and N + P. Beattse the data of Sandier et al imply that odds ratios (and hence the ratios of relative risks) art virtually identical we rsquire that (N+A+2PN(N+A+P)r(N+PyN. In other words A+P-0. This same relation is obtained from relative risks in households with 2 and with 3 or more (other) smokers. A multiplicative model, in which the cancer risk in, for esample, an active smoku in a household with no other smokers is of the general form, N(I+A+P); where A and P are now proportional to the concentration of e$ective carcinogens in active and passive smokers, respectively, also yields the same equality. The relation A+P-0 leaves us .rith tlirec possible interpretations: (])~ Active and passive smoking are both non-carcinogenic (A-P-0); (2) Active smoking is carcinogenic and passive smoking is prophylactic (A--P).1 (3)'Aaive smoking is prophylactic and passive smoking is carcinogenic (P0 -A), The statistical uncertainty in Sandler's table i is large enough to permit slightly less paradcaial'inferences, but let us pursue the tmthinkable a little funher~ Three randomised cvntrolled intervention trials (the Oslo study,/ the Whitehall study' and MRF7Tl) provide a direct epidernio1ogial test of the hypothesis that giving up active smoking reduces the risk of cancer. In the "intervention" (low-smoking) groups in these three trials together there were (including registrations as well as deaths in the White4all study) 149 cancers in a combined entry population of 7746 (l•92eFo); while in the relatively high•smoking control groups there were 121 cancers in 7797 From the orthodox viewpoint-tnmely, active smoking causes at least 30% of all artcers-these findings are as paradoxial'as the inferea7ces from Sandler's study. We might just be able to postulate compliated, though implausible, causal models to account for Sandler and colleagues' table 1, or we may put those results on one side because of their preli.minary, eharacter, It is more difficult to evade the implication of the methodologically reputable nnd'omised trials: active cigarette smoking has little or no net carcinogenic aaion. Depus®m, of M.d~ml iPtry},c., UO/Rn1Iy, of LY'EF, Geoerd lerirm.ry,.. !A.* t.s 1 SE%: PHILIP R. J. $U1lCH I. H ivm.no 1, V i1" ayre K, Holmr 11 t.om P. Flr.n af dr.+ md mwtra6 u,eer.mtwo m.the imdeaceof cororyry t+nn d~: tlepon from tlie OJo Srodt,Gr.up,.of.r.66nuud vul.lolplEYy mm: L.rn t96t; u:~ ISQd-M 2. tlo.e G. Ham,droo P1S, Cehwrll L, Shoplry,Ml. A nndumrad tnnmoWkd vui d.ot: wwtant drrt: 16ycv rtsu6u. J EpJr'wr! Lw.wr+. HhA. 1962,a6; .IU2-06 l. M1tF7T rteurcE Grcup. Mulnph not ficta raer.vouoo wl: IWtf.non cLnan od morWn7.ruuln ftMA 19a7, fYa: 1U65-77, 9x; Dr Sandlerand colleygues' paper on the cumulative effects of lifetime passive smoking on cancer risk appears seriously flawed, They compared 518 out of a total of 740 cancer patients aged 15-59 from a bospitalabased t umour registry, with 510 controls of the same age, se:, and race. 309 controls werefriends or acquaintances of the patients and 209 were randomly selected by systematic telephone sarnpling.. Results are presented for 369 (70%);ases and 409 (79°70) controls. Apart from death before the subjectt could be contacted and refusal; only those persons who had "lived with both natural parents for most of t'he frrst ten yean of life" were analysed. We arc told that they supplied information about the smoking babirts of their spouse and parents. Presumably, they also supplied information about~their own smoking,, „ The results are shown in three ubles,, only tllx of which differentiates between active aad passive smoking, and none off which difTerentiates between sex, age, and race. The belief thatt these factors do not matter because they were matched for is un6ound The reader has no idea how differences between those

