The Effect of Dairy Product Ingestion on Human Immunodeficiency Virus–Related Diarrhea in a Sample of Predominantly Gay Men: A Randomized, Controlled, Double-blind, Crossover Trial

Background  In the highly active antiretroviral therapy (HAART) era, chronic diarrhea remains common in human immunodeficiency virus (HIV) illness. Empirical lactose avoidance is often advised despite lack of evidence of benefit in a population at risk for osteopenia and malnutrition.

Methods  The a priori hypothesis was that moderate lactose ingestion would not worsen diarrhea in this population. We used a double-blind, noninferiority, randomized crossover trial in a community setting of primary and tertiary care HIV clinics. The participants all had chronic diarrhea and were a volunteer sample of 49 predominantly white HIV-infected men who have sex with men. They ingested 240 mL of low-fat milk (12 g of lactose) and lactose-free milk during 2 separate study periods. The primary outcome was mean difference in stool weight between the 2 study periods in the 8 hours after milk ingestion. Lactose was judged not to worsen diarrhea if this difference did not exceed 167 g in 8 hours with 95% certainty.

Results  Forty-eight (98%) of 49 participants were male. Median age, CD4 cell count, and viral load were 42 years (range, 20-62 years), 390 cells/mL (range, 20-1100 cells/mL), and 112 copies/mL (range, <50 to >500 000 copies/mL), respectively. Thirty-nine participants (80%) were taking HAART medication. Ten participants (20%) were lactase deficient. The mean difference in stool weight between the 2 study periods was −41.3 g/8 h (upper 95% confidence limit, −13.5 g) for the entire group and +11.3 g/8 h (upper 95% confidence limit, +47.4 g) for the lactase-deficient group.

Conclusions  Moderate lactose ingestion does not worsen diarrhea in HIV-infected persons with chronic diarrhea, regardless of lactase status. Therefore, avoidance of modest quantities of milk may not be justified in this population.

Diarrhea remains common in persons with human immunodeficiency virus (HIV) despite the use of potent antiretroviral medications.¹ However, the cause of this problem is changing, with a decline in opportunistic infections and an increase in noninfectious causes.² Noninfectious HIV-related diarrhea has been attributed to HIV enteropathy,³ medications,⁴ and lactase deficiency (LD).^5,6

Lactose intolerance is characterized by abdominal bloating, flatulence, diarrhea, and cramping abdominal pains after the ingestion of foods containing lactose. It is caused by a deficiency of lactose, a small bowel brush border enzyme essential to the digestion of lactose. Hydrogen breath testing is a highly sensitive and specific test for LD.⁷ In the general population, it is well recognized that many persons with LD can tolerate moderate quantities of lactose in their diet (eg, 12-18 g, which is equivalent to 1-1.5 cups of milk) without developing symptoms.^8-10 Conversely, subjective perception of milk intolerance is a poor predictor of true LD.^9-11

Although LD is more prevalent in HIV-infected adults than in uninfected adults,¹² there are no blinded and controlled studies evaluating the effect of lactose ingestion on HIV-related diarrhea. Despite this lack of evidence and despite the potential risks of malnutrition¹³ and osteopenia¹⁴ in this population, avoidance of dairy products is often recommended in the management of HIV-related diarrhea.^15-17 In a survey of 21 HIV health care providers in the Toronto, Ontario, area, 15 reported that they recommend empirical lactose avoidance to their patients with HIV-related diarrhea (unpublished data, October 1999).

Our hypothesis was that moderate ingestion of lactose-containing milk would not worsen the diarrhea of HIV-infected persons with chronic noninfectious diarrhea compared with ingestion of lactose-free milk.

This was a randomized, double-blind, noninferiority, crossover trial in which HIV-infected participants drank lactose-containing and lactose-free substrate followed by stool collection and hydrogen breath testing on 2 separate days. Outcomes were stool weight and symptom score. All participants were recruited and studied between September 2000 and March 2003.

HIV-infected persons between the ages of 18 and 80 years who reported at least 3 loose to watery bowel movements per day for the previous 4 weeks or more were eligible. Persons with fewer than 3 bowel movements daily but who required daily antidiarrheal medications were also eligible. The antiretroviral regimen had to be stable over the previous 4 weeks and for the duration of the study.

Exclusion criteria included pregnancy, LD predating the diagnosis of HIV, severe systemic illness, malignancy, hospitalization, use of antibiotics other than for prophylaxis against opportunistic infections in the month prior to study enrollment, any colon-cleansing procedure within 1 week of the first study period, or diagnosis of other causes of diarrhea such as inflammatory bowel disease.

Participants were recruited from primary care and tertiary care HIV clinics in Toronto as well as from the community using a variety of methods including newspaper advertisements, direct communication with treating physicians, and posters in clinics and AIDS advocacy agencies.

