Background: The intestinal microflora is a likely source for the
induction of immune deviation in infancy.
Objective: The purpose of this study was to prospectively relate
the intestinal microflora to allergy development in 2 countries
differing with respect to the prevalence of atopic diseases.
Methods: Newborn infants were followed prospectively through
the first 2 years of life in Estonia (n = 24) and Sweden (n = 20).
By that age, 9 Estonian and 9 Swedish infants had developed
atopic dermatitis and/or positive skin prick test results. Stool
samples were obtained at 5 to 6 days and at 1, 3, 6, and 12
months, and 13 groups of aerobic and anaerobic microorgan-
isms were cultivated through use of standard methods.
Results: In comparison with healthy infants, babies who devel-
oped allergy were less often colonized with enterococci during
the first month of life (72% vs 96%; P < .05) and with bifi-
dobacteria during the first year of life (17% to 39% vs 42% to
69%; P < .05). Furthermore, allergic infants had higher counts
of clostridia at 3 months (median value, 10.3 vs 7.2 log10; P <
.05). The prevalence of colonization with Staphylococcus
aureus was also higher at 6 months (61% vs 23%; P < .05),
whereas the counts of Bacteroides were lower at 12 months (9.9
vs 10.6 log10; P < .05).
Conclusion: Differences in the composition of the gut flora
between infants who will and infants who will not develop
allergy are demonstrable before the development of any clini-
cal manifestations of atopy. Because the observations were
made in 2 countries with different standards of living, we
believe that our findings could indicate a role for the intestinal
microflora in the development of and protection from allergy.
(J Allergy Clin Immunol 2001;108:516-20.)
Key words: Allergy, atopic dermatitis, infants, microflora, bifi-
dobacteria, enterococci, clostridia, Staphylococcus aureus, Bac-
teroides, prospective study
Numerous studies over the past 8 years have demon-
strated that the prevalence of atopic diseases is lower in
the formerly socialist countries of Central and Eastern
Europe than in Western European countries.1-3The rea-
sons for the difference are unknown. They are clearly
environmental, however, and cannot be explained by
genetic differences between populations.
It has been suggested that environmental factors asso-
ciated with a Western lifestyle might decrease the overall
exposure to microbial stimulation in infancy and early
childhood and, as a consequence, lessen the stimulation
of TH1-type immunity.4If this is indeed the case, then the
intestinal microflora would be of particular importance,
inasmuch as it comprises approximately 1014microor-
ganisms—10 times more than the number of cells that
the body consists of.5
We have previously shown infants in Estonia and
infants in Sweden to be different with respect to
microflora, with a low prevalence of allergies in the for-
mer and a high prevalence of allergies in the latter.6,7For
example, the counts of aerobic bacteria (coagulase-nega-
tive staphylococci [CONS], enterococci, enterobacteria,
and lactobacilli) were 10- to 1000-fold higher in Eston-
ian than in Swedish newborn babies during the first week
of life.7Furthermore, lactobacilli were more commonly
found in Estonian children at 1 month7and 1 year.6
There are also differences in the composition of the
intestinal microflora between allergic and nonallergic 2-
year-old children.8Thus the prevalence of bifidobacteria
was low in the allergic infants, whereas the counts of
Staphylococcus aureus and enterobacteria were higher.
Very recently, differences in the composition of the
intestinal microflora between allergic infants and nonal-
lergic infants were also confirmed by other methods.9
Thus, isocaproic acid, which might be an indicator of
colonization with Clostridium difficile, was found in 6 of
25 allergic 1-year-old infants but only 1 of 47 healthy
babies. These studies were all cross-sectional, however,
and did not address the issue of whether the differences
were primary or secondary to disease. Yet in a recent
Allergy development and the intestinal
microflora during the first year of life
Bengt Björkstén, MD, PhD,a,bEpp Sepp, MD, PhD,cKaja Julge, MD, PhD,d
Tiia Voor, MD,dand Marika Mikelsaar, MD, PhDcStockholm and Linköping, Sweden,
and Tartu, Estonia
From athe Centre for Allergy Research, Karolinska Institutet, Stockholm; bthe
Department of Paediatrics, Linköping University; and the Departments of
cMicrobiology and dPaediatrics, Tartu University.
