Types of studies

Types of participants

Types of interventions

Types of outcome measures

See: Collaborative Review Group Search Strategy

Randomized controlled trials of erythromycin treatment in preterm infants were identified from the Cochrane Central Register of Controlled Trials (CENTRAL, The Cochrane Library, Issue 4, 2007, MEDLINE (1966 - December 2007) using the MeSH headings: Erythromycin, infant, newborn; infant, premature; gastrointestinal motility or gastric emptying; motilin; prokinetic agent.tw, EMBASE (1980 - December 2007), CINAHL (1982 - December 2007) and clinical trial registries (www.clinicaltrials.gov;www.controlled-trials.com). Relevant medical journals, bibliographies of articles from the database search, personal files, scientific meeting proceedings, and abstracts published in Pediatric Research (1990 - 2007) were manually searched. Language restrictions were not applied. Attempts were made to contact investigators of studies meeting the inclusion criteria to gather additional data for analysis. An attempt was made to identify unpublished studies from clinical trial registries.

Standard methodology for performing systematic reviews according to the Cochrane Neonatal Review Group was used.

Studies included in the review were randomized or quasi-randomized controlled trials involving preterm infants with a treatment group and a placebo group.

Quality of the trials included were evaluated by the following criteria:

1. Blinding of randomization;
2. Blinding of intervention;
3. Complete follow-up;
4. Blinding of outcome measurement.

The decision to include or exclude a specific study in the review was made by consensus of the two reviewers (EN, VS). Studies involving neonates and older infants and children were excluded if data for neonates could not be extracted.

A data collection form was created and the following information was extracted from the included studies: baseline characteristics of the participants, age at enrollment into study, inclusion and exclusion criteria, sample size, treatment and control group regimen, feeding protocol, and outcome measures. The reviewers were not blinded to the study authors or institutions. Each reviewer (EN, VS) abstracted data independently, and the results were compared and differences resolved by consensus.

Statistical analysis:
When there were at least two randomized controlled studies that evaluated the efficacy of erythromycin for prevention or treatment of feeding intolerance in preterm infants using the same outcome measures, data from those studies were combined to obtain an overall estimate of effect size. Treatment effects were described using risk ratio (RR) and risk difference (RD) for categorical outcomes, and weighted mean difference (WMD) for outcomes measured on a continuous scale. Statistical tests for categorical and continuous variables as recommended by the Cochrane Neonatal Review Group were used.

For details see: table of included studies.

Sixteen studies were identified, including 14 published manuscripts and two abstracts. One of the published manuscripts (Patole 2000) and one of the abstracts (Almonte 1999) were reports of the same study; this study is referred to as Patole 2002. Two studies were excluded because of their design as crossover studies (Costalos 2001; Costalos 2002). One study (Nogami 2001) was excluded because treatment assignment was not randomized (treatment or placebo was given based on the time period during which they were recruited). One study (Curry 2004) was excluded because it only involved infants with post-operative gastroschisis. One study (Jadcherla 2002) was excluded because it was a dose-response study with no placebo group included. Ten eligible randomized controlled trials were included in this review.

These ten studies used different inclusion criteria and used erythromycin in variable doses in different clinical settings. Three studies (Cairns 2002; ElHennawy 2003; Aly 2007) used low dose erythromycin (3 to12 mg/kg/day) as the study drug for preterm infants with feeding intolerance and thus fulfilled the a priori inclusion criteria for this review. The other seven studies did not meet the initial inclusion criteria, either because of the use of high dose erythromycin (> 12mg/kg/day) (Patole 2000; Stenson 1998; Ng PC 2001; Ng SC 2003; Madani 2004; Nuntnarumit 2006) and/or because erythromycin was used as a prophylactic treatment (Patole 2000; Stenson 1998; Oei 2001). Nonetheless, since these studies were still relevant to the objective of this review and might contribute to the knowledge of how erythromycin might be used to promote enteral feeding in preterm infants, a post hoc decision was made to broaden the scope of the review to include these studies.

