Background
Atopic dermatitis (AD), asthma and allergic rhinoconjunctivitis (ARC) are a major cause of chronic disease in childhood. A revised version of the “hygiene hypothesis” suggests that the pattern of colonisation and the diversity of the intestinal microbiota may be an important factor in the increased prevalence of these diseases observed over the past several decades [
1‐
3]. Subsequently, probiotics have been investigated in the prevention and treatment of allergy related diseases [
3‐
8], with the strongest evidence emerging for the primary prevention of atopic dermatitis [
3‐
5]. Throughout this paper we refer AD, asthma and ARC as “allergy related diseases”, recognising that not all presentations of these conditions are related to a classic IgE-mediated inflammatory process.
Randomised controlled trials (RCTs) testing probiotics in the prevention of childhood allergy related disease are heterogeneous and have used a variety of bacterial strains, administration regimes and varying ages of follow-up. Using information from the first published follow up for each trial, a recent meta-analysis concluded that probiotic administration is protective against the development of AD in infancy [
3]. Among the studies with follow-up at or beyond 5 years of age, the greatest protective benefit of probiotics against AD appears to be in early childhood and it is less certain if this effect persists until school age [
9‐
25]. Only one of these studies did not specify a maternal or family history of atopy as an inclusion criteria [
9,
10], and there is therefore a particular need for further longer term follow-up studies to determine the ongoing effect of perinatal probiotics in general populations.
The Probiotics in the Prevention of Allergy among Children in Trondheim (ProPACT) study is a double-blinded RCT investigating the effect of maternal probiotic supplementation on childhood allergy related diseases in a general population. The initial results of the ProPACT study demonstrated a clinically significant reduction in the cumulative incidence of AD at 2 years (OR 0.51, 95 % CI 0.30 – 0.87,
p = 0.013), with the greatest reduction seen in children not considered at “high risk” for allergy related disease based on a negative family history [
26]. Probiotic supplementation did not significantly affect the incidence of asthma, ARC or atopic sensitisation at 2 year of age, although the diagnosis of the former two diseases is uncommon and controversial at such a young age.
The participating children were re-contacted and re-assessed at 6 years of age for the presence of allergy related diseases and allergic sensitisation. The aim of the current paper was to investigate the effect of maternal perinatal probiotic supplementation on the cumulative incidence of AD and ARC and prevalence of asthma and atopic sensitisation at 6 years of age.
Discussion
Maternal probiotic supplementation given to a general population of women appears to have an ongoing preventative effect on the cumulative incidence of AD until school age, however this did not reach statistical significance in the MICE analysis. There was no observed effect of probiotics on the cumulative incidence of ARC, the 12 month prevalence of asthma or current atopic sensitisation.
A significant proportion of cases of allergy related diseases occur in children who are otherwise considered not to be at “high risk” and therefore primary prevention strategies must also be assessed in general populations [
30]. This long term follow-up of the ProPACT trial is an important addition to the literature concerning probiotics in the prevention of allergic disease as one of few studies to recruit participants from a general population [
7,
8,
29]. The only other RCT to have reported long term follow-up in a general population observed no benefit of their probiotic regime on the prevalence of any allergic disease or the cumulative incidence of questionnaire defined AD at 8–9 years of age [
9,
10]. Their regime included
Lactobacillus paracasei spp
paracasei F19 supplementation given to children during weaning. In comparison, our study involved pre- and postnatal supplementation of a mixture of probiotics which included the
L. rhamnosus GG (LGG) strain. Both LGG and pre- and postnatal regimes were found to be associated with reduced RR of AD on sub-group analysis in a meta-analysis [
3]. Additionally, we report UK Working Party diagnostic criteria defined AD which is a more extensively validated method of diagnosis [
31,
32].
Looking to other trials, five RCTs have assessed the cumulative incidence of AD at 5 years of age or beyond in “high risk” populations [
11‐
23]. The trial presented by Kalliomaki et al. [
11‐
13] and the
Lactobacillus rhamnosus (HN001) arm of the study by Wickens et al. [
14‐
16] report a significant reduction in the cumulative incidence of AD, which is sustained, although reduced, at follow-up at 6 years. Together with the current study, these trials indicate that the beneficial effect of probiotics on AD is most pronounced in infancy and may continue into early childhood. Furthermore, they suggest that we are observing a true primary preventative effect, rather than an intervention which delays the onset of AD. Contrastingly, the preventative effect of probiotics seen at 2 years in the large RCT published by Kukkonen/Kuitunen et al. [
17,
18] showed no trend towards ongoing benefit on follow-up at 5 years. The
Bifidobacterium animalis subsp.
lactis (HN019) arm of Wickens et al. [
14‐
16] and 2 other trials [
19‐
23] demonstrated no significant effect of probiotics on the cumulative incidence of AD at any follow-up time point from 1 to 7 years of age. It is interesting to note that all of the studies with an observed ongoing effect administered a
L. rhamnosus strain pre- and postnatally, and that if the child was breastfed, the postnatal supplementation was given solely to the mother during the first months with the exception of Kalliomaki et al. [
11‐
13] where approximate 57 % of participants opted to give the probiotic or placebo capsules directly to the newborn children. In contrast, studies without an ongoing effect have either administered probiotic species other than
L. rhamnosus and or specified that the probiotic supplements were to be given directly to the children regardless of breastfeeding
. Further research is required to investigate if these observations are coincidental or represent strain and or regime specific effects.
