To determine if infants aged 6 to 15 months with frequent parent-reported nighttime awakenings require reduced parental aid during a week of diphenhydramine hydrochloride treatment and 2 and 4 weeks after its discontinuation.
Double-blind, randomized, controlled clinical trial.
The study was conducted from May 1, 2004, through May 1, 2005; patients were recruited nationally.
Forty-four participants aged 6 to 15 months.
Placebo or diphenhydramine was administered in infants 30 minutes before anticipated bedtime.
Main Outcome Measures
The primary outcome was dichotomous: a parental report of improvement in the number of night awakenings requiring parental assistance during the intervention week, which ended on day 14. Secondary outcomes were improved sleep during the 2 weeks before days 29 and 43, parental overall happiness with sleep, and improved sleep latency.
On June 6, 2005, the data safety monitoring board voted unanimously to stop the trial early because of lack of effectiveness of diphenhydramine over placebo. Only 1 of 22 children receiving diphenhydramine showed improvement compared with 3 of 22 receiving placebo. To reach the a priori determined sample size and have a positive outcome (ie, rejecting the null hypothesis), the trial would have needed to enroll 16 more participants in each arm, with 15 of the 16 in the diphenhydramine group and 0 of 16 in the placebo group improving.
During 1 week of therapy and at follow-up 2 and 4 weeks later, diphenhydramine was no more effective than placebo in reducing nighttime awakening or improving overall parental happiness with sleep for infants.
Many parents encounter frustration coping with poor infant sleep, with 20% to 46% in various studies reporting that their infants have problems sleeping.1- 8 To confirm the importance that parents place on adequate infant sleep, one need only visit a local bookstore.9- 14 Most physicians have little formal training to help them advise new parents about sleep. In one study15 of more than 600 pediatricians, the mean score on a test about sleep was 60%, whereas only 25% of primary care physicians report more than 6 hours of sleep education.15- 17
Results of a few studies have shown controlled crying, prolonged crying, and parental education to be effective strategies for modifying infant sleep habits.18- 25 Although the extinction method (commonly referred to as crying out) is popular, professionals and lay authors have criticized it.19,21,26 Another method, used with surprising frequency by parents and often recommended by physicians, is to use drugs to sedate infants before sleep. In one study in 62 children with recognized sleep difficulties, 71% of families reported using medicine for sleep.27 The most widely studied sleep drugs in children are antihistamines, most of which are not readily available.6,28- 31 The antihistamine most often used, diphenhydramine hydrochloride (Benadryl; Pfizer Inc, Morris Plains, NJ), generally has been studied in adults, with only 1 study in children aged 2 to 12 years.32- 34 Although we could find no published studies in children younger than 2 years, a survey of 671 primary care pediatricians in the United States showed that 49% reported recommending antihistamines in children aged 0 to 2 years for sleep problems in the past 6 months.35
Because of the wide use of diphenhydramine and the lack of evidence in this setting, we investigated whether diphenhydramine improves infant sleep. Our goal was to undertake a practical clinical trial to examine short-term and long-term effects of diphenhydramine on infant sleep.36 Because of the wide use and safety of diphenhydramine, we did not believe an efficacy study was indicated and thus undertook a patient-oriented effectiveness study.
We hypothesized that, on the basis of diphenhydramine's sedative properties, treated children would be more likely to fall asleep without their parents on nights they had been treated. In addition, we hypothesized that this treatment would lead children to associate the crib with sleep and comfort, and, therefore, they would need less assistance from their parents when they awakened during the night than would children receiving placebo. Finally, we hypothesized that because diphenhydramine would create a better association of the crib with sleep during the week of treatment, treated children would continue to have better sleeping habits at 2- and 4-week follow-up.
The Trial of Infant Response to Diphenhydramine (TIRED) study was a double-blind, randomized, controlled, clinical trial conducted at multiple centers with allocation concealment. This study was started and completed before registration in clinicaltrials.gov was required. Participants were recruited by using flyers in practice-based research networks, primary care offices, and day-care centers. In addition, national advertisements were placed in magazines and newspapers and on multiple Web sites targeted at parents throughout the United States. The Johns Hopkins University institutional review board approved the study. Written informed consent was obtained from all participating parents. Contact throughout the study was performed via telephone calls and mailings.
Children were eligible for enrollment if they (1) were aged 6 to 15 months, (2) had a parent or caregiver who spoke English or Spanish, (3) awoke 2 or more times per night, and (4) slept in a crib.
Children were ineligible if they (1) had chronic illness (defined as their taking medicine on a regular basis), (2) had seen a sleep or developmental specialist in the past, (3) were known to be allergic to diphenhydramine, (4) had a diagnosis of obstructive sleep apnea or snoring, (5) were actively ill, or (6) were born before 35 weeks of gestation.