For the description of studies and analyses of relevant outcome data, these included studies were grouped based on the clinical settings where erythromycin was used: Prevention (for at risk infants prior to or at initiation of feeds) and Treatment (for infants who exhibit signs of feeding intolerance). Where possible, subgroup comparisons of outcomes from studies using Low dose (defined a priori as 3-12mg/kg/day) versus High dose (> 12mg/kg/day) erythromycin were made.

PREVENTION STUDIES

Studies using low dose erythromycin (3 - 12 mg/kg/day)

Oei 2001 (10mg/kg/day): In the single centre randomized controlled trial by Oei 2001, 50 preterm infants with GA ≤ 32 weeks who were ready to start enteral feeds were enrolled into the study. Infants with major congenital anomalies, perinatal hypoxia, NEC and other abdominal pathological conditions, confirmed sepsis or intraventricular/periventricular hemorrhage more severe than grade I were excluded from the study. Infants enrolled in the study were given either erythromycin or a manufactured placebo.

Erythromycin-ethyl-succinate or placebo was administered enterally 30 minutes before feeding at a dose of 2.5 mg/kg/dose every six hours (10 mg/kg/day) for 10 days. The infants were fed according to a standardized protocol. The study medication was held if the infants' conditions required cessation of enteral feed or if the infant was transferred back to the referral nursery.

Primary outcome was days to achieve full enteral feeds. Secondary outcomes were number of episodes of emesis, gastric residuals > 30% of previous 6-hour feed volume, number of bile-stained gastric aspirates, and duration of total parenteral nutrition. Adverse events associated with the administration of erythromycin, such as increased toxicity to theophylline, abnormal liver functions, or sepsis were also reported. Outcomes were reported on 43 of the 50 infants who completed the trial.

Studies using high dose erythromycin (> 12mg/kg/day)

Stenson 1998 (45mg/kg/day): Seventy-five ventilated preterm infants with GA ≤ 30 weeks were enrolled at birth to determine the effect of erythromycin on reducing the early inflammatory response and, subsequently, the occurrence and severity of chronic lung disease in preterm infants (Lyon 1998). In a separate publication, Stenson (1998) reported the effect of erythromycin on establishment of enteral feeding in this study population. A discrepancy between the two publications (Lyon 1998; Stenson 1998) was noted in the number of infants reported [76 in Stenson (1998) vs. 75 in Lyon (1998)]. The infants were randomly assigned to receive erythromycin or standard treatment without erythromycin or a placebo. Erythromycin was started on day one at 15mg/kg/dose given intravenously three times daily (45 mg/kg/day) for seven days. The infants were fed according to a standardized protocol. The primary goal was to compare the concentration of inflammatory cytokines (interleukins IL-1 beta and IL-8, tumor necrosis factor alpha) in the treatment and control groups. Secondary outcomes included time to full enteral feed (defined as the point when milk intake of at least 150 ml/kg/day was given enterally), feed tolerance as evaluated by net enteral balance (nasogastric feed volume minus nasogastric feed residual volume) for the first 14 days of age, number of infants who vomited, number of glycerin suppositories given because of concern for delayed passage of meconium, and incidence of NEC. Data on feeding pattern were obtained from review of the nursing and medical records after the infant was discharged from the nursery. The staff responsible for the infants' feeding were not aware that the feeding pattern would be studied at a later date. Outcomes were reported on all 76 infants.

Patole 2000 (48mg/kg/day): In the study by Patole (2000), 73 preterm infants with GA ≤ 32 weeks who were ready for enteral feeding were enrolled. Infants with congenital anomalies were excluded. Infants were randomly assigned to receive erythromycin or a manufactured placebo. Erythromycin-ethyl-succinate or placebo was administered in a dose of 12 mg/kg every six hours (48 mg/kg/day) via orogastric tube from the time feeds were begun until full enteral feeds of 150 ml/kg/day were achieved or a total of two weeks of therapy was reached. The infants were fed according to a standardized protocol. Administration of other prokinetic agents was not allowed during the study period. Primary outcome was time to full enteral feed (defined as the point when milk intake of 150 ml/kg/day was achieved). Secondary outcomes included: feeding intolerance, defined by the presence of two major or one major plus two minor criteria or isolated bile-stained aspirate (Major: gastric aspirates > 30% of feed volume given over the previous four hours; bile-stained aspirates. Minor: vomiting; abdominal distension - increase in abdominal girth > 1.5 cm from measurement before feed), incidence of treatment failure (no clinical improvement in feeding intolerance or isolated bile stained/large gastric aspirates despite appropriate management on four consecutive occasions), duration of hospital stay, incidence of NEC, and occurrence of side effects associated with erythromycin [sepsis (positive blood culture with clinical deterioration), loose stool, hepatic dysfunction, cardiac arrhythmias]. Outcomes were reported on all 73 infants.

