Tobacco Documents Online

Epidemiology )uly 2000, Vol. 1 l No. 4 SMOKE EXPOSURE AND FETAL GROWiH 431 TABLE 4. Risks and Adjusted* Odds Ratios (AOR) of Adverse Outcomest by Smoke Exposure luw Brrth Weight Small for Ge~tariouaf Agc Pretennt Vcry PrerennY Smoke Expraurc N$ % AOR 95% CL N§ % AOR 9i% CI. N§ % AOR 95% CL N§ % AOR 95% CL Vunc9 8S 2.9 173 6.0 174 6.0 58 2.1 Low ETSII 20 3.2 1.0 0.61, 1.7 42 6.7 1.1 0.7,, 1.5 46 7.4 1.1 0.79, L6 21 3.8 l.5 0-90, 2.5 Htgh ETS 7 5.2 L8 0.82, 4.1 5 3.7 0.62 0.25. L5 13 9.7 I.6 0.87, 2.9 7 5.5 2.4 1.0, 5.3 Lnw .mok,ng 7 3.9 1.3 0.57, 2.9 19 10.6 2.2 1=, 3-7 5 2.8 0.42 0.17. 1.0 2 1.1 0.43 0.10.1.8 Moderare smokmg 9 4.8 1.7 0.79, 3.4 24 12.9 2.7 L 7, 4.4 10 5.4 0.80 0.41.1.6 4 2.2 085 0.30, 2.4 Heavy smnking 7 8.1 2.6 1. 1. 6.3 16 18.4 4.5 Z._c, 8.1 8 9.2 L3 0.56, 2.8 5 6.0 24 0.91, 6.4 • Ad/,weeA hi lV>ric regrenl,n fur prior prep>..,cy hmory, race. budy ..aw ind.z, It(e euer,n .rd >iucxi,m. t 13W - In. b„th.eRht 412500 pn). SGA - amall ( gnrariund age (<16rh pemmrrle). prererm (O] weeks gcrannnal agel. v<ry prc¢nn (<35 weeks). r(Armrari.w. foc bmh prerenn caterorles i> term bethy z37 wceks. f N n!.v n,.m6er .R~i by e.ch w,cmne i. each c.pµssure grm,p- 9 Referenr cnqury. / ETS = en, vunmend robcr:o sawkec non-whites than in whites. This finding was not consis- tent for active smoking, but when we examined individ- ual race categories, we found that blacks and Hispanics whowere heavier smokers had crude rates of LBW that were six times those of non-exposed based on very small numbers. There was a pattern of excess risk of pteterm and very preterm birth with both high ETS exposure and heavy smoking in non-whire.s (Table 5). Stratifying by maternal age (Table 5) indicated little age mcxlificarion of the association of SGA with heavier active smoking. The risks of LBW associated with active smoking were greater among older women, bur there was little difference in the association with high ETS expo- sure by age. The associations of preterm and very pre- term birth with both active and passive smoke exposure tended to be greater in older women. DISCUSSION Our strong finding of a decrement in birth weight with active smoking is consistent with the literature and was not affected materially by the exclusion of ETS-exposed women from the comparison group. Earlier findings re- garding ETS and mean birth weight have varied, with some studies reporting decrements of 30-100 gm and others showing little effect.1"" ETS exposure has been defined in a variety of ways in these studies and effects may be missed if high exposure is not examined, as suggested here. It is highly likely that the proportion of ETS-exposed women that were highly exposed was greater in studies conducted earlier and in areas such as Europe and Asia where smoking is more pervasive than in California.'s,lt-t9.zt-n,ze-in Offering further validation of our data, we confirmed the consistent Findings of an excess risk of low birth weight, and particularly small for gestational age, with huavier smoking.'-4 While smoking is thought to be more strongly associated with SGA, there is sufficient evidence to indicate it is modemtely associated with preterm de7iverv °.9-1e.f1'N In support of our finding of a dswbled rate of very preterm birth (<35 weeks) among TABLE S. Potential Modifiers of the Relationship Between Smoke Exposure and Fetal Growth Adjusted Odds Ratios (AOR)' and 95% Confidence Limits Low Birth Weight Sma11 for Oesrarional Age Prertrm Very Prererm NJ AOR 95% CL AOR 95% CL AOR 95% CL ACNt 95% CL Race Whitc Unca~ased %Y 1906 2.5 4.6 5.4 1.7 High ECS§ 83 0.44 0.06, 3.2 1.0 0.40, 26 1.1 0.26, 4.9 Hcavy smoktng 76 Z.6 1.0.7.5 4-0 2.0, 8.1 1.0 0.40. 2.7 2.0 0,56.6.8 All o.hcra Unexpnsed %j 980 3.8 3.7 7.4 2.8 HighETS 51 3.8 1.5,9.8 1.3 0.48,3.3 2.4 L1,5.5 3.8 1.3,10.7 Heavy smoking 11 1.6 0.20. 16.1 6.3 L6,.2A7 2.2 0.44, 11.2 4.1 0.74, Z2.4 Age <30 yeao Unexpa>ed %t 1711 3.2 6.0 5.8 2_0 High E7T', 91 1.9 0.73, 5.0 0.54 0.17.1.8 1.3 0.46, 2.6 22 0.75, 6.6 fi<avy smokmg 58 1.6 0.48.. 5.7 5.0 24, 10.2 1.0 0.35. 29 1 4 0.11.64 130 yeon Unexoosed %t 1176 2.6 6.0 6.4 11 High ETS 43 1-6 0.36, 7.1 0.74 0.17, 3.2 2.8 11 6.6 2.7 9.7 0.74 Hca>y smokmg 29 5.3 1.4. 19.6 3.6 1.2, 11.1 1.6 , 0.47, 5.7 3.7 . 0.98, 13.8 • Adin+a.1(or prcgrwr,cy hnmry. bodr mau n.dc.. educ..im, I,(<evenra (anJ eace In ag<.rtrsifl.d modelsl. W,rh no>,ookc e.r.a,.rc ss rde.mce g,oup in cath>rnrumh t Sne nf exr•rure group: nore rhn u smalkr by abnut 4% for very FRterm as in(ann ban in wreks 35 and 36 a.e exeluded. I Pcrum :J&ocd e,,,.,ng e.pwd (+dc,em gr.,up] ir> each mmnn. I ETS = mv,ronmennl mb:.w uroke.