Supported by grants from the Swedish Foundation for Health Care Sciences
and Allergy Research, the Swedish Medical Research Council (#7510), the
Swedish National Heart and Lung Association, the National Association
against Asthma, and Allergy and Glaxo-Wellcome Ltd (Stevenage, United
Kingdom) and by the Estonian Ministry of Education (#0418).
Received for publication March 30, 2001; revised June 11, 2001; accepted for
publication June 21, 2001.
Reprint requests: Bengt Björkstén, MD, PhD, Centre for Allergy Research,
Karolinska Institutet, 171 77 Stockholm, Sweden.
Copyright © 2001 by Mosby, Inc.
0091-6749/2001 $35.00 + 0
CCFA: Cefoxitin-cycloserine-fructose agar
CNA: Columbia agar with colistin and nalidixic acid
COA: Columbia agar with colistin sulfate and oxolinic
CONS: Coagulase-negative staphylococci
FAA: Fastidious anaerobes agar
LAM: Leeds acinetobacter medium
J ALLERGY CLIN IMMUNOL
VOLUME 108, NUMBER 4
Björkstén et al 517
prospective study, more clostridia were detected in the
feces of 3-week-old infants who developed allergy dur-
ing their first year of life.10They also tended to have less
bifidobacteria at 3 weeks.
Because differences in microbial stimulation affecting
the development of the immune system would occur very
early in life, we have tested the hypothesis that early
childhood allergy could be related to differences in the
intestinal microflora. To exclude the influence of
unknown local factors, this study was done prospective-
ly in 2 countries, one with a low and the other with a high
prevalence of atopic diseases.1,2
MATERIAL AND METHODS
Study groups and design
The study group consisted of 18 allergic infants (9 Estonian and
9 Swedish) and 26 nonallergic infants (15 Estonian and 11
Swedish). The 24 Estonian babies (12 of whom were boys) were
born at the Women’s Clinic of Tartu University Clinics between
February 1997 and June 1998. The 20 Swedish babies (12 of whom
were boys) were born at the Linköping University hospital between
March 1996 and August 1998. The 2 groups were selected from par-
ticipants in a prospective study of the development of immune
responses to allergens and the development of allergy in relation to
environmental factors. In each case, a family history of allergy was
obtained, though this was not a criterion for inclusion. A positive
family history was defined as a clear history of allergic rhinitis,
asthma, or flexural, itching dermatitis in the parents. Selection cri-
teria were vaginal delivery at term, an uncomplicated perinatal peri-
od, and availability of stool samples collected 5 to 6 days after birth
and at 1, 3, 6, and 12 months. All of the children were immunized
according to the routine procedures used in their respective coun-
tries. In Sweden, diphtheria, tetanus, inactivated polio, and
Haemophilus influenzae type B vaccination was done at approxi-
mately 3, 5, and 12 months, and rubella, mumps, and measles vac-
cination was done at 18 months. In Estonia, BCG vaccination was
done on day 3, 4, or 5 after birth, diphtheria, tetanus, pertussis, and
oral poliovirus (Sabin) vaccination at 3, 4.5, and 6 months, and
rubella, mumps, and measles vaccination at 12 months.
Clinical follow-up, focusing on allergic symptoms and skin prick
testing, was performed at 3, 6, and 12 months and 2 years. The clin-
ical assessment was done by one of us (T.V.) in Estonia and by a
trained research nurse in Sweden.
Fecal samples were collected at 5 or 6 days and at 1, 3, 6, and 12
months of age. Approximately 1 to 2 g of voided stool was collected
into sterile plastic containers by the ward staff or by the parents.
Samples collected at home were kept in a refrigerator at 4°C for no
more than 2 hours before transportation to the laboratory, where they
were frozen at –70°C until analysis. Samples from Swedish children
were transported to Estonia in dry ice for bacterial analyses.
Atopic dermatitis was defined as flexural, itching dermatitis, as
characterized by Hanifin and Rajka11; recurrent wheezing was
defined as reversible bronchial obstruction occurring at least 3 times
and verified at least once by a doctor. Allergic rhinitis/rhinocon-
junctivitis was defined as the appearance of symptoms at least twice
after exposure to allergen and unrelated to the infection. The aller-
gic children had atopic dermatitis and/or at least 1 positive skin
prick test result. Thus symptoms from the respiratory tract were not
regarded as allergy in the absence of a positive skin prick test result
during the first 2 years of life.
