Eur Respir J, 1996, 9, 26232629DOI: 10.1183/09031936.96.09122623Printed in UK - all rights reserved

Copyright ERS Journals Ltd 1996European Respiratory Journal

ISSN 0903 - 1936

BBrreeaassttffeeeeddiinngg,, mmaatteerrnnaall ssmmookkiinngg aanndd

P. Nafstad*, J.J.K. Jaakkola*+, J.A. Hagen*, G. Botten*, J. Kongerud**

Breastfeeding, maternal smoking and lower respiratory tract infections. P. Nafstad,J.J.K. Jaakkola, J.A. Hagen, G. Botten, J. Kongerud. ERS Journals Ltd 1996.ABSTRACT: The objective of the study was to assess the relationship betweenbreastfeeding and lower respiratory tract infections (LRTIs) during the first yearof life, with special reference to maternal smoking.

A cohort of 3,754 children born in 19921993 in the City of Oslo, Norway wasrecruited and data were collected at birth, 6 and 12 months of age. Complete infor-mation was obtained from 3,238 children (follow-up rate 86%). The main outcomewas an episode of a LRTI, such as pneumonia, bronchitis or bronchiolitis, basedon a self-administered questionnaire addressed to parents when the child was 6and 12 months old. The outcome was specified as physician-diagnosed.

In logistic regression analysis adjusting for confounding, maternal smokingincreased the risk of LRTIs in children breastfed for 06 months (odds ratio (AOR)1.7; 95% confidence interval (95% CI) 1.22.4), but not essentially when the childwas breastfed for more than 6 months (AOR 1.1; 95% CI 0.71.6). Short-termbreastfeeding (06 months) and no maternal smoking was related to an adjustedAOR of LRTIs of 1.3 (95% CI 1.01.7), and short-term breastfeeding combinedwith maternal smoking was related to an adjusted AOR of 2.2 (95% CI 1.63.1),as compared with long-term breastfeeding and no maternal smoking.

The present study indicates a protective effect of long-term breastfeeding on therisk of lower respiratory tract infection during the first year of life. The resultssuggest that the protective effect is strongest in children exposed to environmen-tal tobacco smoke.Eur Respir J., 1996, 9, 26232629.

*Section of Epidemiology, Dept of Popula-tion Health Sciences, National Institute ofPublic Health, Oslo, Norway. +Environmen-tal Epidemiology Unit, Dept of PublicHealth, University of Helsinki, Helsinki,Finland. **Dept of Thoracic Medicine,Rikshospitalet, Oslo, Norway.

Correspondence: P. NafstadSection of EpidemiologyDept of Population Health SciencesNational Institute of Public HealthPostboks 4404 Torshov0403 OsloNorway

Keywords: Breastfeedingchildrenlower respiratory tract infectionspassive smoking

Received: January 16 1996Accepted after revision August 17 1996

Economic support was given by the Nor-wegian Research Council.

Lower respiratory tract illnesses, including infections,in small children have an important public health impactboth in industrialized and developing countries [14].These illnesses in early childhood may predispose forchronic obstructive pulmonary disease (COPD) later inlife [57]. Several environmental exposures, such asenvironmental tobacco smoke, nitrogen dioxide, familycharacteristics and day-care attendance have been stud-ied as possible determinants of the incidence of suchillnesses [3, 4, 812]. The effect of environmental tobac-co smoke, and in particular maternal smoking, is so farbest established [4, 8, 10, 1315]. The protective effectof breastfeeding has been difficult to establish. Resultsfrom developing countries are consistent with a pro-tective effect of breastfeeding [1618], while studiesundertaken in industrialized societies have shown con-flicting results [1931]. There is recent evidence froma cross-sectional study of Australian children that breast-feeding protects from the adverse effects of passivesmoking on acute respiratory illness [29].

A birth cohort was established in Oslo in 19921993to study the environmental determinants of respiratoryhealth in children. A high proportion of Norwegian wo-men breastfeed their children, and the median length ofbreastfeeding is over 7 months [32]. The objective of thepresent study was to assess the relationship between the

length of breastfeeding and risk of lower respiratory tractinfections (LRTIs) during the first year of life, with spe-cial reference to maternal smoking.