A permuted block randomization scheme was used to balance the treatment assignments between diphenhydramine and placebo. In advance, a listing of random assignments was prepared, consisting of blocks of either 4 or 8 assignments. Each block consisted of an equal number of random assignments to each treatment. Two of us (D.M. and M.D.-W.) generated the sequence, and one (D.M.) was responsible for following this assignment and enrolling participants. Identical opaque bottles were labeled as either A or B and consisted of 100 mL of medicine or placebo, both cherry flavored. The pharmacist and a designated Johns Hopkins staff member not associated with the study were the only individuals privy to the label information. All other research personnel were blinded to treatment assignment, and at no time during the study did they obtain label information. Data were analyzed before revealing label information.
Diphenhydramine and placebo were randomly assigned after patient enrollment information and informed consent were obtained. The Community Drug Compounding Center (Pittsburgh, Pa) prepared and bottled the study medication and placebo. Parents were instructed, after a 7-day observation period, to give their children the study medication once daily, 30 minutes before anticipated bedtime, on study days 8 through 14 (the intervention week). Diphenhydramine concentration was 12.5 mg/5 mL and was prescribed at 1 mg per kilogram of body weight. Placebo was flavored purified water, colored to match diphenhydramine and was administered at 1 mg per kilogram of body weight. All researchers were blinded to interventions. Parents were informed of the proper dose by one of us (D.M.). Parents in both groups also were given a sleep brochure produced by the American Academy of Pediatrics (Elk Grove Village, Ill).
The primary outcome was dichotomous: either improved or not improved number of night awakenings requiring parental intervention during the intervention week. Parents reported “yes,” “no,” or “unsure” when asked “Did your child require decreased parental intervention during the past week?” The responses of “no” and “unsure” were considered treatment failure in calculating differences between groups and the number needed to treat. Parents also tracked their child's sleep for the first 28 nights with a diary. If 2 parents lived with the child, they were encouraged to talk about outcomes, but only 1 parent relayed information to the study coordinator. In addition, on days 0, 15, 29, and 43, parents were asked, “On a scale of 1 to 10, where 1 is very unhappy and 10 is extremely happy, what would you rate your scale of happiness in regard to your child's overall sleep, in the past 2 weeks?”
All outcomes were assessed on days 15, 29, and 43. The primary outcome was improved sleep during intervention week, referred to as day 15. Secondary outcomes were improved sleep during the 2 weeks before days 29 and 43, parental overall happiness with sleep, and improved sleep latency.
Other successful interventions showed improvement in more than 50% of children per study.29,31,37 Sample size was determined on the basis of the primary outcome of improvement of nighttime sleep after 1 week of intervention. We anticipated a 10% improvement in the placebo group and 40% in the intervention group. Setting statistical significance at .05 and 80% power, we calculated that the trial needed 38 participants per group to reject the null hypothesis.
Baseline characteristics were compared between the placebo and diphenhydramine groups by using either a χ2 test or Fisher exact test for categorical variables. For continuous variables, the t test was used; the Wilcoxon rank sum test was used for ordinal or skewed variables. All analysis was performed by using the intention-to-treat principle. The difference in the proportion improving between the placebo and diphenhydramine groups was estimated along with the associated 95% confidence interval (CI). The number needed to treat was calculated at the end of the study. All statistical analysis was conducted by using Stata 8.0 software.38
On June 6, 2005, approximately 1 year after the first participant was enrolled, a preplanned analysis was conducted by the data safety monitoring board, and they voted unanimously to stop the trial early. This decision was reached because of the lack of effectiveness diphenhydramine demonstrated in the multiple measured outcomes.
During the 12-month enrollment, from May 1, 2004, through May 1, 2005, 109 parents inquired about the study. Eleven infants did not meet inclusion criteria because of the child's age or because the child's sleep problem did not include nighttime awakenings. An additional 9 parents declined participation after receiving preliminary study information. Thus, 89 were mailed a written informed consent form, and 44 parents agreed to participate after reviewing it (Figure). One participant in the placebo group and 2 in the diphenhydramine group were lost to follow-up 2 weeks after the intervention. All 3 participants failed to improve immediately after the intervention and at days 29 and 43 treatment was considered to have failed (“not improved”) in the intention-to-treat analysis.
Participant flow diagram.
There were no major differences with respect to any of the baseline characteristics between the treatment groups (Table 1). In the placebo group, 3 participants (14%) had improved nighttime awakening requiring parental aid compared with 1 (5%) in the diphenhydramine group after the intervention week (difference between treatment and placebo group, −9%; 95% CI, −26% to 8%, number needed to treat, 11; range, 12 to infinity) (Table 2).