Skin prick testing
Skin prick tests were performed in duplicate on the volar aspects
of the forearms with natural egg white and fresh skimmed cow’s
milk (lipid concentration, 0.5%) at 3 and 6 months. At 12 months
and 2 years of age, the children were also tested with standardized
extracts of Dermatophagoides pteronyssinus, cat and dog dander,
birch, timothy (Soluprick, ALK, Hørsholm, Denmark), and cock-
roach (Bayer, Spokane, Wash). Histamine hydrochloride (10
mg/mL) was used as a positive control and glycerol as a negative
control. The test result was regarded as positive if the mean diame-
ter of one of the wheals was ≥3 mm.
Weighed samples of feces were serially diluted (10–2to 10–9) in
prereduced phosphate buffer (pH 7.2) in an anaerobic glove box
(Sheldon Manufacturing, Inc, Cornelius, Ore) with a gas mixture
consisting of 5% CO2, 5% H2, and 90% N2; they were then culti-
vated on 11 freshly prepared media. The order of analysis of sam-
ples from the Estonian and Swedish allergic and nonallergic infants
was random, except that samples from the same infant were ana-
lyzed as closely in time as possible.
Yeast extract agar was used for total aerobes; yeast extract agar
with 6.5% sodium chloride for staphylococci; Endo agar for entero-
bacteria; Leeds acinetobacter medium (LAM) with vancomycin at
10 mg/L, cefsulodin at 15 mg/L, and cephradine at 50 mg/L for
acinetobacteria12; de Man-Rogosa-Sharpe agar (Oxoid, Hampshire,
United Kingdom) for microaerophiles such as lactobacilli and strep-
tococci; Columbia agar with colistin sulfate and oxolinic acid sup-
plement (COA; Oxoid) for β-hemolytic streptococci; fastidious
anaerobes agar (FAA; LAB M, Bury, United Kingdom) for total
anaerobes; BBL Schaedler agar (Becton, Dickinson and Company,
Franklin Lakes, NJ) with vancomycin and nalidixic acid supple-
ment (Oxoid) for gram-negative anaerobes; BBL Columbia agar
with colistin and nalidixic acid (CNA; Becton, Dickinson and Com-
pany) for gram-positive anaerobes; cefoxitin-cycloserine-fructose
agar (CCFA; Oxoid) with egg yolk and sodium taurocholate for C
difficile; and Sabouraud dextrose agar with penicillin (50,000 U/L)
and streptomycin (40,000 U/L) for yeasts and fungi. The counts of
clostridia were estimated on FAA after ethanol treatment.13
Seeding of anaerobes and incubation of FAA, Schaedler agar,
CNA, and CCFA for 5 to 6 days was performed in the anaerobic
glove box. The yeast extract agar, salt-yeast-extract agar, Endo agar,
LAM, COA, and Sabouraud medium were incubated aerobically by
37°C and inspected after 24 and 48 hours. The de Man-Rogosa-
Sharpe agar medium was incubated in a microaerophilic atmosphere
(IGO 150 CO2incubator, Jouan Inc, Winchester, Va) for 72 hours.
Colonies that differed morphologically and were growing on the
plate with the highest dilution of bacteria were Gram-stained and
subjected to microscopy. The microorganisms were identified pri-
marily at the genus level (CONS, enterococci, streptococci, acine-
tobacteria, Candida, bifidobacteria, Bacteroides, Eubacterium,
clostridia) and the species level (lactobacilli, β-hemolytic strepto-
cocci, enterobacteria, S aureus, C difficile). The detection level of
the various microorganisms was 3 log10CFU/g.7
Standard methods were used for identification of enterobacteria
and other gram-negative bacteria.14A coagulase test (Oxoid) was
used for differentiation of S aureus and CONS. Streptococci and
enterococci were identified by the absence of catalase production
and differentiated by fermentation of esculin. β-hemolytic strepto-
cocci were identified by their growth on COA medium and through
use of a latex test (Oxoid). Colony and cellular morphology and a
negative catalase production identified lactobacilli grown on selec-
tive media. The anaerobes were identified up to the family or genus
level by growth on selective media, colony and cellular morpholo-
518 Björkstén et al
J ALLERGY CLIN IMMUNOL
gy, and Gram-stain reaction. We diagnosed anaerobic gram-nega-
tive rods as Bacteroides, gram-positive rods as eubacteria, bifi-
dobacteria, or clostridia, and gram-positive cocci as anaerobic
cocci. C difficile was identified by its ability to grow on CCFA, by
colony and cellular morphology, by positive Gram staining, and by
the typical smell. All anaerobic microorganisms were tested for the
absence of growth under aerobic and microaerophilic conditions.