Epidemiology July 2000, Vol. I I Nn. 4 SMOKE EXPOSURE AN17 FETAL GROVLTH 429 TABLE 1. Percent Distribution of Subjects by Level of Smoke Exposure for Various Demographic Categories, Pregnancy Outcome Study. 1990-91 Smoke Exposure (%) Variablc N None Mrd ETD` (I-6hr/day) High E'IS (-7hr) Low Smokung (<5cigs/day) Mai Smoking i5-10) Heavy Smoking (>10) R..ce White 2715 70.2 1~1.7 3-1 5.3 5.9 2.8 Nomwhire 1382 70.9 20.2 3.7 2.5 1.9 0.8 Age <25 1091 59.8 22.3 4-7 5.8 5.7 1.8 25-34 2584 73.9 13.1 2.6 3.9 4-3 2.2 ~35 424 76.4 99 4A 3.8 3.3 2.1 Previous pregnancy hisrory No prinr 1061 73.5 1 i.1 3.5 2.7 2.4 0.9 1 pg I k.ca 21147 69.8 14.7 3.2 4.6 5.L 2.6 -21,mc 163 64.0 11.7 3.3 9.3 7.7 2.2 Maritnl sratus Mamed 3460 74.1 14.0 2.9 3.4 3.6 1.9 Nor mnrricd 639 50.6 21.9 5.3 9.7 9.4 3.1 Elucarw.n sHigh achcxd 1663 57.3 20.1 4.9 6.4 7.4 3.8 S.me cullcge 1300 i 1.9 15.4 2.6 4.6 3.8 1.5 C.dlebr goduare 1132 87.6 8.0 1.6 L2 1.2 0.3 Exnplo'm ~ during pregnancy EmP uy 3211 61.8 17.2 3.8 4.4 4.7 Z.I Nor employed 895 801 R4 1.2 4-3 3.8 2.2 Alcohol consumprion Nonc-0.5 3842 71.9 151 3.3 4.0 3.9 1.7 >1 dnnk(wczk 256 48.1 17.2 3.1 10.6 13.3 7.8 t.nffcinc cowumprirm Nnne 2074 79.5 12.8 2.5 2.6 2.0 0.5 5150 myfday 1578 64.9 k8.4 3.9 5.1 5.4 2.1 >150 mg/day 447 46.8 15.2 4.7 10.3 13.4 9.6 Life rnrs 0-Z 3361 73.1 14.7 3.2 3.5 3.8 1.7 3-4 573 62.5 16.9 3.1 8.0 5.9 3.5 z5 165 44.2 20.0 4.2 9.1 15.8 6.7 Wnrk ttresst High 242 61.6 19.8 5.0 7.0 6.2 0.4 le,w 2887 68.4 16.8 3.6 41 4.6 2.3 13.v1v mas+ index l.ow (0-19.0) 430 68.1 13.7 3.7 4.9 6.3 3.3 M<J (19.1-27.3) 2993 72.4 14 5 3.0 4.1 4.2 1.7 liigh (27.4+) 627 62.5 19.9 4.2 5.4 4.8 3.2 Tnrzl'- N 4099 2887 625 134 180 186 87 94 100 70.4 15.2 3.3 4.4 4.5 2.1 `ETS = environmmtal tnbacc,.mokeo t L3nu mly fnr "w=an dw warkad aurncnmc durin6' FrcRn.ncv. I4Swne varublea do nnr add ro nxsl due ru msalne rabws nr <vclutim N x,.ne valuei e 6w-ry lighr ak.drnl <unwm<n. quarters of the women worked during the first trimester and 60% had some education past the high school level (Table 1). In rhis cohort of women receiving prenatal care at a health maintenance organizacion, 3.2% of infants had low birth weight, 6.9% were small for ges- tational age, and 6.1% were burn preterm, 40% of which were very preterm (or 2.4% of total). Of those with low birth weight, 65% were born preterm. About two thirds of the womet, reported no smoke exposure, 18% were nnn-smokers with some ETS expo- sure (or 21% of non-smokers reported ETS), and about 11% smoked during the first trimester. As shown in Table l, reported exlxa,ure to any form of robacco smoke decreased with increasing age and education. Asians were the least likely to report any smoke exposure whereas blacks were most likely to report exposure, particularly to ETS. Women who were not married were more likely to be exposed to either form of tobacco smoke, and women who worked were more likely to be exposed to ETS. Consumers of alcohol or caffeinated beverages were more likely to smoke, but they were not more likely to report ETS exposure. IvIFAN BIRTH WFXiHT Infants of active smokers had reductions in crude rncan binh weight on the order of 100-200 gm, with a dose response rrend. These weight decremenrs were greater in the adjusted models (Table 2). Adding gesrational age to the model still yielded large decrements in weight for all smoking categories. Mean birth weight varied little by FT.i exposure (Table 2). When we examined home and work ETS expnsures separatcly, we found a slight weight decrement with high exposure at home, but the confi-

428 Windham et uf they called to make their fn.t prenatal appointment. A prospective study design was used to obtain information on exposures early in prrgnancy to avoid problems of reporting bias Eligibiliry criteria includ.d age 18 years or older, 12 weeks gestarion ru less and English or Spanish speaking. Telephone inrerviervs were completed within a few wceks of initial .untact for 5,.342 women; 18% of the rotal refused, about 10% were ineligible and .3% could not be re-contacted. The median gestational age at interview was 8 weeks. Pregnancy outcomes were ascertained prtmarilv from compurerized hospital admission files as well as by ab- srraction uf medical records fnr pregnancv losses. Less thau 1% of outcomes could not be determined. We linrited this analysis of birth weight and preterm delivery to singleton live births. Birth weight was abstracted from the birth certificate fnr 4,454 births, which excludes 103 (2.34G) that could not be linked. We calculated gest,i- tional age using the date of birth as reported on the hirt h certificate and the lst day of the last mcnstntal perird'as ascertained at mtervicw preferentially, to maintain dat- ing used throughout the intcrview. These were com- pared with gestational age as cepoued un the birth certificare, and in 48 births we suhsriruted thr certificate gestational age tu correct:rn iorpruhablr calcrdated ges- tational age. We defined low birth weight (LBW) as