The study protocol was approved by the research ethics boards at St Michael's Hospital, the University Health Network, and the University of Toronto. All participants gave written, informed consent. This investigation was carried out in accordance with the regulations set out in the Declaration of Helsinki.

Prior to study entry, infection and other non–HIV-related causes of diarrhea were ruled out with microbiologic studies (negative findings from 3 stool cultures, 3 microscopic examinations for ova and parasites with special stains for cryptosporidium, and 1 stool assay for Clostridium difficile toxin) and endoscopy.

For the 24 hours prior to each study period, participants followed a lactose-free diet and avoided all antidiarrheal medications. They ate an evening meal that was low in hydrogen-producing foods. Prior to each study period, they abstained from smoking for 3 hours and fasted for 12 hours. Human immunodeficiency virus medications were taken at the same time on both study days.

On both study days, participants ingested test substrate and underwent concomitant 5-hour hydrogen breath testing and an 8-hour stool collection period. They were randomized, using a computer-generated list created by the hospital research pharmacist, to the order in which they received the test substrates. The primary investigator (J.T.) obtained written informed consent and enrolled the participants. Study personnel and participants were blinded to the randomization schedule. There was a minimum 1-week washout period between the 2 study days, but both had to be completed within 6 weeks.

Alveolar breath samples were collected using an AlveoSampler (QuinTron Instrument Co, Milwaukee, Wis). A baseline breath specimen was taken; the test substrate was ingested; and subsequent specimens were collected hourly. Concentrations of hydrogen and methane in the breath samples were measured by gas chromatography (model SC MicroLyzer; QuinTron). If the hydrogen concentration rose by 10 ppm or more from baseline^9,18 or if there was a rise of 100% or more in methane levels from baseline in participants with a baseline methane level of 5 ppm or more¹⁹ on the lactose study day, a diagnosis of LD was made. Otherwise, participants were categorized as lactase persistent (LP).

At the start of each study period, participants drank 240 mL of test substrate. On the lactose-free study day, the test substrate was lactose-hydrolyzed 2% milk (Lactaid; Parmalat, Brampton, Ontario) and on the lactose study day, it was sweetened 2% milk containing 12.5 g of lactose. To render it indistinguishable from the lactose-free milk, which has a sweetened taste as a result of the lactose hydrolyzation, 0.23 g of aspartame was added to the lactose-containing milk. Using the sensory triangle test,²⁰ 30 untrained volunteers, who were not known to be HIV-infected, were not able to distinguish the 2 test substrates (data not shown).

The primary outcome was the stool weight in grams over an 8-hour period after test substrate ingestion. This duration of stool collection was based on reported orofecal transit times after lactose ingestion.²¹ To ensure that our collection period was not too short, participants completed the symptom questionnaire described below for the 16 hours after each study period.

The secondary outcome was self-reported participant symptoms, scored using a questionnaire previously used in the evaluation of lactose intolerance⁹ in which 4 symptoms (abdominal bloating, flatulence, pain, and diarrhea) were evaluated on a 6-point Likert scale (0-5). The number of bowel movements and of flatuses during each study period were recorded.

For the primary outcome, a noninferiority boundary of an increase of 167 g in stool weight during the lactose period was selected, reflecting a change of approximately 50% in stool output for HIV-infected persons taking protease inhibitors.²² Our hypothesis would be supported if the 1-sided upper 95% confidence limit (CL) of the mean difference in stool weight between the study periods was less than this noninferiority boundary.^23-25

For the secondary outcome, the 4 symptom scores were tallied, and the mean difference in the total score between the 2 periods was calculated. Two subgroup analyses were performed by analyzing the subgroup-intervention interaction term.²⁶ The first subgroup analysis, which was prespecified, compared stool weight and symptom scores in the LD and LP groups. In the second subgroup analysis, stool weight and symptoms were compared for those who reported lactose avoidance or definite symptoms associated with lactose ingestion to determine whether lactose ingestion worsened their diarrhea.

Mean differences with 95% confidence intervals (CIs) are reported. Nonparametric bootstrapping techniques²⁷ were used for the CIs where the data were markedly skewed. Comparisons were made using the t test, the Wilcoxon signed rank test, or the Wilcoxon rank sum test depending on the distribution of the data. A P value of less than .05 was considered significant. All data were analyzed using the SAS statistical package, version 8.2 (SAS Institute, Cary, NC).

The sample size was calculated for a noninferiority type trial assuming no difference in stool weight between study periods, a standard deviation of 258 g in 8 hours,²² and a moderate correlation between the 2 study periods (ρ = 0.3). For our sample size of 50 participants, there was a 96.4% probability that the 1-sided 95% upper CL for the mean difference in stool weight between study periods would exclude an increase of 167 g.