The total count (log10CFU/g) of microorganisms and the counts
of various genera and species were calculated for each stool sample.
In addition, the relative amount of each of the particular microbes
was expressed as a percent of the total count in that sample.6
The Fisher exact test was used to compare allergic and nonaller-
gic infants with respect to the prevalence of colonization at differ-
ent ages. The counts (Mann-Whitney rank sum test) and the pro-
portions of different microorganisms (Student t test) were compared
through use of the computer program Statgraphics (Statistical
Graphics Corp, Rockville, Md).
Informed consent was obtained from the parents of the babies.
The Institutional Review Boards at Tartu and Linköping Universi-
ties approved the study.
The prevalence of colonization with enterococci was
lower at 1 week and 1 month in infants who developed
allergy during the first 2 years of life than in those who
did not (67% vs 96% [P = .02] and 72% vs 96% [P =
.03]; Table I). Similarly, the prevalence of colonization
with bifidobacteria was lower in allergic infants through
the first year of life, though significantly so only at 1
week (17% vs 50% [P = .03]) and at 3 (28% vs 62% [P
= .04]) and 12 months (22% vs 69% [P = .005]). Bifi-
dobacteria were isolated from 21 of 26 healthy and 8 of
18 allergic infants (P = .02) at 5 days and/or 1 month and
in 24 of 26 healthy and 9 of 18 allergic infants at least
once during the first 3 months of life (P = .003).
The differences between those infants who developed
allergy and those who did not were less pronounced after
the first month of life. As shown in Table II, however,
higher counts of clostridia were recorded in the allergic
infants at 3 months (10.3 vs 7.2 log10[P = .01]). Fur-
thermore, the prevalence of S aureus was higher at 6
months (Table I), and the counts of Bacteroides were
lower at 12 months (Table II).
The prevalence and counts of lactobacilli increased in
healthy children during the first month (Tables I and II).
The high proportion of aerobic CONS decreased from 1
to 3 months (mean, 20% of the flora at 1 month vs 2% of
the flora at 3 months [P = .02]). At the same time, the
proportion of anaerobic bifidobacteria increased from
15% (mean of total bacterial count in the sample) at 1
month to 40% at 3 months (P = .01).
Two children with recurrent wheezing and negative
skin prick test results in the group of nonallergic children
had a colonization pattern similar to that seen in the other
nonallergic children. The exclusion of these infants did
not appreciably alter any of the relationships between
microflora and allergy.
The duration of breast-feeding was 6 months (median)
in both the allergic and the nonallergic infants. The num-
bers of babies treated with antibiotics during the first year
of life were 9 of 18 in the allergic group and 12 of 26 in
the nonallergic group. We also analyzed the different sub-
groups in Estonia and Sweden separately. Although the
numbers then became too small for statistical analysis, the
differences between allergic and nonallergic infants were
similar in the 2 countries (data not shown).
This prospective study extends previous reports of dif-
ferences in the composition of the intestinal flora
between allergic infants and nonallergic infants8,9and
demonstrates that they are already present during the first
week of life. Differences in diet and antibiotic treatment
cannot explain the findings; the allergic and nonallergic
infants were similar with respect to diet and the use of
antibiotic drugs in both countries.