TREATMENT STUDIES

Studies using low dose erythromycin (3 - 12 mg/kg/day)

Cairns 2002 (12 mg/kg/day): In the multicenter randomized controlled trial by Cairns 2002, 60 preterm infants with GA < 37 weeks who failed to initiate or advance feeds for three or more days were enrolled in the study. Infants with NEC (suspect or confirmed) and ileus were excluded. Enrolled infants were given erythromycin at 3 mg/kg/dose or a normal saline placebo intravenously every six hours (12 mg/kg/day) until full enteral feeds (150 ml/kg/day) were achieved. The infants were fed according to a standardized protocol. Primary outcome was time to reach full feeds. Outcomes were reported on all 60 patients.

ElHennawy 2003 (6 mg/kg/day): In the single center randomized controlled trial by ElHennawy 2003, 27 preterm infants with GA between 29 to 36 weeks and birth weight (BW) between 900 to 2000 grams who were unable to achieve full feeds (150 ml/kg/day) within eight days of feeding initiation were eligible. Infants with NEC, intestinal myopathy, hepatitis, hepatic failure, congenital or chromosomal abnormalities, and sepsis within 48 hours of study entry were excluded. In addition, infants with unstable cardiorespiratory status requiring inotropic support, infants on opioids, or continuous feeding and those with arrhythmia were also excluded. Enrolled infants were randomized to receive erythromycin or a manufactured placebo.

Erythromycin-ethyl-succinate or placebo was administered in a dose of 1.5 mg/kg/dose every six hours (6 mg/kg/day) by orogastric tube 30 minutes before feeding for eight days. On the first day of the study (day 0), antroduodenal motor contractions were assessed using manometric probe two hours during fasting, two hours after a dose of erythromycin or placebo, and during and until one hour after a 10 ml/kg milk feed. Whole gut transit time was measured using 2 ml of carmine red dye one hour after the above milk feed, followed by gastric emptying time measurement according to method of George (George 1968) 20 minutes after the milk feed.

From days one to eight, feeding was advanced using a standardized protocol. Feeding intolerance was defined by the presence of abdominal distention (abdominal circumference > 15% of baseline measurement), gastric residuals > 25% of the volume of feed, or frank blood detected in the stool. Feeding volume was not advanced for 24 hours in the presence of feeding intolerance, although the criteria for withholding feeds were not explained.

Primary outcome was volume of feeding achieved by study day eight. Short-term secondary outcomes included the acute changes in antroduodenal motor pattern and gastric emptying in response to a single dose of medication; longer term secondary outcomes included number of hours where feeding was held, number of gastric residuals over 1 ml 2 ½ hrs post-feed, number of stools, weight and number of infants reaching full enteral feeds at study day eight, days to reach full enteral feeds and full oral feeds. Outcomes were reported on all 27 patients.

Aly 2007 (3 mg/kg/day): In the single center randomized controlled trial by Aly 2007, 60 preterm infants with GA < 37 weeks who were fed bovine protein-based formula and presented with feeding intolerance (defined by repeated gastric residual > 30% of previous six hour feed) were included in the study. Infants with major congenital anomalies, structural gastrointestinal anomalies, birth asphyxia, presence or history of NEC, confirmed sepsis, and organic abdominal illness were excluded.

Erythromycin-ethyl-succinate diluted to 4 mg/ml or equal volume of normal saline placebo was mixed into the milk feed prior to administration. Erythromycin was given at a dose of 1 mg/kg every eight hours until full enteral feeding (150 ml/kg/d for at least 24 hours) was established. Enteral and parenteral feedings were administered according to a standardized protocol.