Methods

Study population

The source population included children born duringa period of 15 months in the city of Oslo, Norway, witha population of approximately 500,000. The recruitmentfor the study was carried out by midwives in the mater-nity wards at Aker and Ullevl Hospitals, the two mainbirth clinics in Oslo. About 90% of the mothers withpermanent address within the city of Oslo attend mater-nity ward at these two birth clinics. Only in periodswhen capacity is insufficient do normal deliveries takeplace at another hospital. This birth clinic was not includ-ed for organizational reasons. The recruitment periodwas from January 1 to December 31, 1992 at UllevlHospital (2,694 children born live), and from March 10,1992 to March 9, 1993 at Aker Hospital (3,706 chil-dren born live). A delay in authority approval necessi-tated the later start at Aker Hospital.

The inclusion criteria were: 1) permanent address with-in the city of Oslo; 2) no plans to move from Oslo with-in the near future; 3) birth weight more than 2,000 g;4) no serious illness which might impair respiration (sev-ere respiratory, cardiovascular, neuromuscular or metabo-lic disease); 5) no assisted ventilation or oxygen therapyafter 6 h of life; 6) at least one family member able tospeak and write Norwegian (assessed by the midwives);7) at least one biological parent living together with thechild; and 8) no known drug abuse in the family. Familiesfulfilling these criteria were informed about the mainobjectives of the study.

The families were asked to sign a parental consentform and to complete a self-administered questionnaireinquiring about pregnancy, smoking habits, details ofhome environment, and family and socioeconomic char-acteristics. A reminder was sent to those eligible fami-lies who did not return the questionnaire. Most of theimmigrant families (identified by hospital records) werefound to have insufficient knowledge of the Norwegianlanguage. Immigrant families not responding to the firstapproach were, therefore, considered not eligible forthe cohort (1,045 out of 1,172 immigrant families). Thenumbers of children in the source population eligiblefor the cohort, entered in the cohort, and of childrenwith complete information on respiratory health bothfrom 6 month and 12 month follow-up surveys arepresented in figure 1. Subjects with no information onrespiratory health in the first or second half year oflife were considered lost to follow-up. A maximum of4,973 children were found to be eligible for the cohort.

Approximately 150250 of these children (exact num-ber not available) were not included due to plans tomove out of Oslo after birth. A total of 3,754 children(76%) were enrolled in the study.

The study was approved by the Norwegian Data In-spectorate and the Regional Ethics Committee for Medi-cal Research.

Data collection

When the child was 6 and 12 months old, the parti-cipating families received a questionnaire with questionsabout the child's health, use of health services and detailsof their environment. A reminder was sent within 3weeks. Based upon hospital records, information on fa-mily characteristics, pregnancy and birth was registeredfor 3,743 (99.7%) of the participants (table 1). Corres-ponding information was also obtained from 1,050 non-participating families, who were eligible for the cohort.

Outcome measures

The primary outcome was an episode of a LRTI, suchas pneumonia, bronchitis or bronchiolitis. The operatio-nal definition of the outcome was based on the follow-ing question addressed to parents when the child was 6and 12 months old: "Has the child been examined by a

P. NAFSTAD ET AL.2624

Fig. 1. Recruitment and follow-up of children in the Oslo BirthCohort 19921993.

Sourcepopulation(n=6,400) Not eligible

(n=1,427)Address outside Oslo (n=180)Birthweight

physician (during the period of interest)? If yes, fill inat what age and which diagnosis". Due to a commonpractice among some physicians to describe symptomsof bronchiolitis during infancy as respiratory syncytialvirus infection, parental report of such a diagnosis wasconsidered as an episode of LRTI. The outcome wasspecified as occurring during the first 12 months of lifeor, alternatively, either during the first 6 months or from712 months of life. Parents were also asked to reportepisodes of LRTIs needing hospitalization, and thesewere treated separately.