Results are reported for only 49 of the 50 participants because 1 participant violated the study protocol. This person was unable to complete the protocol within a 6-week period because of hospital closures related to a public health emergency in Toronto.

Baseline demographics are reported in the Table. Forty-eight (98%) of the participants were male; 45 were white; 1 was African Canadian; 1 was Latino; and 2 were Aboriginal. The median CD4 count was 390 cells/mL, while the viral load was undetectable in 21 participants (43%). Thirty-nine (80%) of the participants were taking highly active antiretroviral therapy (HAART) medication while 42 (86%) were taking antidiarrheal medications. Participants reported a median of 4 bowel movements per day.

Nineteen participants (39%) reported “definite gastrointestinal symptoms” with intake of lactose-containing foods, while 15 (31%) were unsure about this relationship, and the remaining 15 (31%) felt that there was no relationship. Eighteen (37%) of the participants reported avoiding lactose-containing foods.

The baseline characteristics of the 10 participants (20%) found to be LD did not appear to differ from the overall group (Table).

In each study period, participants had a median of 2 bowel movements (range, 0-7) and 2 episodes of flatus (0-15 in lactose-free period; 0-16 in lactose period).

For the entire group, the mean ± SD stool weights were 126.3 ± 132 g and 167.6 ± 194 g during the lactose and lactose-free study periods, respectively. The correlation in stool weight between the study periods was ρ = 0.12. The mean difference between the 2 periods (stool weight_lactose – stool weight_lactose-free) was –41.3 g, with a 1-sided upper 95% CL of –13.5 g, which was less than the +167-g noninferiority boundary (Figure 1).

The LD group had a mean difference in stool output between the 2 periods of +11.3 g (upper 95% CL, +47.4 g), while in the LP group, the mean difference between the 2 periods was –54.8 g (upper 95% CL, –17.8 g). The stool output in the LD group differed significantly from the LP control group (66.1 g difference between groups; 95% CI, 11.84-122.3 g; P = .03) (Figure 2).

For the entire group, the mean symptom scores were 4.7 and 5.5 during the lactose and lactose-free study periods, respectively. The mean difference between the 2 periods was –0.8 (95% CI, –1.9 to 0.3).

For the LD group, the mean difference in symptom scores between the 2 periods was 1.0 (95% CI, –2.5 to 4.5), whereas for the LP group, the mean difference was –1.2 (95% CI, –2.3 to –0.1). There was no significant difference in symptoms between the 2 groups (–2.2 difference between groups; 95% CI, –5.0 to 0.5; P = .11).

There were no significant differences between the study periods in the total and diarrhea-specific symptom score for the 16 hours immediately following each period for the overall group or for the LP and LD subgroups.

Self-reported lactose avoidance and participant perception of symptoms were not significantly associated with a change in stool weight nor were they associated with increased symptoms during the lactose period (data not shown).

In this study, we have shown that empirical ingestion of moderate quantities of lactose-containing milk does not increase stool output or worsen symptoms in a sample of patients with diarrhea, predominantly white HIV-infected men who have sex with men, when compared with ingestion of lactose-free milk. In the subgroup analysis, there was a significant difference in stool output between the LD and the LP groups, largely due to a greater stool output in the latter group on the lactose-free study day (Figure 2). The reason for this finding in the LP group is unclear. We did not identify a missed lactose effect in this subgroup in the immediate poststudy period. Some might interpret our results to mean that lactose-hydrolyzed milk causes diarrhea in LP HIV-infected persons; however, there is no recognized biological mechanism to support this conclusion.

Despite the statistically significant difference in stool weight between the LD and LP groups, the clinical importance of this finding is questionable. The upper 95% CL of the mean difference in stool output between the 2 study periods suggests that at most, lactose would produce a mean increase of 48 g of stool in an 8-hour period in the LD group. This difference represents approximately 3 tablespoons of stool and is unlikely to be clinically significant. Similarly, the upper 95% CL of the difference in symptoms between the 2 study periods for the LD subgroup is small.

Our findings echo those from the non-HIV literature where moderate lactose ingestion has not been shown to worsen symptoms in otherwise healthy individuals.^9,11 In a nonblinded and uncontrolled study, Corazza et al¹² demonstrated an association between lactose ingestion and symptoms in the LD subset of a group of HIV-infected adults. Our findings are particularly robust because our randomized and controlled study was a noninferiority trial.^23-25 We can be 95% certain that the true difference in stool output between the 2 periods is no more than upper CL of this difference, −13.5 g.