In at least 3 recent studies,differences in the composition
of the gut flora were observed between allergic and nonal-
lergic infants.8-10Thus the counts of S aureus were higher
and the prevalence of Bacteroides and bifidobacteria lower
TABLE I. Prevalence of certain species of fecal microorganisms in allergic (n = 18) and healthy (n = 26) children during
the first year of life
1 week1 mo 3 mo6 mo12 mo
Microorganisms HealthyAllergic Healthy AllergicHealthy AllergicHealthyAllergic HealthyAllergic
Prevalence of microorganisms in allergic infants vs healthy infants at a given age: *P = .02; †P = .03; ‡P = .04; §P = .05.
Prevalence of microorganisms in healthy children at different ages: ?P = .02; ¶P = .05.
J ALLERGY CLIN IMMUNOL
VOLUME 108, NUMBER 4
Björkstén et al 519
in allergic children at 2 years.8In the present prospective
study, lower counts of Bacteroides were observed at 12
months and the prevalence of bifidobacteria was lower
through the first year of life. Furthermore, the allergic
infants had higher counts of clostridia at 3 months but not
later. Very recently, higher counts were reported in allergic
infants at 3 weeks but not at 3 months or later.10Support for
a higher prevalence of colonization with C difficile in aller-
gic infants at 12 months was obtained in a considerably larg-
er study, though an indirect method was used.9
The differences between the findings of the present
study and those of previous studies might be due to dif-
ferences in statistical power and the use of different
methods. It should also be noted that the composition of
the intestinal microflora is largely unknown; the
observed differences were therefore possibly due to alter-
ations in other, unknown components of the flora. Colo-
nization with bifidobacteria and low counts of Bac-
teroides and C difficile appear to be associated with
protection against allergy, however.
An early and more extensive colonization with aerobic
bacteria in healthy infants could conceivably induce a
strong stimulation of the immune system, including a
stimulation of IL-12 by gram-positive bacteria.15The
higher prevalence of bifidobacteria in healthy children
very early in life is also of interest, because these
microorganisms are known to elicit a TH1-type immune
response15,16According to the “hygiene hypothesis,”
atopy is associated with less microbial stimulation and as
a consequence, with a prolonged propensity towards
TH2-skewed immune responses.4Interest has recently
been focused on the protective role of early infections on
the development of allergy, possibly through TH2-
inhibitory effects of cytokines released during host
defense responses to infections.17It would seem unlike-
ly, however, that a stimulus that is potentially harmful to
the host should be necessary for the postnatal maturation
of a balanced immune system. It has therefore been sug-
gested that the primary signal for such maturation during
infancy and early childhood is provided not by pathogens
but by stimulation from the commensal microbial flora—
in particular, the flora of the gastrointestinal tract.18The
present study could support this hypothesis. Further-
more, a very recent prospective study suggested that lac-
tobacilli given during the first 6 months of life might pro-
tect against the development of allergy.19
It is tempting to speculate that the differences in gut
flora between babies who will and babies who will not
develop clinical manifestations of atopy indicate that these
flora are primary causes of allergy—either promoting
immune deviation in healthy infants or delaying it in aller-
gic babies. Our findings do not exclude the possibility,
however, that atopy is associated with mucosal conditions’
facilitating a certain microflora. Thus it is possible that the
differences in the gut flora and the differences in the devel-
opment of immunity are both consequences of one or
more other, as-yet-unknown factors associated with the
atopic genotype. Only a prospective intervention study or
studies in gnotobiotic animals could settle this issue.
In conclusion, our results show that differences in the
composition of the gut flora are demonstrable very early
in life and before the development of any clinical mani-
festations of atopy. If the microbial flora drives the mat-
uration of the immune system, changes in its composi-
tion as a consequence of altered lifestyles and diet in
industrialized societies might play a role in the higher
prevalence of allergy. Because our observations were
made in 2 countries with different standards of living, we
believe that these findings could indicate a major role for
the intestinal microflora in protection from allergy.
TABLE II. Counts of certain fecal microorganisms in allergic (n = 18) and healthy (n = 26) children during the first year
Count (CFU log/g)
1 week 1 mo3 mo6 mo 12 mo
Microorganisms HealthyAllergic HealthyAllergic Healthy AllergicHealthy Allergic Healthy Allergic
8.2 7.18.8 188.8.131.52.15.35.4 4.6
The median values are given for positive samples.