Breastfeeding and maternal smokingThe length of breastfeeding and maternal smoking

were the main determinants. The length of breastfeed-ing, expressed in months, was based on information onthe endpoint of breastfeeding in the 6 month question-naire, and in the 12 month questionnaire if the childwas still breastfed at 6 months of age. In the 6 monthquestionnaire, the mothers were also asked if they hadstopped breastfeeding, if they still gave their child onlybreast milk, or a combination of breastmilk and formulasubstitute. Information on smoking was based on the 6month questionnaire, which was considered the mostrepresentative for the exposure during the first year oflife. The mothers were asked to state whether they werenonsmokers, occasional smokers or daily smokers. Ifthey were daily smokers, they were asked to state theaverage number of cigarettes smoked per day. Occasionalsmokers were those who reported smoking less oftenthan daily.

Covariates

The following covariates included in the analyseswere from hospital records or questionnaire informationat birth: gender; birth weight; season of birth; maternalage; single parenthood; family income; maternal edu-cation; mother's nationality; parental asthma; distancefrom nearest road open to public traffic; and crowded-ness. Information on siblings, sharing of bedroom withsiblings and paternal smoking habits were from the 6month questionnaire, and day-care attendance (at least10 h per week) both from 6 and 12 month question-naires. The categories and frequency distributions of thecovariates are presented in table 2. Nationality wasdichotomized according to mother's country of origin(third world, or other). Crowdedness was defined asfloor area (m2) per person living in the child's home.As combustion products of motor vehicles are the mainsource of air pollution in Oslo (industrial air pollutionis low), the distance between the child's home and thenearest road open to public traffic was used as a crudemeasure of exposure to air pollution. Parents were askedto report smoking habits of the father, like those of themother.

Statistical methods

The cumulative incidence was used as a measure ofthe risk of a LRTI episode. Cumulative incidences from

BREASTFEEDING AND RESPIRATORY INFECTIONS 2625

Table 2. Characteristics of the study populationaccording to length of breastfeeding (Oslo Birth Cohort19921993)

Breastfed Breastfed06 months >6 months

(n=989) (n=2,200)% %

Male gender* 55 51Birth weight**

10 hweek-1) 1 1Day-care outside home at 12 months(>10 hweek-1) 15 15Maternal age**

birth to 12 months and from 7 to 12 months were esti-mated for different categories of breastfeeding stratifiedby maternal smoking. The odds ratio (OR) was used asa measure of association between the occurrence ofrespiratory infections and breastfeeding. The adjustedodds ratios (AORs) were estimated in logistic regres-sion analyses, controlling for the covariates in table 2.Based on previous knowledge, eight core covariates (gen-der, siblings, sharing of bedroom, day-care attendance,maternal education, parental asthma, maternal and pater-nal smoking habits) were always included in the model.The other covariates were included, if their inclusionhad an impact on the accuracy and precision of the esti-mate of the relationship studied [33].

The independent effects of breastfeeding and mater-nal smoking, as well as their joint effect, were studiedin logistic regression analyses using long-term breast-feeding (>6 months) and no maternal smoking, the largestand most preferable category, as the reference category.The following three categories were contrasted to thereference category: 1) long-term breastfeeding (>6 mon-ths), maternal smoking; 2) short-term breastfeeding (06 months), no maternal smoking; and 3) short-termbreastfeeding, maternal smoking. The children of occa-sionally smoking mothers (n=262) were excluded.

The analyses were performed using the StatisticalPackage for the Social Sciences (SPSS) for Windows,version 6.0 [34].

Results

Participation

The children and families entered in the study werenot different from those not participating as to birth wei-ght, maternal age and percentage of families with firstborn child. Smoking during early pregnancy and mater-nal history of allergy were more common in nonpartic-ipating than participating mothers (40 versus 33% and20 versus 17%). Of the 3,754 families entered in thecohort, 3,563 (95%) answered the 6 month question-naire and 3,452 (92%) the 12 month questionnaire.Nonresponders at 6 months received a 12 month ques-tionnaire in the same way as the others. Due to an errorin the data-handling procedure, information collected onrespiratory health during the first 6 months of life waslost for 124 (3%) of the children. This left a total of

3,238 (86%) children with information on respiratoryhealth both at 6 and 12 months.