weight less than 2,5CY? gm. We examtned the usual classification nf preterm delivery as less than 37 weeks' gestation, as well as an earlier cur-off at less than 35 weeks ("very pre- rerm") to rdent ify infants at even higher risk of neonatal morhidiry arid murtaliry.'t We defined small for gestational age (SGA) as birth weight less rhan the 10th percenrile for gestational week. For this purpose, wc calculated a standard from the weight distribution of thc over 2 million singleton births that occurred in Califor- nia during years comparable with this study (1990-93) separately fnr malrs and females, for each gestational week from 24-44. EXPC~tiuAE ASCtv"MENT The interview asked abtxrt a number of topics including demographics (eg age, tacee education); reproductive his- tory; job characteristics; phyxical and psychosocial stress; and water, tobacco. caffclne, and alcohol consumption. Most of the consumprion yuestiuns (including amount smoked) were asked abuut two time periods: the week before inrervicw, (referred to here as "during" the first trimestcr (if pregnancy) aed the week at last menstrual period ("lefnre" pregnancy). The smoking status of the infant's father was -also asce rta ined for the same two ti nre periods. The number of hours per day that non-smoking resprmdenrs were near omer people smoking both at home :ard at work, since the last menstrual period, was also asked. We sunnned these two variables to estimate rntal daily ETS expo.sure. These ETS variables were examined on)v among women who reported nnt smok- ing both hefure and during pregnancy- Thus, women who reported smoking at one of thc timv inirrvals, but Ep~demidogy )ulp. 2000, Vol. I I Nn. 4 not the other, were excluded (N = 35!) to avoid pn- rential misclassiCication. ( I nfants of these women did not generally appear at increased risk for any adverse nut- come.) We creared a categorical variable with six levels of smoke exposure (excluding an additional four women missing one of rhe ~moking variables) as follows: Among nnn-smokers: 1. No smoke exposure (reference): none or <0.5 hours ETS exposure/day 2. Moderate ETS- 0.5-6.5 hours ETS/day 3. High ETS; ?.7 hours ETJ'/day Among smoken irrespective of ETS exposure: 4. Low smoking: <5 cigarertes/day at interview 5. Moderate smoking! 5-10 cigarettes/day at inter- vieu 6. Heavy smoking> 10 cigarettes/day at interview STATISTICAL ANALYtitS We calculated crude rates for the four categorical end- priiuts as well as rate ratios and adjusted odds ratios and rheir 95% confidence limits (95% CL) by smoke expo- sure, relative to non-exposed. We assessed a variety of potential confounders, identified from the literature and univariace analyses, in logistic regression models (ie the "full" model) for each endpotnt These included: mater- nal age, proo-pregnancy body mass index (BMl = weight/ height'), parity and prior pregnarxy loss, race, educa- tion, marital status, employment stanrs (and hours worked), stressful life events and social support (assessed using modificarions of standard instromenrrs),rs and caf- feinated and alcoholic beverage consumption during the first crimester. We used the change in estimate methodr" to identify covariares whose exdusion from the full models changed the adjusted rwlds ratio for high envi- rontnental tobacco smoke nr heavy smoking by 10% or more, to create a"rcduced" mudel. The variables that met this ctirerion wcre pregnuncy history for all end- points, race for small for gestational age, stressful life events for ptetenn delivery, and BMI and education for very preterru delivery. We included all five in the re- duced mrx9els used to calculate the adjusted odds ratios (A(1R) presented, unless othesavise indicated. Potential zffect modifiers of a priori interest from previous smoking :audies were maternal age and race, assessed by stratifi- cazinn and separate models. We detemrined mean birth weight for each exposure category, and we calculated the difference from the non-tobacco-exposed group. We used multivariate re- gression models to control for all the variables in rhe full model noted above. In addition we added gestarinnal age to some mrrdels to examine effccrs on weight, control- ling for rhe influence of age at delivery. Results -lhe mothers of these live births were on averagc 27.8 years old, with most (90%) under 35 years. About two tirirds of the women were white, rme quarter were nul- liparous, and most were married (83.4°4,). Over three