At intention-to-treat analysis at 29 and 43 days, a statistically significant difference emerged with 2 additional parents reporting improvement in the placebo group, an addition to the 3 who previously reported improvement. These 5 parents reported improvement at days 29 and 43 (Table 2). Parents also reported their overall happiness with their child's sleep on a scale of 1 (very unhappy) to 10 (extremely happy). The results on days 0, 15 (immediately after the intervention), 29, and 43 were nearly identical for the 2 groups (Table 3).
At day 0, parents in the placebo group reported a mean (SD) number of 3.6 (1.4) nighttime awakenings requiring intervention, whereas those in the diphenhydramine group reported 3.2 (1.0; estimated mean difference, 0.4 awakenings; 95% CI, −0.38 to 1.1). During the intervention week, parents in the placebo group reported a mean (SD) number of 2.1 (1.6) nighttime awakenings, compared with 2.0 (1.1) in the diphenhydramine group (estimated mean difference, 0.2 awakenings; 95% CI, −0.65 to 1.0).
Parental report of sleep latency, the time from when the child was placed in the crib until initial sleep, for the 7 days before intervention was a mean (SD) of 22.5 (17.9) minutes for the placebo group and 21.6 (38.5) minutes for the diphenhydramine group (estimated mean difference, 0.8 minutes; 95% CI, −17.8 to 19.5). During the week of intervention, the mean (SD) time in the placebo group decreased to 16.0 (13.5) minutes, whereas that in the diphenhydramine group decreased to 11.9 (15.2) minutes (estimated mean difference, 4.2 minutes; 95% CI, −4.7 to 13.0).
The data safety monitoring board reviewed interim data, but the researchers remained blinded. To achieve power to detect the originally hypothesized effect of 40% improvement with diphenhydramine and 10% with placebo, we would need to enroll another 16 participants per group. The null hypothesis of no difference between groups then could be rejected only if no participants in the placebo group improved and 15 of the 16 in the diphenhydramine group improved. The data safety monitoring board considered this so unlikely that it advised us to stop the trial.
If the 3 people lost to follow-up were not considered treatment failure, thus not in accordance with conducting intention-to-treat analysis, our results at days 29 and 43 remain almost identical, with only a 1% difference. If we assumed that these children actually improved, then at days 29 and 43 there would no longer be a significant difference (−18%; 95% CI, −40% to 4%).
No parents reported adverse effects that caused them to stop study participation early. One patient in the diphenhydramine group acquired hand, foot, and mouth disease during the study and stopped after 5 days of intervention. Investigators and the data safety monitoring board judged that this was not related to study intervention. Two other children in the placebo group had mild adverse effects, one with hyperactivity and the other with diarrhea, and one in the diphenhydramine group also was reported as having hyperactivity. All conditions were reported by the parents to be mild.
According to parental report, 37 (84%) of the 44 infants complied with all 7 days of intervention. Four of the remaining 7 missed only 1 day of medicine, whereas the other 3 missed 2, 3, and 4 days.
Many in the medical and lay community accept diphenhydramine as effective treatment for sleep problems.39 Unfortunately, this attitude is based on anecdote and studies of adult physiological interactions. However, the TIRED study results demonstrated that at the most commonly used dose, diphenhydramine may play no role in treating infant sleep problems. Although parents did not report increased hyperactivity in the diphenhydramine group compared with the placebo group, it is also possible that diphenhydramine caused low-level hyperactivity in children, thereby negating the sleep benefits seen in some adults.
We believed that to evaluate diphenhydramine in the setting in which it is most commonly used and to assess effectiveness, a practical clinical trial would need to be conducted. Thus, we did not use methods such as actigraphy, video recording, or any measures foreign to the natural setting. In addition, we asked parents not to vary their routines, so the time the infant was placed in crib, bedtime routines, feeding schedules, and so on were not standardized, with the exception that parents were asked not to stay with the child once placed in the crib and to place the child in the crib awake. We asked parents not to stay with the infant because of the theory that changes in long-term sleep are achieved with the child learning to fall asleep alone. We believe that, rather than detracting from our study, this practical patient-oriented approach is an advantage of our study.
Our findings are consistent with those of a recent study by Paul et al,40 who studied whether over-the-counter medications were superior to placebo for treating nocturnal cough and sleep difficulty. They enrolled patients aged 2 to 18 years who had respiratory tract infections, with 33 patients in the diphenhydramine group, 33 in the dextromethorphan group, and 34 in the placebo group. They found no differences in parental or child sleep in any of the groups. However, our results differed from those of the only other study with diphenhydramine solely meant to study sleep.34 We can only speculate on the differences, but one of the major differences is that we enrolled only patients aged 6 to 15 months with difficulty staying asleep at night, whereas the other study's investigators enrolled children aged 2 to 12 years with various sleep disorders. The children in their study may have had more serious disease, and diphenhydramine may have positive results in these types of children and in older children. However, these investigators did not use intention-to-treat analysis, so they excluded 14% of patients from their final analyses.