The baseline characteristics of the children in the 1year cohort and those lost to follow-up are presented intable 1. Children in the 1 year cohort (n=3,238) did notdiffer from those entered in the cohort (n=3,754). Theproportion of mothers reporting a history of asthma wassimilar among the 1 year cohort and those lost to fol-low-up (6.9 and 6.0%, respectively).

BreastfeedingInformation on the length of breastfeeding was obtained

from 3,189 (99%) of the 1 year cohort. Only 30 (1%)children were not breastfed at all, and 837 (26%) werestill breastfed at the age of 12 months. Of 2,200 chil-dren still breastfed after 6 months, 1,738 (69%) werebreastfed without infant formula at 6 months of age, and453 (21%) were breastfed and received infant formula(missing=9). The characteristics of the children and chil-dren's families according to breastfeeding status at 6months are presented in table 2. Breastfed children hadolder and more educated mothers than children breast-fed for 06 months, and their parents were less likelyto smoke. Daily smoking mothers reporting breastfeed-ing at the 6 month questionnaire smoked on average 9cigarettesday-1 (SD=5, median=10), compared with 11cigarettesday-1 (SD=5, median=10) for the mother whohad stopped breastfeeding (p6 monthsMaternal smoking n CuI% 95% CI n CuI% 95% CI p-value

No 575 20 (1623) 1751 16 (1417) 0.03Occasional 85 19 (1027) 177 15 (920) 0.39114 cigarettesday-1 209 23 (1728) 206 16 (1121) 0.0915 cigarettesday-1 112 32 (2341) 54 15 (525) 0.02Total 981 22 (1924) 2188 16 (1417)

1924), compared with children breastfed for more than6 months (CuI 16%; 95% CI 1417). The difference inCuI between the breastfeeding categories increased withincreasing amount of cigarettes smoked by the mothers.

In logistic regression analysis controlling for the corecovariates, the AOR of LRTIs increased on average bya factor of 1.05 (95% CI 1.021.08) per 1 month decre-ase in duration of breastfeeding. Consequently no breast-feeding increased the AOR of LRTIs to 1.7 (95% CI1.22.5) compared with 12 months breastfeeding. Thecorresponding estimate for children of nonsmoking moth-ers was 1.03 (95% CI 1.001.07), and for children ofdaily smoking mothers 1.07 (95% CI 1.021.13) per 1month decrease in duration of breastfeeding.

In logistic regression analysis adjusting for the core co-variates, maternal smoking increased the risk of LRTIsin children breastfed for 06 months (AOR 1.7; 95%CI 1.22.4). Corresponding figures for hospitalized LRTIswere AOR 3.2, 95% CI 1.66.7. The effect of maternalsmoking showed an exposure-response pattern (114cigarettesday-1: AOR 1.5, 95% CI 0.92.4; 15 ciga-rettesday-1: AOR 2.3, 95% CI 1.43.9 for all LRTIs).However, maternal smoking did not essentially increasethe risk of LRTIs, when the child was breastfed for morethan 6 months (AOR 1.1, 95% CI: 0.71.6). Paternalsmoking was not significantly related to the outcome(figures not given).

The relationship between the occurrence of LRTIs andbreastfeeding was elaborated for all LRTIs, and sepa-rately for hospitalized infections, in different categoriesof maternal smoking (table 4). Breastfeeding had a pro-tective effect on infections in children of nonsmokingmothers. Short-term breastfeeding resulted in a greaterrisk of all infections (AOR 1.3; 95% CI 1.01.7), andhospitalized infections (AOR 1.4; 95% CI 0.82.5) thanlong-term breastfeeding. Short-term breastfeeding com-bined with maternal smoking was related to an AORof 2.2 (95% CI 1.63.1) for all infections and of 4.6(95% CI 2.58.3) for hospitalized infections, when com-pared to long-term breastfeeding and no maternal smok-ing.