T 430 Wmdham et <il Epidemiology July 2000, Vol- I1 No. 4 TABLE 2. Mean Birrh Weight and Differences by First Trimester Smoke Exposure Smake Er(ts,.re N Meen (g) (SD) Adjust J Di(terence• v5%CL f:ouv 2887 3514.1 (533.Y) cet Msdrmne ETSt (IE hr/duy) 675 349i8 (5255) 068 -46.7,48.1 High ETS (z7 ht/day) 134 3516.6 (662.7) 8.2 -86.1.1025 Low smoking (1-4 c~s/day) 180 34L0.5 (567.2) -14L4 -225.3.-57.5 Mtd srrn,king (5-10/day) 186 3367.7 (482.5 ) - 144.3 -227.4, -61.2 Hea,y svukiuy (>10/day) 87 33(Z0 (682.6) -'_38-3 -J58.6,-118.0 dence intervals were wide. Examining an even higher ETS exposure level, infants of the 28 women reporring 12 hours or more per day had a weight decrement nf -128 gm (SE 101) or -88 gm (SE 103) when adjusred. Had we used paternal smoking status as the ETS expo- sure variable, tlte adjusted difference in mean birth wcight wnuld hee about -32 gm (95% CL =-8l, I8). Stratifying by race, infants of Asians and Pacific Is- landers who reported high FT'S exposure (?7 hours/day) had crude birth weight reductions of up to 500 gm. Infants of exposed blacks had about a 100 gm reduction, Htspetnics had little reduction and exposed whites had somewhat increased weighrs, compared rvith unexposed women nf the same race. (A weight reduction of about 100 gm was seen among whjte infants at the higher ETS exposure level o(z l2 hours/day.) Because of small num- hers in some .d these categories, we compared whites with all other races. The nnn-whites had much greater absolure and relative crude weight decrements than whites with both high ETS exposure and heavier smok- ing (Table .3)- Including gestational age in the models reduced the magnitude of the weight decrements more tnr non-whites than for whires. Stratifying by matemal age revealed a greater weight reduction with heavier smoking among older women (Table 3), which was magnified at an even greater smoking level (>20 ciga- ettes/day, datx not shown). This age modifteacion was nor :as apparent with high ETS exposure (Table 3). Low BIRTH WEIGHT, PRETERM DELI\'ERY AND SMALL FOR GFSTATR)NAL AGE Increasing amount smoked was associated with an in- eteased risk in low birth weight and small for gestarional uge; infants uf heavy smokers had adjusted risks 2.6-4.5 times those of unexposed women (Table 4)- The risk of pretcrm birth was not greatly increased by smoking, but the risk of very prererm birth was more than doubled among heavier smokers. High ETS exposure was associated with an increased risk of LBW on the order of that of low-moderatc smok- ing, which aas changed little by adjustment (Table 4). The adjusted odds ratio for any ETS exposure (>1 hr/day) and LBW was 1.1 (95% CL = 0.71. 1.7). The increased risk from high ETS exposure appeared more sirongly related to preterm delivery, partieularly very preterm, than to growth retardation (Table 4). The a.ljusted odds ratio for any ETS exposure and preterm dalivcry was 1.2 (95% CL = 0.90, 1.7). For both low b.rth weight and preterm hirth, risks w•cre some.what greater with ETS expasurc at home rarher than work. Stratifying by race (Table 5) indicated that the risk of LBW from high ETS expoeure was much greater in TABLE 3. DiHerences in Mean Birth Weight* by Smoke Exposure among Demographic Sub-Groups High ETS L17 hr/day) Hea.y Smoking (>10 dgsfda.) Wrl,ht Di(ference W<ighr 9`.%C:L Difference 95% CL Maremal tacrt White Cmdc 89.3 -27-7. 206.2 -170.5 -292.5. -48.5 (N = 2715) Adl.t 97.6 --21.8, 217.0 -159.5 -287.5. -31.5 Adt-§ 89.7 -18.5, 197-9 -158.8 -274.8. -42.8 All others C-rudc -122.4 -274.5. 29.7 -687.1 -1i)D8.2, -366.1 (N= 1382) Adj.S -1189 ' -272.3. 347 -595.9 -939.1, -252.7 n al M Ad(.§ -22.9 -162.1, 116.3 -437.7 -749.1, -162-3 n rn age <30 Cmdc 19-7 -92.7.13L1 -10L7 -247.I.37.8 (N - 2574) AdjFf 31 4 -8L7 144.5 -108.4 -2529, 36.1 Adtll 45-3 -57.Z 147.8 -108.1 -239.0, 228 ,30 Ciudc -20.6 -188.i. 147.4 -39Z-1 -595-5, --188.1 N (N = 1523) Adj.T -3.4 -1720, 165.2 -440.0 -651.1. -228.9 Ull Ad'1.1I 85.9 -664 238.2 --357.1 -5480, -166.2 0 • l:.,myav..n +s grwF -th no ,e, i N, 6u hi¢a en.ironmrnul tokac nke .ap.w.e ,n ram. co amokc and hn.y n avan. ndine, m.pcc,i.ahas: :,mong whi.u (eJ.76/. orhee (S Llq: ag. <30 (91 58); agc .30(4329). ~ f 4di-rc.d ks aqs~. hdy m+n inde x. ps<8nanq h:uo.y, cd-awn. lifr evmo, oNeinr' and akdwl e.nrytw,,. 9 AJfuarrJ tnt (3) rFcse. plut uest 9 Adrwased Fac race. body mass ind auaral sye. em. Cregnaancy hura.y, eduor:.m. nw¢aI ac,m ., Lfc e.enn, caffein. =nJ alrn6nl nm mpflon. N II Adjusred (.w (5) ahove, plus erer ai«,.1 ape. 0

Prenatal Active or Passive Tobacco Smoke Exposure and the Risk of Preterm Delivery or Low Birth Weight Gayle C. Wirtdham,' Barbara Hopkins,' Laura Fenster,' and Shanna H. Swanz We examined the association of exposure to environmental robaccn smoke mnh birth weight and gestational age in a large, pnnpective study. We also compared these endpoints between infants uf active tnaternal smokers an;l those of non-smoking. non-EfS exposed women. Pregnant Women were interviewed by tclcpbuue during dte first trimester, and pregnancy outcome Was determined for 99%_ Among the 4,454 singleton live btrdu that cuuld be linked to their birth certificam, we con- firmed mcrea.ed risks of low Mrrh weight and small for gesta- tionai age with heavier maternal snioking (>10 cigarettes/ day), as well as nonng an increased ris6 for "ve~ prcrerm" birth (,'3i weeks). These :,ssociations werr generally stronger ammng infants of older (?30 years) than thaue of younger mothers, as welt as among