There are several limitations to this study. Primarily, there is no gold standard method to evaluate diarrhea. We used a change in stool weight as our primary measure because (1) it is easily quantifiable; (2) it is a logical choice for the osmotic diarrhea of LD; and (3) abnormal stool weight is generally agreed by clinicians to objectively define diarrhea.²⁸ Patient symptoms were also assessed because stool weight has been criticized as a criterion for failing to capture the more subjective symptoms associated with diarrhea such as stool frequency or abdominal cramps.²⁹ Our conclusion that lactose does not worsen HIV-related diarrhea is strong because it was consistent across 2 independent measures: one, objective and quantifiable; the second, subjective but patient centered.

The size of and the timing of the lactose load are also potential limitations. This is particularly a concern in a noninferiority trial where a smaller lactose load could introduce bias (ie, toward a finding of no difference between the 2 test substrates). However, this dose is equivalent to that found in 1 glass of milk, which is representative of the amount of lactose that might be found in a normal diet; for this reason, this dose is extremely clinically relevant. Our choice of 12.5 g of lactose is in keeping with other studies where a similar physiologic dose of lactose was used.^8,9,11,18 Additionally, we evaluated the effects of lactose ingestion at a single point in time only. However, long-term lactose ingestion is likely to cause even fewer symptoms because persons with LD may develop tolerance to the effects of lactose (due to adaptation of colonic bacteria).³⁰

The crossover design we used has some potential limitations, including period and sequence effects.^31,32 Formal statistical tests of period and sequence effects are discouraged because they are difficult to interpret.³³ We did not perform these tests but note that our study results appear similar when reported by order of intervention and study period (data not shown). We believe that the nature of the clinical problem being studied and the trial design we selected are appropriate to a crossover trial,^33,34 making important period or sequence effects unlikely.

Finally, the generalizability of our findings may be limited because our study sample was predominantly white men who have sex with men, with 80% taking antiretroviral medications. The composition of our sample may be of concern because the prevalence of LD is known to vary by racial group. In addition, most of our participants were taking antiretroviral medications; it is conceivable that LD related to HIV enteropathy may be more common in those who are not taking antiretroviral agents.

Deliberately, we evaluated the effect of lactose ingestion on a general population of HIV-infected individuals with diarrhea because we were interested in testing our hypothesis under the conditions that we had observed in clinical practice, that is, the recommendation of empirical lactose avoidance. In fact, we found that more than one third of our participants reported avoiding lactose, although in our sample, neither self-reported lactose avoidance nor perception of symptoms after lactose ingestion was associated with objective evidence of worsening diarrhea after lactose ingestion.

In summary, we have shown that moderate 1-time ingestion of lactose-containing milk does not worsen diarrhea significantly in HIV-infected persons, regardless of the presence or absence of LD. Therefore, avoidance of moderate quantities of milk (ie, 1 cup, equivalent to 12 g of lactose) may not be justified in this population. Further studies would be useful, specifically studies in different patient populations and with larger doses of lactose given over longer periods, to make more definitive recommendations.

It is well recognized that elimination diets such as a lactose-avoidance diet may lead to weight loss and that lactose avoidance is associated with osteoporosis.³⁵ In HIV-infected persons, diarrhea is also a risk factor for weight loss,³⁶ and even in the HAART era, weight loss remains a predictor of mortality.³⁷ Given the risks of malnutrition¹³ and osteopenia in this population,¹⁴ practitioners may wish to encourage their HIV-infected patients with diarrhea to include moderate quantities of dairy products in their diets.

Correspondence: Jill Tinmouth, MD, PhD, FRCPC, Division of Gastroenterology, Sunnybrook & Women's College Health Sciences Centre, 2075 Bayview Ave, Room HG40, Toronto, Ontario M4N 3M5, Canada (jill.tinmouth@sunnybrook.ca).

Accepted for Publication: January 24, 2006.

Author Contributions: Dr Tinmouth had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Financial Disclosure: None.

Funding/Support: This study was funded by the Canadian Association of Gastroenterology, Oakville, Ontario; Canadian Institutes for Health Research, Ottawa, Ontario; Canadian HIV Trials Network, Vancouver, British Columbia; and Ontario HIV Trials Network, Toronto.

Role of the Sponsor: None of the study's funding agencies had a role in the study design or conduct; in the collection, analysis, or interpretation of data; in the preparation, review or approval of the final manuscript; or in the decision to submit the paper for publication.

Acknowledgment: We thank the Canadian Association of Gastroenterology, Canadian Institutes for Health Research, Canadian HIV Trials Network, and the Ontario HIV Trials Network for their support of this study. In addition, this study benefited tremendously from the expertise and assistance of Gail Stewart, RN, Laura Parsons, CPhT, Ann Kosinski, CPhT, APMR, Shawn Launders, BScPharm, and Cindy James, RN, BSCN.