To elaborate a meaningful time sequence betweenexposure and outcome, analyses were carried out focus-ing on exposure during the first 6 months and outcomefrom 7 to 12 months of age (table 5). The risk estimatesfor the protective effect of breastfeeding and adverseeffect of maternal smoking on LRTIs did not differessentially from those focusing on the whole year.

Discussion

In this cohort study of 3,754 Oslo children, short-termbreastfeeding (06 months) without maternal smokingincreased the risk of LRTIs, as compared to long-termbreastfeeding without maternal smoking (AOR 1.3; 95%CI 1.01.7). Maternal smoking increased the risk ofLRTIs in children breastfed for 06 months (AOR 1.7;95% CI 1.22.4), and this effect showed an exposure-response pattern. For children breastfed more than 6months, maternal smoking did not increase the risk sig-nificantly (AOR 1.1; 95% CI 0.71.6). Short-term breast-feeding combined with maternal smoking was relatedto an AOR of 2.2 (95% CI 1.63.1) compared to long-term breastfeeding and no maternal smoking.

The relationship between breastfeeding and the riskof LRTI is complicated in several ways [30]. Breastfeed-ing may be associated with other measured determi-nants of LRTI, and the association observed may be theresult of selection bias or confounding. Information onLRTI may be related to breastfeeding habits becauseof parental reporting, use of health care, or physician'sdiagnostic habits. A prospective cohort study avoidssome of the validity problems, such as bias in the as-certainment of exposure, but losses to follow-up couldintroduce bias. In the present study, this was not likelyas the follow-up rate was high (86%), and the baseline

BREASTFEEDING AND RESPIRATORY INFECTIONS 2627

Table 4. The risk of lower respiratory tract infections during the first year of life in relation to the length of breast-feeding (long is >6 months) and maternal smoking (Oslo Birth Cohort 19921993)

All infections Hospitalized infections

Breast- Maternalfeeding smoking n CuI% COR AOR 95% CI CuI% COR AOR 95% CI

>6 months No 1751 16 ref ref - 3 ref ref ->6 months Yes 264 16 1.0 1.1 0.71.6 3 1.1 1.1 0.52.706 months No 575 20 1.3 1.3 1.01.7 4 1.4 1.4 0.82.506 months Yes 326 26 1.9 2.2 1.63.1 9 2.6 4.6 2.58.3

: logistic regression analysis adjusting for gender, having siblings, sharing of bedroom, day-care outside home, maternal edu-cation, parental history of asthma and paternal smoking. CuI: cumulative incidences; COR: crude odds ratio; AOR: adjusted oddsratio; 95% CI: 95% confidence interval; ref: reference category. Missing information: on length of breastfeeding=49; on mater-nal smoking=11. Children of occasional smoking mothers (n=262) excluded.

Table 5. The risk of lower respiratory tract infectionsfrom 712 months of life in relation to the length ofbreastfeeding (long is >6 months) and maternal smok-ing (Oslo Birth Cohort 19921993)

All infections Hospitalizedinfections

Breast- Maternalfeeding smoking AOR 95% CI AOR 95% CI

>6 months No ref - ref ->6 months Yes 1.0 0.61.5 1.0 0.33.606 months No 1.4 1.01.8 1.5 0.73.306 months Yes 1.9 1.32.7 5.0 2.211.5

: logistic regression analysis adjusting for gender, having sib-lings, sharing of bedroom, day-care outside home, maternaleducation, parental history of asthma and paternal smoking.For definitions see legend to table 4. Missing information: onlength of breastfeeding=49; on maternal smoking=11. Childrenof occasional smoking mothers (n=262) excluded.

characteristics of those followed for 1 year were simi-lar to the total cohort.