non-whites. High envirorunental tobacco smoke exposure (?7 hours/day in non-.mokers) was mtuderately associated with low birch weighc (adjusted odds ratio (AOR) 1.8, 95% confidence limits (9SR, CL) = 0.82, 4.1 ), preterm birth (AOR 1.6, 95% CL = 0.87, 2,9), and most snouglv with very preterm birrh (AOR 14, 95% CL = L0, 5.3). These associations were generally greater among non- whites than whites. The data support carlier studies mggrsting chat prenatal environmental robacco smoke exposure, in ad- dinon to maternal smaking, affects inf.mt health. (E-ndemiology 200411:4Z7-433) Keywords: maternal smoking, passive smoking, environmental tobacvo smoke, low birth weight, preterm delivery, birth weight, small for gestational age. Maternal active smoking has been associated with a number of adverse developmental and reproductive end- points.1-4 Infants of women who smoke during preg- nancy are estimated tu have twice the risk of low birth weight or an average weight decrement of 150-200 gm at blrth, compared widi chose of non-smokers. The dec- rement in birth weight appcars m Fe due primarily to intra-uterme growth retardation, and to a lesser extent to preterm birth.5-9 Because of these relations, exposure to environmental tohacco smoke (ETS) has been of increasing concem.10a1 Persons exposed to environmen- ral rnbacrn smokce are subjected to most of the same coruNtuents as tht.so contained in mainstream smoke, hut the pattem and amounts of exposure differJ7 Studies of environmental tobacco smoke exposure tend to show a decrement in mean birth weight of a small magnitude and suggest a slight tncrease in the risk of low birth weight.t'"-'h Many studies, however, were based on only crttde measures of ETN exposure, such as parernal smok- ing. Fewer snulies have examined ETS exposure and pmcerm birth, some of which found moderately in- cre.ased risks.","," Some studies of maternal smoking or ETS exposure have shown greater effects in older women-'7-" Further- more, a few studies have indirated that blacks have higher levels of cotinine, a nicotine metabolite, than do whites at the same reported smoking level, suggesting possible differences in metabolism.~-u Age and race may thus modify effects of tobacco smoke exposure. This report has several objectives. The primary one is ro examine the association of ETS exposure (or passive smoking) with birth weight and gestational age in a prospective study conducted during the first trimester of pre.gnancy, with ETS exposure reported at home and work. Our second objective is to estimate the effects of active smoking using a non-smoke-exposed comparison group, instead of including possibly ETS-exposed women in the comparison as is done in most studies of active smoking. Another objective is to examine these as.snci- atioru among demographic suh-grou(ts defined by mater- nal age and race. Subjects and Methods From rhe'Reproiw,,.. Ep:J.,,.,,,,l.q-y Sec,.on, tlprtmenr nf Hrahh Sr,vi~. Study methods have been blished elsewhere,tl but are tAlaland, Czhfurn,a, n,d rhe 4 ;n,veniry oi Mav,ur~, CalunSm, Mi,ui_ pu described briefly below. AJJr,ru cone.ln.nd<o<e r.r U.,rtr C. Wmdha,n, ReproEn:orc Epidcmoloer iec,i.n, [h~panmcm . n Hcal, I, 4r-,c~.,. 1515 Cl.r V rs~r, Suis 17M. tbk6nd. <SlnLmia. 9MI2. Subo,bu-J lun, 21. Ivo9: i„ul .e..,no a.«Lnd b„~,•ry 4.20011 G,pmgh. ti'_tVC hy I.,ppirc..u v4911- 6 W,Iktr,., Inc. St/BJECT RECRt11TMFNT ANn ENOICJINTS Pregnant women were recruited during 1990-1991 from a large pre-paid health plan (Kaiser Permanente Medical Care Program) in three regions of California at the time 427

432 Windham et ul heavy smokers, other studies have reported a greater association of smoking with delivery before 33 weeks gestation than with moderate preterm delivery (33-36 weeks).'r"" High ETS exposure wa:; moderately related to low birth weight and particularly to very preterm birth, but not to small for genstational age. Other studies of ETS expeuure and low birth weight or small for gestational age have found varying results from no effect up to about a doubling of risk.'o.it.u.n.ic.rr.iv:c.n.zs A nreta-analysis of studies published by 1995 reponed pooled adjusted odds ratios from I.1-1.4 for these endpoints." Studies of preterm delivery and ETS exposure are fewer and several have found moderate associatinns, but none has exam- ined earlier prererm binhsJ1."