The primary outcome was days required to achieve full enteral feeds (150 ml/kg/d for at least 24 hours). Secondary outcomes were number of episodes of feeding intolerance (as defined), weight gain since enrollment into study, duration of TPN, incidence of NEC, and length of hospital stay. Outcomes were reported on 49 of the 60 infants who survived to reach full enteral feeds.

Studies using high dose erythromycin (> 12mg/kg/day)

Ng PC 2001 (50 mg/kg/day): In the single center randomized controlled trial by Ng PC 2001, 56 preterm infants with BW < 1500 g who achieved less than half of full enteral feed (or < 75 ml/kg/day of milk feeds) by day 14 of age were enrolled into the study. Infants with lethal congenital anomalies, structural gastrointestinal abnormalities, NEC, or cyanotic congenital heart diseases, and those with major gastrointestinal surgery within the first two weeks of age were excluded. Enrolled infants were randomized to receive either erythromycin or normal saline placebo.

Erythromycin-ethyl-succinate or placebo was mixed into the milk feed and administered orally at a dose of 12.5 mg/kg every six hours (50 mg/kg/day). Treatment was initiated on all subjects on day 15 of age for a total of 14 days. The medication was held if the infants' conditions require cessation of enteral feed during the study period. Administration of other prokinetic agents was not allowed during the study period. Enteral and parenteral feedings were administered according to a standardized protocol.

Primary outcome was time (days) to achieve half, three quarters, and full enteral feeds. Secondary outcomes included adverse effect from erythromycin administration (prolonged QT interval, pyloric stenosis, bacterial colonization in the stool) and from the use of parenteral nutrition (culture positive sepsis and TPN cholestasis). Length of hospital stay and mortality were also recorded. Outcomes were reported on all 56 infants in the study.

Ng SC 2003 (15 mg/kg/day): In the single center randomized controlled trial by Ng SC 2003, 24 preterm infants with BW ≤ 1500 grams who either fail to start minimal enteral feeding by one week of age or to advance feeding up to 20 ml/kg/day one week after initiation of feeds were included. Infants with sepsis, major congenital anomalies, or growth restricted infants with a history of absent or reversed end-diastolic flow in the umbilical Dopplers in-utero were excluded from the study. Enrolled infants were randomized to receive erythromycin or a manufactured, isotonic placebo.

Erythromycin-ethyl-succinate or placebo was administered at a dose of 5 mg/kg every eight hours orally until one week after full enteral feeding (130 ml/kg/day) was achieved.

Primary outcome was time taken to achieve full enteral feeds. Secondary outcomes included time to regain birth weight, days of total parenteral nutrition, incidence of cholestatic jaundice, incidence of necrotizing enterocolitis, number of glycerin suppositories used, adverse event associated with the use of erythromycin (cardiac arrhythmias, pyloric stenosis, sepsis with multi-resistant organisms) and duration of hospital stay. In addition, with continuous pH probe monitoring from the time of study entry to one day after full feeds was achieved, the reflux index (percentage of monitored time with esophageal pH < 4 for 15 or more seconds) before and after the study was measured and the number of infants with reduction in reflux was calculated. Outcomes were reported on all 24 patients.

Madani 2004 (50 mg/kg/day): In the single center randomized controlled trial by Madani 2004, 57 preterm infants with GA < 36 weeks at birth who failed to tolerate half (75 ml/kg/d) of the full enteral feeds within five days of initiation of feeding were eligible. Infants with hypoxic injury, cyanotic congenital heart disease, previous gastrointestinal surgery, intestinal atresia or other congenital gastrointestinal conditions were excluded from the study. Enrolled infants were randomized to receive erythromycin or no treatment. A placebo was not used in this study.

Erythromycin-ethyl-succinate was administered at a dose of 12.5 mg/kg every six hours orally until full enteral feeding (150 ml/kg/day) was achieved.

Primary outcome was days required to achieve full enteral feeding (150 ml/kg/day), with subgroup analyses done on this outcome measure (infants ≥ 32 weeks, and < 32 weeks). Secondary outcomes included age at discharge from hospital, incidence of necrotizing enterocolitis and adverse effects associated with erythromycin use (stool colonization, prolonged QT interval, and pyloric stenosis). Outcomes were reported on all 57 infants.