The assessment of the outcome depends on parentalbehaviour in relation to the health services, availabilityof health care, physicians' diagnostic habits, and paren-tal reporting in the questionnaire. For a severe disorder,parental behaviour is likely to have less influence onascertainment. To minimize the effect of parental healthbehaviour, we included only those episodes of LRTIdiagnosed by a physician. Also, a separate analysis wascarried out for hospitalized cases. The cumulative inci-dence of LRTIs in this study was lower than in someprevious studies using the term "lower respiratory ill-ness" [2, 4], and indicates a more severe outcome. Itcould be that the present approach resulted in a highproportion of true LRTIs and, thus, justifies the use ofthe term "lower respiratory tract infection" rather than"lower respiratory illness". In Norway, expenses formedical treatment of children are paid by the govern-ment, which ensures practically equal availability ofhealth care and, consequently, also a uniform parentalpractice in contacting doctors when their children areill. The frequency of physician visits for respiratorysymptoms was not related to the length of breastfeed-ing. This indicates that the use of health services wasnot a likely source of bias.

There may be a considerable variation in the diag-nostic practice of LRTIs, but there is no reason tobelieve that breastfeeding of a child per se, would affecta doctor's diagnosis of respiratory disease. Except forthe first months of life, it is unlikely that doctors areconcerned about the length of breastfeeding when theydecide on diagnosis or hospitalization.

The parental reporting of LRTI could be related tobreastfeeding. It has been argued that active and fre-quent surveillance is necessary to prevent bias in report-ing (information bias). On the other hand, this wouldfocus on the objectives of the study and, thus, introducethe possibility of reporting bias both by parents and datacollectors. Blinding of the objectives, as in the presentstudy, might be a better strategy for reducing informa-tion bias. The validity of the outcome was assessed bya physician-administered telephone interview of 100randomly selected participants, who reported no LRTIsshortly after (14 days) receiving the questionnaire. Theinterview revealed no LRTIs and none of the interviewedfamilies had visited a doctor for lower respiratory tractsymptoms during the last 6 months before the interview.

Length of breastfeeding was associated with severalknown or suggested determinants of LRTIs, such as ha-ving siblings, maternal age, socioeconomic status (mea-sured here by maternal education and family income),and parental smoking habits. The potential confoundingby these factors was taken into account in logistic reg-ression analysis. However, the crude and adjusted oddsratios were similar. The variance in socio-economic sta-tus among Norwegian families is small compared tomany other countries and probably reduces the risk ofconfounding by socioeconomy. Due to a long full-paidmaternity leave, very few children in Norway attendday-care before the age of 10 months. Logistic regres-sion models with or without day-care practice at 6 or12 months gave the same estimates of the protectiveeffect of breastfeeding.

Smoking habits within the family were used as acrude measure of the child's exposure to environmentaltobacco smoke. Questionnaire information on maternalsmoking at the end of pregnancy was compared withbiomarkers in umbilical cord serum taken at birth for202 randomly selected mothers [35]. An excellent agree-ment was shown, which indicated that the mothers repor-ted their smoking habits correctly. In agreement withprevious studies [8], the cumulative incidence of LRTIwas significantly related to maternal smoking, but notto paternal smoking. A closer relationship between themother and the child than between the child and anyother person could explain why maternal smoking is ofparticular importance during the first year of life. Ourown unpublished results indicate that occasional smok-ers are, on average, light smokers [35]. Their smokingrate varies over a large range. Consequently, their chil-dren's exposure to tobacco smoke products would vary.Therefore, these children were excluded from some ofthe analyses.

Earlier studies of the protective effects of breast-feeding on children's health have been inconsistent, anda recent review concluded that the evidence of theprotective effect of breastfeeding against infections indeveloped countries is weak [36]. Some studies havepresented a significant relationship between the lengthof breastfeeding and outcomes, such as respiratory syn-cytial virus infection, pneumonia, wheezy bronchitis orLRTI [2226, 28]. On the other hand, in a prospectivecohort study, RUBIN et al. [20] found no protective effectof breastfeeding on LRTIs during the first year of life.