=' An effect of ETS on very prererm delivery that is of a similar magnitude as that of heavy smoking appears inconsistent. Non-smokers exposed to ETS generally have levels of cotinine, a merabolite of nicotine, that are 1-2 leveLs of magnitude lower than acrive smokers, suggesting that effects of ETS exposure should be corre- spondingly lower than those of active smoking. This theory presupposes a linear relation.hip, however, which may not be approprinte.ratsi" Furthermore. ETS is com- ptsed of hundreds of compound+ contained in exhaled mainsrream, as well as sidestream, smoke of which others besides nicotine are probable toxicants of interest (eg carbon monoxide, cadmium, polycyclic aromatic hydro- carbons)1° These compounds are nor all present in the same relative ratio as nicotine in mainscream versus sidesrream smoke," nor in aged FTS, so rhar a linear trend of effects by cotinine level might not he expected. ETS was not associated with SCA, as smoking is, but SGA has a different etiology thm preterm delivery. A recent study of breast eancer similarly found an odds ratio for passive smoking that was of the.ame magnitude as that for acrive smoking.'0 The associations we found with high ET5 exposure appeared limited to non-whites for the most part, which was not explained by a differential exposure distribution within this high category. There was generally a similar pattern of racc-modification for heavy smoke exposure, bur the number of non-whites in this group was small and estimates less stable. A previous U.S. study19 rc- ported that ETS and LBW were more strongly associated in non-whites than in whites. Several other studies",4t-" found few racial differenccs in rhe effects of active smok- ing on low birth weight or small for gestational age, btn mean birth weight does not appear to have been exam- ined in this way. There is scxne evidence that blacks menrb,li:e nicotine differently than whites.x`" Dara on orher specific ethnic groups is not readily available, so our preliminary data suggesting some differential effects by ethnicity (including Hispanic> and Asians) is of in- n'iesr. Our finding of generally greater effects of heavy smok- ing amonR older mothers did not appear to be explained by a differential distribunon ot smoking within this high category and i+ supported by several other studics.r-"L9e1 Such an effect may Ix due to cumulative exposurc over Epiderniology July 2000, Vol. II No. 4 years or perhaps to diminished reserves to compensate `or the toxic effects of smoking among older mothers. A :ecent U.S. study'-7 found some age modification of ETS exposure with both LSW and preterm delivery in a low income population, but two other studies did not and we tound it only for prererm birth.'" Although our ascenainment of ETS exposure was more derailed than in many previous studies, it was based on self-report of hours exposed and did not include cxposure outside home or work. The associations of low birth weight and prererm birth with high ETS exposure were not dependent on the cutpoint selected, as similar increases in risk were seen at 5 or more hours/day. Decreases in mean birth weight, however, were not observed until ETS exposure levels were much greater, highlighting possible difficulties in across-smdy compar- iutns of self-reported exposure. Another limitation was that exposure was assessed during the first trimester of pregnancy. Many pregnant women have already changed their smoking habits by this time!6 but in an incerview we conducted during the thinl trimester (or post-nacally) on a sub-set of this sample, about one third cf first-ttinrester smokets did not report any smoking in the last 3 months of pregnancy; some of this may reflect r.trospectlve reporting errors. Nevertheless, the associ- arions we found based on firsc trimester smoking may be underestimated (eg a conservative bias) if quitting later decreases risk. It is unlikely that many women started smoking later in pregnancy, so associations with ETS exposure examined among non-smokers should not be greatly affected. We did not ask about ETS exposure in the third trimester re-interview and there is little other data on the consistency of ETS exposure during preg- nancy, so some misclassification may exist. Another possible source of error is misclassification of nurc„me. We did not have information from medical records to verify gestational age. As several authors1P.4s have recommended updating fetal growth curves and California has an unusual racial distribution (more His- p.mic births), we calculated a new standard for Califor- nia- The weighc-for-age cutpoints were similar to, but consistendv slightly higher than, recent U.S. and Cana- d an srandards.^^s This HMO-ba.sed popularion had gcnerally low risks of the adverse outcomes examined, pshap due to early, routine prenatal care and/or lack of extreme socioeconomic disadvantages, which limited the power for sub-group analyses at higher smoke expo- sure levels- Although we were able to examine many potential confounders, including some on stress, exer- tijn, and socioecunontic factors, residual confounding may influence the results. We had no data on history of other diseases, particularly sexually transmitted diseases, which may play a role in preterm delivery, or on the speci/ic causes of prerenn birth. The strengths of this study arc many, including the piospective design, nearly complete follow-up of all ptegnancy outcomes, and a population with equal access te medical care that should decrease possible confound- ing cffccts of socio-demographic variables on exposure st-ms.