'tHELANCET,ArRIL 13,1985 initially chosen and those finally analysed could have influenced the results. Moreover, as soon as the groups Bre stratified by active smokers and never smokers, the matching is broken. Until, the results have been presented for non-smokets by age, sex, and nce no conclusions can be drawn, Furthermore the trunated age group (15-59 years) has resulted in an ttnrepresentative selection of' cancer tues. Even so, the distribution by cancer site seems strange: there were 62 (17%) cancers of the cervix uteri but only 19 (5%) cancers of the respiratory tract and 48 (13%) breast cancers. The trend in cancer risk',from multiple household exposures to cigarette smoke is least impressive for cancer of the respiratory system, where an effect might be expected to be greatest, and most striking for leukaemia and lymphoma where any biologial explanation is, to say the least, obscure. The most extraordinary finding appears to have been the very similar trends in cancer risk with number of bousehold members who smoke irrespective of whether the cases smoked or not. Indeed, Sandler and colleagues' publication on the aame material in t he Rmeruan jot+rnaloJEprdemlclogy(1985;121: 37-48)', shows that the effect of pauive smoking on cancer risk appears to have been greater than the effect of active smoking. &hool erPublk Hedtki, Unr.cniryof M,cLµn,: •. AmA1Eec,M,ceuaan9alq9, VSA IAN HIGGINS;•These letters have been shown to Dr Sandler and oollagves,. whose reply fol lows.-ED, L StR; Professor Burch presents an algebraic rearrangement ofour data that suggests that smoking is protective. His approach assumes that childhood and adulthood exposures are interchangeable. As we indicated in oun paper, the apparently linear trends in table i simplify a complex set of relationships. Our data illustrate that childhood and adulthood exposures may contribute independently to cancer risk in adulthood, but this does not imply that these two e:posures are equivalent. Data we present elsewhere suggest the two risks may, in fact, be different (ref 1,,and unpublished). As shown in the accompanying e2pznsion of tablt n, the odds ratio assoeiated with pacsive exposure only as an adult was 1• 8 for nonamokers but only, 1• 2 for aaivesmokers(not equal, as Burch's analysis requires). For childhood exposures, the opposite was true: the odds ratio was 1' • 9 for smokers and 1• 3 for non•smokers: Thus, passive exposure in childhood seems to have its greatest effect among persons later exposed to a carcinogen (their own smoking), while passive exposure in adulthood has its largest effect among persons notiaaively exposed. In shortl our data do not support the timplified biological assumptions Burch requires for his analysis. Dr Higgins raises concern about possible biases in the study and reduests additional data. The information be seeks is provided' in a paper be cites.l As explicitly stated in both papers„there was no confounding by the variablts mentioned by Higgins. The OVERALL GSNCER RISR FROM:HOUSEHOID EXPOSURE TO CtlGARETTE: SMORE IN CHnDHOOD AND ADULTHOOD ABe period of ezposure 'i, No Childhood AdultDood exposure o017' eatj^t Both Aaiar nmoknr . atieau . 22 63 25 92 :,,,nlrols 38 58' 37 65 Oddsr.tio 1,0 1•1. 2-6$,S 1•6 Noa.nwaRen Panients 32 44 33 53 Controls 61 66 35 33 OddYratio 1•0 1a$ 3•1$,S - 1•S , •E:paure m amotiea matLc. 6rLw; a aan 6ws6old mem6e,:durioa etifd6wd.. tFiposure w 9met,y epou.e. #S~ta,lioee dilraeum M.ara eit.irh.pm6d:apsm and m e:psun. (V<o' a5).tp-o tnl. SSu„aieo4r, eepzifiom lmur amt m eapawc, ezp.un, m aelry aer cme per,od, . e:poaon 0 baL ome perqd. (I' Yor v. ad.pC.p-o1):.. 