In contrast to some of the earlier studies, the presentstudy design made it possible to take into account thepossible time sequence of breastfeeding and occurrenceof infections, because relevant information was collect-ed at birth and at 6 and 12 months. Most informationwas taken from the birth or 6 month questionnaire, asit was considered to be the most representative andaccurate for children's exposure during their first yearof life. Focusing on infections between 712 monthsrevealed a plausible time sequence supporting a trueprotective effect of breastfeeding. Our observation of aweaker effect of exposure to tobacco smoke productson LRTIs in long-term breastfed children, is consistentwith results of a cross-sectional study of acute respira-tory illness among children in Adelaide, Australia [29].The results do not support a hypothesis that exposureto tobacco smoke products through breastmilk increa-ses the risk of LRTIs. A modification of the effect ofenvironmental exposure to tobacco smoke products onsymptoms or infections of the lower airways, could bea possible effect of breastfeeding.

In conclusion, the results of the present study indi-cate a protective effect of breastfeeding on the risk oflower respiratory tract infections during the first yearof life. The results suggest that the preventive effect isstronger in children of smoking mothers compared withchildren of nonsmoking mothers.

Acknowledgements: The authors would like to acknowledgethe staff of the maternity wards at Aker and Ullevl Hospitalfor their support in data collection, and L. Bakketeig, K.C.Ldrup Carlsen, K.H. Carlsen and L. ie for their contributionin the planning of this cohort study.

P. NAFSTAD ET AL.2628

References

1. Denny FW, Clyde WA. Acute lower respiratory tractinfections in nonhospitalized children. J Pediatr 1986;108: 635645.

2. Wright AL, Taussig LM, Ray G, Harrison HR, HolbergCJ, and GHMA. The Tucson children's respiratory study.II. Lower respiratory tract illness in the first year of life.Am J Epidemiol 1989; 129: 12321246.

3. Monto AS, Ross H. Acute respiratory illness in thecommunity: effect of family composition, smoking andchronic symptoms. Br J Prev Soc Med 1977; 31: 101108.

4. Fergusson DM, Horwood LJ, Shannon FT, Taylor B.Parental smoking and lower respiratory illness in thefirst three years of life. J Epidemiol Commun Health1981; 35: 180184.

5. Samet JM, Tager I, Speizer F. The relation betweenrespiratory illness in childhood and chronic airflowobstruction in adulthood. Am Rev Respir Dis 1983; 127:508523.

6. Burrows B, Knudson RJ, Lebowitz MD. The relationof childhood respiratory illness to adult obstructive air-way disease. Am Rev Respir Dis 1977; 115: 751759.

7. Shaheen SO, Barker JP, Shiell AW, Crocker FJ, WieldGA, Holgate ST. The relationship between pneumoniain early childhood and impaired lung function in lateadult life. Am J Respir Crit Care Med 1994; 149:616619.

8. United States Environmental Protection Agency. Res-piratory health effects of passive smoking: lung cancerand other disorders. United States Environmental Pro-tection Agency, EPA/600\6-90\006F, 1992. WashingtonDC.

9. Samet J, Lambert WE, Skipper BJ, et al. Nitrogen dio-xide and respiratory illnesses in infants. Am Rev RespirDis 1993; 148: 12581265.

10. Morgan WJ, Martinez FD. Risk factors for developingwheezing and asthma in childhood. Pediatr Clin NorthAm 1992; 39: 11851203.

11. Leeder SR, Corkhill R, Irwig LM, Holland WW. Influenceof family factors on the incidence of lower respiratoryillness during the first year of life. Br J Prev Soc Med1976; 30: 203212.

12. Louhiala PJ, Jaakkola N, Ruotsalainen R, Jaakkola JJK.Form of day-care and respiratory infections among Fin-nish children. Am J Public Health 1995; 85: 11091112.

13. Martinez FD, Cline M, Burrows B. Increased incidencesof asthma in children of smoking mothers. J Pediatr1992; 89: 2126.

14. Wright AL, Holberg CJ, Martinez FD, Taussig LM, andGHMA. Relation of parental smoking to wheezing andnonwheezing lower respiratory tract illnesses in infancy.J Pediatr 1991; 118: 207214.

