Treating infant colic with the probiotic Lactobacillus reuteri: double blind, placebo controlled randomised trial

Trial design

This was a phase III, double blind, randomised placebo controlled trial. Details of the protocol are described elsewhere.³⁴

Participants

Between August 2011 and August 2012 we recruited participants from a range of services accessed by carers of newborn babies in Melbourne, Australia: the Royal Children’s Hospital emergency department, the Royal Children’s Hospital Unsettled Babies Clinic (a referral based outpatient clinic), Tweddle Child and Family Health Centre (a mother-infant parenting centre), two maternal child health centres (universal nurse health checks), and paediatricians at the Royal Children’s Hospital and in private practices. Families could also contact the study team to be involved.

We recruited healthy term infants less than 13 weeks of age with infant colic, defined by modified Wessel’s criteria of crying or fussing for three hours or more a day for three days or more over seven days. The study doctor assessed eligibility through a structured telephone or face to face interview. Fussing was defined as “behaviour that is not quite crying but not awake and content either.”³⁵ We excluded infants who weighed less than 2500 g at birth; failed to thrive; had major medical problems; had an allergy to cow’s milk protein (confirmed by symptom resolution after a two week trial of hypoallergenic formula or a cow’s milk protein-free diet in mothers of breastfed infants); or took solids, antibiotics, or L reuteri at the start of the trial. We also excluded breastfed infants whose mothers took L reuteri at trial start, and caregivers with insufficient English to complete questionnaires.

Intervention

The treatment was L reuteri DSM 17938 (0.2×10⁸ colony forming units per drop) in an oil suspension, given once daily (five drops orally) for one month. The placebo was maltodextrin in the same oil suspension with the same appearance, colour, and taste as the treatment, identically packaged and stored. Caregivers were instructed to dispense five drops of study liquid onto a spoon to administer to their infants at the same time each day. The dosing of L reuteri is the same as that used in the three previous trials showing effectiveness of the probiotic in managing colic.^{32 33 36}

Randomisation, allocation concealment, blinding

Randomisation was stratified by method of infant feeding (breast fed v formula fed) and age (≤6 weeks v >6 weeks, owing to the natural peak in crying at around six weeks of age). We allocated infants who were both breast fed and formula fed to the formula fed stratum. An independent statistician prepared the computer generated randomisation schedule using a block size of two to maintain balance between treatment arms within each stratum. The trial was double blind, whereby treatment allocation was concealed from all study investigators and participants at every phase, including measurement of outcomes. Study investigators allocated infants to one of four randomisation strata according to their age and feeding type. An independent pharmacist (who was not part of the study team and had no access to the study database) then allocated the study identification number to each infant and dispensed the study product according to the randomisation schedule. This randomisation schedule was not revealed to the study investigators until all outcomes analyses at one month were complete. The treatment and placebo drops were packaged identically such that study investigators and families were blinded to treatment allocation throughout the study.

Outcomes, adherence, adverse events

Table 1 shows the outcome measures at the different time points. Funding allowed analysis of a subset of faecal samples for microbial diversity, E coli load, and calprotectin levels collected at one month post-intervention. We measured adherence by weekly parental report using a diary, with non-adherence defined as carers reporting discontinuing the intervention within two weeks of trial start. We weighed all bottles before and after intervention to determine equivalence in adherence between the groups. We also assessed for L reuteri in 65 faecal samples. Parents recorded any adverse events weekly. For the six month outcomes, we allowed the last 91 carers in the trial to fill in one day of the study diary instead of two to reduce the study burden and improve retention rate.

Statistical analyses

Participant flow and baseline characteristics

Overall, 167 families were randomised (51% of 329 eligible families; 85 probiotic, 82 placebo). Retention to primary outcome was 76% (127 infants analysed, 67 probiotic, 60 placebo); 11 and 29 families were excluded for loss to follow-up and missing data, respectively (fig 1). Other than more boys in the placebo group, the baseline characteristics of both groups were similar (table 2). Recruitment occurred mainly from the Royal Children’s Hospital emergency department (73%).

Primary outcome

Figure 2 plots the steady decline in daily duration of infant cry or fuss in both groups over the study period (see supplementary table 1 for values). Table 3 shows and compares all primary and secondary outcomes at 1 and 6 months.