867 "trtmatcd" age group was chosen because a primary aim of ourr study was to evaluate effects ofrnothers' smoking. Since few women smoked before 1920; we studied cases whowere younger than age 60 in 1980. It will be several years before cancer risk associated with tnothen' smoking can be evaluated for older persons. The distribution of cancer sites studied resulted from methodologic decisions and hospital referral patterns which werediscvssed in the papers. Nonetheless this group was not preselected on the basis of any prior hypotbesis, and it is not easy to see bow inclusion of more ancers that are "unrelated"'to cigarette smoking would lead to an inflated overall cancer risk from passive smoking.. Higgins cites our finding of a leukaemia effect as onefor.rhich "any biological explanation is, to say the least; obscure". On the contrary, cigarette smoke contains chemicals known to be leukaemogenic, and has been associated with increased leukaemia risk in many,2-5 although not all° studies. Some oftbese studies find an apparent dose-response, especially for rson.lymphocytic kukaemiu.2-s•5' The 40-50% iltcrease in kukaemia risk among smokers is much smaller than that reported for other sites such as the lung, which may account for the lack of interest in smoking and leukaemia. Experimental and biochemical data from a variety of sources (refs 7 and 8, and discusse d in our papers) are also consistent with a possible usoci>ttion between leukaemia risk and!ezposure to cigarette smoke. There is one final poiat: Our study was not designed to compare the effects of active and passive smoking. Our method of choosing friend controls inadvertently matched for active smoking, which ezases the possibility ofobservirtg an effett due to active smoking. Higgins' comparison of the effects of aetive and passive smoking is not meaningful inour data. The questions raised by Burch and Higgins do not persuade us that our study is "seriously flawed". We stress the need for additional studies and strongly urgeiavestigators to consider that the range of possible effects from active and passive exposure to cigarette smoke may be broader than has been thought. Nkrion.u teomm of. D.. P. SANDI.FR Enwroomenu! He.hA samcn, A. 1t Rne.rch Trisngk P.rk, • Q'[l,eox NbnbCymlJn.27709;.USA R...B. EVER.SON 1. SndkrDP: E.er.oo RB, R'IkmAl.i Pa.ire ~ne In~a6'rhliood and onm ri.R. An J.Ep.4-1 1965; 121: 37-4a. 2: kahn HA. The Doro uvdy odmrok,neand romnd,p.®onr US v.,er.o.:. Re¢rn on a- 5 yun of oMerrs"wn. N.N G+u+ lnrr Me.or, 1966, 1a: 1-125. 3. Hatmmood ~EC, Horn,D. SmnRme vd deeth r.,n:. Rrpon ee fortTfmnmmtSj or fo,dov-upp of 167.763men. aAMA 19e5, tM: 129a-M.. •: tD,llume RR,.. Hem la': A.ocuLpn of ooce, ,mn .,rL rabaomd .kohiol' ooo.umpewn..nd .oc.ir<couanu qnw of pa,ena lnsen- audy.remVr TL,rd Nuionsl Gnnr Sur.ry. ).Nal Crwur G,r 19.77+M: 525-17: S. P.lrentiu{er RS, WmrAL, HydeRT. Chnetterw:w m.rovtbpredKtiw a[.dul,- ae.n malian.nr lrmp6om.., .cf„wmu, amd kuL~, Br,d oommueiorrao.. J.NwLCG.n.l.u 197a: M: a9-92. 9. Doll R Peo R. Monarry m rsl.rwn~.romoo4ng. 207ean otwv+usas m m.k Dntyhh donon, L, Med.) 1976; u: 1525-M ~ 7. YnoraLr E, Ame. BN. CaoeetMra,m ofmuuarn hem mme b7 abwycwn ah. the fonpnlc ee~m xAD•2: C~prene emekcn i..e muuBeoicwrer. Aee NalAndSn ' US! 1977;,74: 3555-59. a. DeMarini DM. Genomaia7 of oo6.oromke and wb.aeede aMm..ee M.ur Aes 19a3; I14::S9-B9: REDUCED RESPONSE OF URAEMTC BLFEDING TIME TO REPEATED DOSES OF DESIMOPRESSIN StR,-Treatment with dessrapressin (l-deamino-8-D-arginine vuopressini DDAVP), shortens prolonged bleeding times in patients with von lS?illcbrand diseue,l platelet deCects,2 and nraemia,}5 and in healthy subjects.2 The bleeding time correction: by desmopressin has been attributed to the raising of plasma concentrations of high-moletvlar-weight forms of factor V1IIi (FVIII) related antigen and von Willebrand foctor (vWF) activity.° In uraemia FVIII-va'F concentrations are normal to high even before desmopressin; here, the presence of abnormal multimers may explain wh~ normal multimers, provided via dcsmopressins or cryopreripitate 7 are effective. Published experience with desrnopressin in uraemia is limited to short-term responses. When we gave ayoprecipitate (30 imL!kg) in I