Epidemiology July 2000, Vol. 11 No. 4 Acknowledgments We rhank Ki.rren Waller 6- her ..orL in arigntng geurlonal age and fmal Ivcynancv ourcome. Financial .upport from rhe Packard Family Fswmdarroo allowed rompleem •afdara collecrisn. We alro ackrwe.kdge she conniburtorss n( rhe Kaiser cbmo mvolved m panent rmuisnmt ad invnrigamrs ze Kaiser Divisinn uf Rcsenreh, induding Ca.henne Schae(er and Roben Hian. References 1. U3 L4partmenr of Healrh and Hunan Serwce (USDHHS). The Health C.ane.pcnas r! Smn6ng (nr Women: A Repon of dse Surgeun (lenual. DHHS. Atlanta. GA: U.S. Public 1 kdrh Srrvicc, OFicc uf Snxduug and Healrh. 198(1. 2. Amencsn Collega of g]bscerrics ard Gm«ology (ACOG) edu®.iasJ bulletin. Smoking arrd wvrnen i health. Int J Gyncad CHa¢r 1997:6@71- 81. 1. ITFanra JR. Lew RA. Effect d maternal cigarare smuking r,n pregnancy a.pli.-. ns and addrn mfana d.arh ayndrome 1 Iimily Pnce 1995,40: 185-3941rn 4- W.IsA RA. EXecrs uf maremal smnking m adverae pregnancy nuscome.: n of rhe erireru of camarinn. Human Biol 1994;66aJ59-1092. 5. ClimrnS.~M.resrul smukrng and feal gmwsh Assuc-urion of Rsprm3ucos< Healrh Prtafevkmah (ARHP) Clinical Pr«eedings 1996;11-Ie. 6. Kramer MS. Decermsn.vsa of low bsnh werghe merhodolugrcd assessmenr and meo-analyais. WHO Buller 1987;65663-737. 7. Kcame* MS. Sr.:ioscorunuic deimoinana d imnurerine grasvih mnrda- nnn. Eur 1 Clin Nuvi 199&52(Sq.529 S33 6. Wnglo SP, Mirchcll EA, Thorryvrm JMD,ClensennMS. Pord RPK, Ssew- are AW. Ruk faernn ht pms<m NM: e New Zealand scudy. NZ Ma31 199F;111:14-1a. 9. WlaMrg K. Henaksas TB, HcJegsaN M, Sechrr N). Smnking during pregnancy and prerenn birth. Be J Obseer Gynaecol 1996;103,800-605. 10. Natwnal Cancer Inssimm. Health Effects rI Expmure ru Envuonmennl Tnharrn Gm,4c The Rzporr J rhe. Glifnmia Envimnrnenml Pnsunion Agenrv. Srmhing and Tobacco Conrml. Mwsogr:q.ls Nu. 10. DHHS Pub. Nu. INIH)9:9-4645. BerAesda. MIL National Cancer lnscimm, 1999. 11. Wundhan OC, Eatsm A, Hnpki,n B. Evidersce (nr ar• xvocoriwr berween mironmeural mbacrs sri cepawre and birthweighn a mera-analyse and new daes. Paed'uv I'erirur Ep/demiol 1999:11d5 57. 12. Gucnn MR, Jenkins BA, T,nnk,m RA. The Chrmuvy of Environmemal Tobacro Smoke ebmr.uk,m a..3 Measuremene Ilo:a Ramn: Le,,is Pub. Ilahns, 1992. 13. Ahllorg G Jr. Heahh effecs if enviranmennl mlvscrn rrm n ehe of5pring d n.mamoking women. 1 Smnking-Re6red Dis 1994;i107-112. 14. Estrnsri R, Prehn AW, Cluisuanson RE Pasnve and acuve ma¢mal smakmg as muisured bY ssrum c•nmine: The eNesr nn binh weu,•RS Am 5 Puhl Healrh 1995:85:395-39& 15. Hald,w JE, Knighr GJ, Pahmaki OF, MeCenhy JE Second-mmexrer semm cammne kveb in n nsmoken in relarion ro bsrrh weighr. Am J CMsrn Gynecal 1988;159(2):191-481. 16. Fwrhr 1, Marn.ux 5, R.is,m J. Paaslvc smoking durtng pmgnnncv arsd the ruk u( deliverine a smaBdor-geunNnnal-age lnhnr. Am I EPSdumol 1994; 1390),294-301. 17. JerWchowaki W, Flak E. Con(nnsing the prrnasal effecrs of auive and powive mbarco smoking un rhe b,rrh weigbe o(chddren. Cenr Fiu j PuNic Health 1996N:2C1-2J5. 18. Lau:uuni F. B,n.an; S, Manmello E. Mnrcaldi L, Repuro R.. R,,.can A, Calvi .1, Cnrelkua 6. Effecr d puuve smoking dusing p¢gnuncY On aelecred perinanl par;srneren. Inr J F{odeminl 1990:19i41:960-9(s6. 19. Main.uwn AG, I leuxon Wl- Pxw,.•< smoke and lu»btrch weighe Arch Fam Med 1994:3:875-878. 2~. Murrm TR, Bracken Me. As.so.iar;<.n of Io.- b,rrh wcighr wirh p.aive smoke exposurc in prqmanq. Am I Epidcmiol 198QI24(4):633-642. 71. Marrlnes FD, W'righi AL, Tnus,usg LM, and the Group Healda Medical Asnsciaees. The eRecv nl pammal smohing on rhe birth ae,ghr of newMms w4une modrers did nur smoke. Am J Publa Heilrh 1994:N41489_1491. 22. Mathai M. Viiaywsri R, Rahu S, Jc3zeelan L. Pasar.e mammal smuk;ng and Hnh wcighr in a Snth Indun pnpulati,m. Br I CXeres Gynuuol 1992: 99(41:34-34 1. U. Nakamum M. Chhima A. Hiyam:, T. Kuhia N. Wad;r K. YenL• K Efkar 4 ryssrn .wnnking dunn¢ ~rrersanry on birth we;gA, and geemn;m: A pu{w- SMGKE EXPOSURE AND FETAL GROWTH 433 lacion-based prospeccive study in Japan. In: Aoki M, Hisamlchi S. Tominaga S. e.r.. Sn..king and simlrlN, t997. Proceedrnga of rhe brh World Cun" enee on Smoking and Health, Tokyo. 