Overall, the two groups were similar on almost all outcomes. At 1 month, the probiotic group cried or fussed an adjusted mean of 49 minutes/day more than the placebo group (95% confidence interval 8 to 90, P=0.02). This was mainly due to more fussing, not more crying, in the probiotic group (adjusted mean difference 52 minutes/day, 19 to 84, P=0.002) at 1 month. The probiotic group fussed significantly more than the placebo group at all time points from day 7 to 1 month (see supplementary table 1). Although daily duration of cry or fuss decreased over the study period in both groups, the decline by 1 month was greater in the placebo group than probiotic group, with a mean difference in reduction of 46 minutes (95% confidence interval −7 to 99, P=0.09). At 6 months, the groups did not differ for duration of cry or fuss (adjusted mean difference 7 minutes/day, 95% confidence interval −47 to 34, P=0.75).

By 1 month, 40% (n=27) of the treatment group and 48% (n=29) of the placebo group showed at least a 50% reduction in duration of cry or fuss (odds ratio 0.7, 95% confidence interval 0.4 to 1.3, P=0.23). Results from per protocol and intention to treat analyses were similar, as were regression models performed with and without bootstrapping. No treatment effects were evident in the longitudinal analysis using random effects regression models, and no trends in treatment effect over time were evident.

Secondary outcomes

The two groups were similar on all secondary outcomes at 1 and 6 months (table 3). At 1 month, the probiotic group slept 47 minutes less a day than the placebo group (95% confidence interval −90 to −3, P=0.04). Similarly, laboratory analyses of faecal samples at 1 month (treatment group n=31, placebo group n=34) showed no differences between groups in faecal microbial diversity, calprotectin levels, or E coli load. However, infants in both groups with at least a 50% reduction in duration of cry or fuss at 1 month had significantly lower calprotectin levels than non-responders (responders n=50, non-responders n=52, mean difference 96.6 mg/kg, 95% confidence interval 5.1 to 188.1, P=0.04), even though microbial diversity scores and E coli load were comparable.

Subgroup analyses

At 1 month, the total duration of cry or fuss was similar between treatment and placebo groups among exclusively breastfed infants (adjusted mean difference 19 minutes, 95% confidence interval −49 to 87, P=0.57, n=50). Durations of cry or fuss did not differ significantly between the two groups for infants aged more than 6 weeks, and for infants with and without a family history of atopy.

In contrast, formula fed infants in the probiotic group cried or fussed an adjusted mean of 78 minutes more (95% confidence interval 25 to 132, P=0.005, n=71), and infants aged 6 weeks or less in the probiotic group cried or fussed an adjusted mean of 88 minutes more (16 to 160, P=0.02, n=50) than corresponding infants in the placebo group. In both cases, this largely resulted from an increase in fussing (adjusted mean difference 63 minutes/day, 21 to 104, P=0.004 for formula fed infants, and 67 minutes/day, 8 to 126, P=0.03 for infants aged ≤6 weeks).

Adherence and adverse events

Reported adherence was 82% (probiotic) and 80% (placebo), with equivalent ingested weights in both groups (mean total product ingested per baby 3.3 g v 3.9 g, P=0.19). Analysis of the faecal samples showed that 45% of the probiotic group and none of the placebo group were colonised with L reuteri at 1 month post-intervention. No study related adverse events occurred.

Strengths and limitations of this study

This is the largest randomised controlled trial of probiotic intervention in infants with colic to date. It is the first to include formula fed infants and to extend outcomes to infant sleep, maternal mental health, and family functioning. We used rigorous methods and high quality measures; in particular, the well validated Baby Day Diary reduces carer opportunity for recall bias and is sufficiently detailed to permit separated analyses of infant crying versus fussing. Retention and reported adherence rates were high, with the final sample size adequate to achieve the predetermined statistical power and significance.

The trial has some limitations. Most of the infants were recruited from an emergency or urgent care setting; whereas our results may not therefore generalise to infants whose carers do not seek assistance, treatment in this group may be less relevant. Funding permitted completion of laboratory analyses of a subset of faecal samples collected at 1 month only. The low colonisation rate of L reuteri in only 45% of the treatment infants may reflect low sensitivity of the assay used. It could also reflect low adherence, although we ascertained good adherence by parental report, and equivalent adherence between the two groups by weights of study bottles before and after ingestion of study product. It remains possible that the infants did not receive a dose of L reuteri large enough to establish an effect; however, the trial’s probiotic dosing was the same as that of three previous trials reported as effective.^{32 33 36} Moreover, the rate of colonisation of L reuteri in normal healthy infants given the probiotic is unknown and possibly highly variable. We excluded infants with a likely diagnosis of allergy to cow’s milk protein, so our results cannot generalise to this subset of infants.