15. Stoddard JJ, Miller T. Impact of paternal smoking onthe prevalence of wheezing respiratory illness in chil-dren. Am J Epidemiol 1995; 1412: 96102.

16. Chen Y, Shunzhang Y, Li W. Artificial feeding andhospitalization in the first 18 months of life. Pediatrics1988; 81: 5862.

17. Victora CG, Fuchs SC, Flores JAC, Fonseca W, KirkwoodB. Risk factors for pneumonia among children in a Bra-zilian metropolitan area. Pediatrics 1994; 93: 977985.

18. Foreman MR, Graubard BI, Hoffman HJ, Beren R, HarleyEE, Bennett P. The Pima infant feeding study: breast-feeding and respiratory infections during the first yearof life. Int J Epidemiol 1984; 13: 447454.

19. Howie PW, Forsyth JS, Ogston SA, Clark A, FloreyCV. Protective effect of breastfeeding against infection.BMJ 1990; 300: 1116.

20. Rubin DH, Leventhal JM, Kransilnikoff PA, et al.Relationship between infant feeding and infectious ill-ness: a prospective study of infants during the first yearof life. Pediatrics 1990; 85: 464471.

21. Leventhal JM, Shapiro ED, Aten CB, Berg AT, EgerterSA. Does breastfeeding protect against infections ininfants less than 3 months of age? Pediatrics 1986; 78:896903.

22. Holberg CJ, Wright AL, Martinez FD, Ray G, TaussigLM, Lebowitz MD and GHMA. Risk factors for res-piratory syncytial virus-associated lower respiratory ill-nesses in the first year of life. Am J Epidemiol 1991;133: 11351151.

23. Wright AL, Holberg CJ, Martinez FD, Morgan WJ,Taussig LM, GHMA. Breastfeeding and lower respira-tory tract illness in the first year of life. Br Med J 1989;299: 946949.

24. Ford K, Labbok M. Breastfeeding and child health inthe United States. J Biosoc Sci 1993; 25: 187194.

25. Pullan CR, Toms GL, Martin AJ, Gardner PS, WebbJKG, Appleton DR. Breastfeeding and respiratory syn-cytial virus infection. Br Med J 1980; 281: 10341036.

26. Rylander E, Pershagen G, Eriksson M, Nordwall L. Par-ental smoking and other risk factors for wheezing bron-chitis in children. Eur J Epidemiol 1993; 9: 517526.

27. Beaudry M, Dufour R, Marcoux S. Relation betweeninfant feeding and infections during the first six monthsof life. J Pediatr 1995; 126: 191197.

28. Pisacane A, Graziano L, Zona G, et al. Breastfeedingand acute lower respiratory infection. Acta Paediatr1994; 83: 714718.

29. Woodward A, Douglas RM, Graham NMH, Miles H.Acute respiratory illness in Adelaide children: breast-feeding modifies the effect of passive smoking. JEpidemiol Commun Health 1990; 44: 224230.

30. Baucher H, Leventhal JM, Shapiro ED. Studies onbreastfeeding and infections: How good is the evidence?JAMA 1986; 256: 887892.

31. Cunningham AS, Jelliffe DB, Jelliffe EFP. Breastfeed-ing and health in the 1980s: a global epidemiologicreview. J Pediatr 1991; 118: 659666.

32. Oshaug A, Botten G. Human milk in food supply sta-tistics. Food Policy 1994; 19(5): 479482.

33. Kleinbaum DG. Logistic regression: a self-learning text.Ann Arbor, Springer-Verlag, 1994.

34. SPSS for Windows. Release 6.0. Chicago, USA, SPSSInc.

35. Nafstad P, Kongerud J, Botten G, et al. Fetal exposureto tobacco smoke products: a comparison between self-reported maternal smoking and concentration of coti-nine and thiocyanate in cord serum. Acta Obstet GynecolScand (in press).

36. Statement of the Standing Committee on Nutrition ofBritish Paediatric Association. Is breastfeeding benefi-cial in the UK? Arch Dis Child 1994; 71: 376380.

BREASTFEEDING AND RESPIRATORY INFECTIONS 2629