9-12 NovemLcr 1987. Amnerdam. Exceqtta Medka. Inremanonal Congmss Series 780, 1989;267-269. 24. Rel..gliaru M, do V Florey C, Bulumsr F. Exposurc ro enviraunenml nsha.rn amoke In nonaroking pregnanr women in relation ru birth weighr. Am I EPdemiul 1995;142:511-537. 15. Royuer 1M. Fieuens J. Boter F. limenes R. InBumce on (eral gmwrh of c:Pn»sre m m6acco smoke doring pregnanq. Acra Peed,an 1995:84a 16- ISI. 26. Rublrr DH, Krasibsikc.T PA, Le-nrhal JM. Waile e, Bager A. Flfecr of pauive smdcrssg on birth wcighr. Lancer 1986;2(8504):415-417. 27. AMumRa IB, Gn.mmer-5m».r L. Fanlon KS Eaposure m o,vironmanal s Ivscu emnke and birth <wrcome: lncreased e(leeu m pregnant .vnen agcd Sd years rx nkler. Am J Epidemml 1997;146:42_47. 28 Foa iH, Koepsc8 TD, Daling JR. Rirth weight ad snwkiny dmrng png. nancy-e6ece malificatlon by maremal age. Am J EpidrmLd I994;139: 1009-1015. 29. Wen SW, Goldenberg RL, Curter OR, Ho(fman HJ, CGvet SI'. kAvu RO. LlrB:ud MB. Sm~kinq, mnamal age. (enl gm•a.h. aml gestaci•uul age ar debvery. Am 1 Gbsset Gynecol 199P,162:53-58. 3C. Engl.d. PB. Eskenasi B. Chr,asiaeam RE. Black-whire di&erenc.a in rmm H.e Iwd. amung p.cgnanr snmen and subse9uent e(fece nn irdant bmhwe@tu- Am J PnM,c H.alrh 1994;84d419-1443. 31. US 1)epa.rmena •ti I lcalch avuf Hanun Senlces. Tubacco Usc among U.S. Raci.rllEdmic MlnoriryGruups-African Arnancvn., Amencan IMiam and Alaska Nariva, A.ian Americaru znd Pacdic Islaralers, ard Ilispanws: A Rcprn of nce Suryenn Genaal. At6nra, G.S: U.S- Oeparvmem of Health and Iiumsrs Servicn, Cenrers fnr Duease Gmrrol and Pmve.rri.n, Narional Cenrer for Chrons Disease Pre.emroey and Health Pnmsorion Ofhce an Smuh;ng vcl Heahh, 1978. 32. Klrb;moR MA, Levine RJ, Clemeas ID, [hrSmonian ft, Wilkina IXi. Sermn cminine eoncenrrannn and.el(-rc.pmrs3 wm,king d,.r.nq pe,lna. y. Am, Eryd.mid 1998;148:259-262. 33. W,n,lham GC, Penuer L Hupkms B. Swan SH. Moderate ovtemal and pare.nal d,mhol c•nunryriun ard ruk of spununenus abnrrion. Epdrmiol 1997.8:509-514. 34. Ber6wua G5, Paplern,k E EpidemiobagY u( Preeersn birth. Epdrrniul Rcv 1993.15:414_443. 35. Fessser L, Schaefer C, Maehur A. Hiarc RA, Pieper C, Hubbunl AE, Vrm Behn-n J, Swan SH. Pryr7robgic ureu in rhe workPlaee and s3.mtxneous aHnsicm. Am J Ei:,leminl 1995;142:I176-Nf3. 3G G,ec dand S. Modding arW variabk aclcctiu, in epi.kmWlugic nnahsb. Am J ILbhc HWth 1989;79:340-349. 37. Shio,u. PH, Kkbnnoff MA, Rhoads GG. Smok.ng ard drmk,ng durinq preg arscy; their ef(eces on Premrm birth. JAMA 1986:255:82-84. 38. Kyrklurvi-Blomherg NR, Cnarringiu.s S. Prerenn birth and m.,crnal smok~ trsg: uks relaced ro gesntional ege and ons.r nF deli•cry. Am 1 Oluer Gvr. cnl 1996;179,1051_1055. 39. Harh- BL, Frideao FD, Cnmer DW, E- JK, LeFevm ML. Bain RP, McN {lis 1) and rhe RADIUS Sndy Group. Derrrminann N prerenn dcliv.vy in I.w-risk p,c5nancics. 1 Cliu Fj,uluniul 1996;49:441-44A 40 Lash 71. Aschengnu A. Active aed passlve cip.eue smuking and the ocemrence of breast cancor- Am J Epidemid I999;149:5-12. 41. Wan¢ X, Tager IR, van V'unakis H. Speiser FE, Hannhan 3P. Masemal ms,kmg dunng pregnancy. urine r.nioirse ons, a.aJ birth wr- mn. A pre.pecnve cohnrt smdy, lor J Epidcmiol 1991:26:978-987. 42 Negg_•r. Y, GnldenRerg RL, Cliver SP, H.,Rman HJ, ~PPn RL. The rdad mshrp besween marunal akin(old rhicknrsc snakissg and hinh weight in hl..rk and -hire wumen. Paedixr Per,sar Epidev.iol 1994;8:516 .21. 43 CG»re LC. Aaen C. Hobcl Cl, Plan LD. Maeersal mhscco use nsd wl.uance aMur: Rii prevalence nres md assncnnnnc wirh rhe deLv«ry of.mall foe gssrarinnal agc nc•nase. 06ares Gynecd 199}81:396-4CI. 44. Cnac ingim S. Axels.s,n 0. Ekland 0. L'usdmvk G. Smoking, natemal age. ard (aal grnwth. Obssa Gyrsccol 1985;66:449-452. 45. Hun;. BL. Vicrors CC:, Menaes AM, lialpem R, Barrrs H1. Low binh weighi. Premms binhs and i e gro»+h ..rdano rdanon .o mal smoking. Pxdiaer Pe~naml Ey,deminl 1997:11:14J' 151. 46. Fingrrhut IA, Klanman JC, KesslndcJS. Smaking before, dur:ng, and a(rer pregnancy Am J Public Health 1990;80:541-544. 47. Alex.rndes GR, thms JH, KauFmzn RB. Mra J, Kopn bl. A l1mnJ Snres usl nhrcsce for fml gro.rh. Oksra Oynrcd 1996;97:Ifd-168. 48, Arbu:kl< TE, Wdkirc R. Sherman Gl, Birth we,g}t Irrrennlc> by gesra- ri„na aG in C nada. OMrur G3necol 199i;81:39-48.