Interpretation in light of other studies

Despite using the same probiotic strain, formulation, and dosing, our negative trial results contrast with the three previous trials of L reuteri in infants with colic,^{32 33 36} even when considering only breastfed infants. Our trial also failed to demonstrate positive probiotic effects in formula fed infants with colic. There are several possible reasons for the differences in results. Firstly, our sample size was larger than any previous trial and was adequately blinded, unlike Savino and colleagues’ trial.³² Secondly, ours was the only trial that used a validated diary to measure the primary cry or fuss outcome, reducing significantly the likelihood of recall bias and enabling us to uniquely separate out crying and fussing outcomes. Thirdly, the baseline gut microbiota of infants from Melbourne, Australia, could differ from that of Polish and Italian infants, or could be less susceptible to the effects of probiotics. Previous studies have shown that infants with allergic disease have different gut microbiota according to different geographical locations, even among countries with westernised lifestyles.^{44 45 46 47 48} Fourthly, 60% of our infants had a family history of atopy, in keeping with local epidemiology.⁴⁹ However, the prevalence of a family history of atopy was balanced between groups, and subgroup analysis of infants with and without an atopic family history was similar. Fifthly, it may be that L reuteri in formula fed infants is only effective if infants are also receiving a dairy-free diet. Finally, we included infants who were receiving proton pump inhibitors at the start of the trial. However, proton pump inhibitors are ineffective in reducing crying in irritable infants^{50 51} and therefore should have no impact on results.

Unexpectedly, infants in the probiotic group fussed significantly more than infants in the placebo group at all time points from day 7 to 1 month in the adjusted analyses. Further analysis revealed that carers from the placebo group had 40 to 60 minutes/day more contact (carrying/moving) with their infants at all time points, although these differences never reached significance (all P>0.05). The relevance of this is uncertain, as increased contact with carers has been shown to reduce fussing and crying in infants without colic^{52 53} but not in infants with colic.⁵⁴ Previous studies have shown that in exclusively breastfed infants the use of hypoallergenic formula or dairy elimination diets in mothers is somewhat effective in reducing irritability in infants with colic.^{10 11 12} Interestingly, more infants in the placebo group were receiving hypoallergenic formula at 1 month (eight in placebo group v four in probiotic group, P=0.08), and more exclusively breastfed infants in the placebo group had mothers who were on dairy elimination diets at baseline (table 2) and at 1 month (10 in placebo group v 3 in probiotic group, P=0.03). However, these numbers were all small, and in further analyses the results were not altered by controlling for dairy exclusion and use of hypoallergenic formulas.

In our preliminary analysis of faecal samples, calprotectin levels, although unrelated to trial arm, were lower in those with resolved colic. This finding is in keeping with a previous small study showing faecal calprotectin levels to be higher in infants with colic than in those without,⁸ but contrary to a larger study showing no differences in faecal calprotectin levels between infants with and without colic.⁹ Considering that faecal calprotectin levels are associated with intestinal neutrophilic inflammation, it is conceivable that the more than 50% reduction in crying or fussing may reflect lower levels of intestinal inflammation by a mechanism unrelated to probiotic ingestion. These can be further explored by better designed studies investigating faecal calprotectin levels and infant crying.

When the results of our trial are added to the meta-analysis of previous randomised controlled trials of L reuteri for the management of infant colic,³¹ the overall effectiveness of the probiotic is diminished: mean difference 48 minutes/day (95% confidence interval -84.8 to −11.3) compared with 68 minutes/day (−99.8 to −35.6); figures 3 and 4). However, it should be noted that the meta-analysis puts significant weighting on the study by Savino and colleagues,³² which was an unblinded, open label trial.

Conclusions and future directions

Based on the null findings from this large, rigorous trial, and the results from the updated meta-analysis, which has potential biases, we recommend caution in formulating recommendations regarding use of L reuteri DSM 17938 for reducing crying or fussing in breastfed, and particularly in formula fed, infants with colic. Currently, there are at least five other similar randomised controlled trials taking place in different parts of the world.⁵⁵ It will be valuable to pool data from all existing and ongoing rigorous trials into an individual patient level meta-analysis to investigate with greater certainty which subgroups of infants with colic, if any, could benefit from probiotics.