The lack of effect on asthma, ARC and atopic sensitisation may reflect a true lack of effect or that the current study is under-powered to observed smaller differences in less frequent diseases. This is a universal problem for probiotic trials reporting asthma as an outcome [
8]. Two recent meta-analyses concluded that there is not enough evidence to support perinatal probiotic supplementation in the prevention of childhood asthma or wheeze [
7,
8]. Reassuringly, neither our study nor these meta-analyses found that probiotics increased the risk of asthma or wheeze, a concern which arose after long term follow-up by Kalliomaki et al. [
8,
11] Whilst one of these meta-analyses suggested that probiotics may increase lower respiratory tract infections [
8], our study does not support this conclusion. On the contrary, we observe a trend towards lower cumulative incidence of pneumonia in the probiotic group. Atopic sensitisation was found to be significantly reduced in a sub-group meta-analysis of regimes which combined pre- and postnatal administration [
7]. Consistent with the ProPACT study, none of the individual longer term follow-up studies have observed a significant effect of probiotics on sensitisation at school age. Interestingly,
Lactobacillus acidophilus was found to be associated with an increased rate of atopic sensitisation in a multivariate meta-regression analysis [
7]. This observation requires further investigation and may, in part, explain the lack of effect of the probiotic regime on sensitisation in the ProPACT trial.
The major limiting factor of this RCT is the high proportion of missing data which naturally raises concerns regarding the generalisability of the results and introduction of bias. Following the four point ITT analysis strategy recommended by White et al. [
29] we have attempted to follow-up all participants, performed a primary analysis using MICE and a sensitivity analysis using PMM. Both of the latter two models account for all randomised participants under a range of assumptions about the cause of missingness and in doing so attempt to minimise bias from covariate-related and outcome-related drop-out, respectively. The reasons for loss to follow-up are primarily unknown, however very few participants actively withdrew from the study. An estimated 73 participants had moved from the study region, which would have precluded their attendance at the clinical examination in a presumably random manner. A number of these participants were followed-up through the questionnaire. In terms of generalisability, the original ProPACT population was similar to the total PACT population, which in turn was representative of the general population in Trondheim, Norway [
26]. At the 6 year follow-up, the remaining participants were more likely to have a family history of atopy, older siblings and a pet and less likely to have a father who smoked. As these differences were not large we believe that the results are still generalisable to westernised populations where there is a reasonably high rate of allergy related disease. The PMM sensitivity analysis is particularly pertinent in this case because atopic sensitisation and or a diagnosis of AD at 2 years are associated with both attendance at the 6 year clinical examination and a diagnosis of AD at 6 years. This raises suspicions that the data is partially MNAR which would lead to biased estimates under both the complete case and multiple imputations analysis models. The PMM analysis suggests that results of this study for ARC, asthma and atopic sensitisation would have only been affected if there was an unrealistically strong association between disease and missingness in a single treatment arm. On the other hand, the magnitude of the preventative effect attributed to probiotics for AD at 6 years must be considered with caution, as the observed benefit is sensitive to the assumption that the relationship between AD and loss to follow-up is essentially identical in both the probiotic and placebo groups. Regardless of whether the outcome variables are partially MNAR, the MICE estimates are expected to be less biased than the complete case analysis. Another limitation is that the participants were informed of their treatment allocation and the observed reduction in AD after publication of the results from the 2 year follow-up in 2010 [
26], although we do not believe that this to have significantly affected the current results. Firstly, the knowledge of treatment allocation has not affected participant behaviour with equal numbers from the probiotic and placebo groups attending the clinical follow-up at 6 years. Secondly, the UKWP diagnosis was based on assessment by research nurses who were unaware of treatment allocation.
Acknowledgements
We thank all the children and their parents for their participation in this study along with the project assistants, Guri Helmersen and Else Bartnes, Dr Anne Rø, and the midwives of the Trondheim region.
Competing interests
T.Ø., O.S., C.K.D and M.R.S. participated in seminars sponsored by Tine BA. All other authors declare that they have no conflict of interest.
Authors’ contributions
OS, RJ and TØ designed the study and directed its implementation. MRS conducted the statistical analysis and drafted the manuscript. CKD contributed to the interpretation of results and writing of the manuscript. All authors were involved in the interpretation of the results, revision of draft manuscript and have approved the final manuscript.