Counts of microorganisms in allergic infants vs healthy infants at a given age: *P = .01; †P = .03.
Counts of microorganisms in healthy children at different ages: ‡P = .02.
520 Björkstén et al
J ALLERGY CLIN IMMUNOL
1. Björkstén B, Dumitrascu D, Foucard T, Khetsuriani N, Khaitov R, Leja
M, et al. Prevalence of childhood asthma, rhinitis and eczema in Scandi-
navia and Eastern Europe. Eur Resp J 1998;12:432-7.
2. Bråbäck L, Breborowicz A, Julge K, Knutsson A, Riikjärv M, Vasar M,
et al. Risk factors for respiratory symptoms and atopic sensitization in the
Baltic area. Arch Dis Child 1995;72:487-93.
3. von Mutius E, Martinez FD, Fritzsch C, Nicolai T, Reitmeir P, Thiemann
HH. Skin test reactivity and number of siblings. Br Med J 1994;308:692-5.
4. Holt P, Sly P, Björkstén B. Atopic versus infectious disease in childhood:
a question of balance? Pediatr Allergy Immunol 1997;8:1-5.
5. Midtvedt T. Microflora-associated characteristics (MACs) and germfree
animal characteristics (GACs) in man and animals. Microecol Ther
6. Sepp E, Julge K, Vasar M, Naaber P, Björkstén B, Mikelsaar M. Intestinal
microflora of Estonian and Swedish infants. Acta Paediatr 1997;86:956-61.
7. Sepp E, Naaber P, Voor T, Mikelsaar M, Björkstén B. Development of
intestinal micro-flora during the first month of life in Estonian and
Swedish children. Microbial Ecol Health 2000;12:22-6.
8. Björkstén B, Naaber P, Sepp E, Mikelsaar M. The intestinal microflora in
allergic Estonian and Swedish 2-year old children. Clin Exper Allergy
9. Böttcher M, Sandin A, Norin E, Midtvedt T, Björkstén B. Microflora
associated characteristics in faeces from allergic and non-allergic chil-
dren. Clin Exp Allergy 2000;30:590-6.
10. Kalliomäki M, Kirjavainen P, Eerola E, Kero P, Salminen S, Isolauri E.
Distinct patterns of neontal gut microflora in infants developing or not
developing atopy. J Allergy Clin Immunol 2001;107:129-34.
11. Hanifin J, Rajka G. Diagnostic features of atopic dermatitis. Acta Derm
Venereol (Stockholm) 1980;92(suppl):44-7.
12. Jawad A, Hawkey P, Heritage J, Snelling A. Description of Leeds acine-
tobacterium medium, a selective and differential medium for isolation of
clinically important Acinetobacter spp. J Clin Microbiol 1994;32:2353-8.
13. Marler L, Siders J, Wolters L, Pettigrew Y, Skitt B, Allen S. Comparison
of five cultural procedures for isolation of Clostridium difficile from
stools. J Clin Microbiol 1992;30:514-6.
14. Balows A, Hausler W, Herrmann K, Isenberg H, Shadomy H. Manual of
clinical microbiology. Washington (DC): ASM; 1991.
15. Hessle C, Andersson, Wold A. Gram-positive bacteria are potent induc-
ers of monocytic interleukin-12 (IL-12) while gram-negative bacteria
preferentially stimulate IL-10 production. Infect Immun 2000;68:3581-6.
16. Miettinen M, Matikainen S, Vuopio-Varkila J, Pirhonen J, Varkila K,
Kurimoto M, et al. Lactobacilli and streptococci induced interleukin-12
(IL-12), IL-18, and gamma interferon production in human peripheral
blood mononuclear cells. Infect Immun 1998;66:6058-62.
17. Björkstén B. Allergy priming early in life [commentary]. Lancet
18. Björkstén B. The environment and sensitisation to allergens in early
childhood. Pediatr Allergy Immunol 1997;8(suppl 10):32-9.
19. Kalliomäki M, Salminen S,Arvilommi H, Kero P, Koskinen P, Isolauri E.
Probiotics in primary prevention of atopic disease: a randomised place-
bo-controlled trial. Lancet 2001;357:1076-9.