Learning to fall asleep without parental intervention is the basis of most infant sleep theory.10,41 Although the results of the TIRED study did not confirm this theory, we also examined other aspects of sleep such as parental happiness or satisfaction, parental belief that medicine worked, and sleep latency. Diphenhydramine was not superior to placebo for any of these outcomes. Although not statistically significant, the diphenhydramine group demonstrated a larger decrease in sleep latency (relative decrease of 45% vs 29% in placebo; data not reported). Such an observed decrease in the time it takes an infant to fall asleep may explain why some physicians and parents have believed diphenhydramine improves sleep.
Several important limitations of our study deserve mention. Most importantly, we used a dose of 1 mg per kilogram of body weight. Dose recommendations range from 1 to 1.5 mg per kilogram of body weight. We used a conservative dose for 2 reasons: (1) we consulted with many local physicians, and this dose was commonly used; (2) we could find no previous studies in this age group, so we wished to err on the side of safety. At a higher dose, diphenhydramine may be more sedating and would show some positive outcomes with infant sleep. Also, we used no biomarkers to assess compliance. However, parents had no incentives for participation in the study except to see their children's sleep improved, and they were highly motivated, as indicated by the fact that only 5 parents did not complete the entire 28-day diary. Our chosen effect size was purposely large, which may have resulted in our missing smaller effects of diphenhydramine on infant sleep and which was exacerbated by the data safety monitoring board’s stopping the study early. Finally, as discussed, our study relied on parental reports of sleep behaviors, and no objective criterion to validate these reports was used. Proper infant sleep clearly is defined culturally, and many parents would never consider methods such as extinction or sedation; therefore, this selection bias needs to be considered when interpreting study results. However, regardless of the disease state being investigated, drug studies by definition are composed of a select population that not only has a disease but also is willing to take medicines.
Our study results reinforce the need for rigorous studies of common primary care interventions and illustrate that such studies are feasible.36 Despite dozens of published books about infant sleep, we found only 10 level 1–evidence studies about sleep in healthy children. Because of a lack of evidence, parents and physicians often are forced to make important decisions on the basis of anecdotal experiences.
We have shown that it is possible to conduct quality patient-oriented primary care research in young children to provide parents and physicians with evidence to help them make their decisions. Important aspects of a young child's health care, such as proper food intake and sleep routine, often are based on assumptions. Our study results illustrate that such assumptions may be wrong and should be subjected to study in life laboratories not university laboratories. Increased emphasis on pediatric effectiveness research would allow treatment of children to become more evidence based.
Correspondence: Dan Merenstein, MD, Department of Family Medicine, Georgetown University, 3750 Reservoir Rd, 220 Kober Cogan, Washington, DC 20007 (firstname.lastname@example.org).
Accepted for Publication: January 12, 2006.
Author Contributions:Study concept and design: Merenstein, Diener-West, Halbower, and Rubin. Acquisition of data: Merenstein and Krist. Analysis and interpretation of data: Merenstein, Diener-West, Krist, Halbower, and Rubin. Drafting of the manuscript: Merenstein, Diener-West, and Krist. Critical revision of the manuscript for important intellectual content: Merenstein, Diener-West, Halbower, and Rubin. Statistical analysis: Merenstein and Diener-West. Obtained funding: Merenstein. Administrative, technical, and material support: Merenstein and Halbower. Study supervision: Merenstein, Diener-West, Halbower, and Rubin.
Funding/Support: The study was funded by the Robert Wood Johnson Clinical Scholars Program, Robert Wood Johnson Foundation (Drs Merenstein and Diener-West). All work was independent of funder.
Acknowledgment: We thank the data safety monitoring board members for their dedication and hard work: Alicia Bazzano, MD, MPH; Leigh Culver, PhD; Paul Lipkin, MD; and Thomas Newman, MD, MPH. We also thank Alicia I. Arbaje, MD, MPH, for invaluable help with translation and study input. We thank Community Drug Compounding Center of Pittsburgh, Pa, and pharmacist Susan Freedenberg, RPh, for drug development. We thank Fairfax Family Practice Centers, Southeast Denver Pediatrics, and Kathryn Casey, MD, for help with recruitment.
Merenstein D, Diener-West M, Halbower AC, Krist A, Rubin HR. The Trial of Infant Response to DiphenhydramineThe TIRED Study—A Randomized, Controlled, Patient-Oriented Trial. Arch Pediatr Adolesc Med. 2006;160(7):707–712. doi:10.1001/archpedi.160.7.707