Nuntnarumit 2006 (enteral administration 40 mg/kg/day for two days, then 16 mg/mg/day for five days): In the multicenter randomized controlled trial by Nuntnarumit 2006 , 46 preterm infants with GA < 35 weeks and BW < 1800 g who were at least five days of age and clinically stable (normal blood pressure and no recurrent hypoxemic/bradycardic episode) and who had feeding intolerance (defined by gastric residual > 50% of the previous three hour feed volume on more than occasions in 24 hours) were included in the study. Infants with major congenital anomalies, suspected or proven NEC within seven days prior to the onset of feeding intolerance, cyanotic congenital heart disease, major gastrointestinal surgery within two weeks, sepsis, metabolic or electrolyte abnormalities were excluded. Infants treated with fentanyl, indomethacin, pancuronium or vecuronium at the onset of feeding intolerance were also excluded. Enrolled infants were randomized to receive either erythromycin or a manufactured placebo. Erythromycin-ethyl-succinate diluted to 40 mg/ml or placebo was administered 30 minutes prior to feeding. Erythromycin was given enterally at a dose of 10 mg/kg/dose orally every six hours for the first two days, and 4 mg/kg every six hours for another five days. Administration of other prokinetic agents was not allowed during the study period. Enteral and parenteral feedings were administered according to a standardized protocol.

Primary outcome was time after initiation of treatment to achieve full enteral feeding (150 ml/kg/day) and maintained for at least three consecutive days. Secondary outcomes included incidence of NEC, septicemia, length of hospital stay, adverse effects associated with erythromycin (elevated liver enzymes, hypertrophic pyloric stenosis), and complications relating to the use of TPN (cholestasis and catheter-related sepsis). Outcomes were reported on all 46 infants.

For details see: table of included studies.

PREVENTION STUDIES

Oei 2001: Randomization was performed by parallel group design into blocks of 10 and was stratified by gestational age (over or less than 30 weeks). The randomization codes were concealed and were only available to the trial pharmacist off-site. The placebo was made to look identical to the treatment drug. Thus, blinding of randomization and intervention were accomplished. However, it was unclear whether outcome assessments were blinded. A sample size calculation was performed.

Stenson 1998: Infants were randomized to treatment or no treatment group by sealed envelopes. Blinding of intervention was not ensured, as a placebo was not given. Although the staff providing care were unaware of the investigators' intention to study the effect of the study drug on feeding tolerance, blinding of outcome assessments could not be ascertained. A sample size calculation was performed for the original study (Lyon 1998) to detect an effect of erythromycin on measures of lung injury and inflammation, but this sample size was inadequate to detect any effect of erythromycin on the primary outcome of time to full enteral feeds in the subsequent study (Stenson 1998).

Patole 2000: Infants were randomized using computer generated random numbers to treatment or placebo groups. Infants were allocated into the groups using sealed, coded envelopes. Blinding of intervention and outcome assessment were achieved. A sample size calculation was performed.

TREATMENT STUDIES

Cairns 2002: The randomization was stratified by study center and by postmenstrual age (over or less than 32 weeks). No other methodological information was available from the abstract.

ElHennawy 2003: Study infants were allocated to the treatment or placebo group using random number assignment generated by the pharmacy staff. The placebo was made to look identical to the treatment drug. The laboratory technician who analyzed the antroduodenal manometry recordings was unaware of the treatment assignment. Staff physicians, nurses, and investigators were all blinded to the treatment assignment. Therefore, blinding of randomization, intervention and outcome assessment were achieved. A sample size calculation was performed.

Aly 2007: Infants were randomized to erythromycin or placebo by opaque sealed envelopes. Randomization was done by parallel block of 10 stratified by gestational age (≤ 32 weeks and > 32 weeks). Blinding of randomization, intervention and outcome assessment were achieved. A sample size calculation was performed.

Ng PC 2001: Computer-generated randomization was performed by a designated staff member not involved in clinical care. The treatment drug and saline placebo were mixed into the milk feeds by non-clinical care staff as an attempt to mask the difference in appearance between the treatment drug and placebo. This method of masking was incomplete, as the treatment drug still had a distinct odor, so blinding of intervention could not be achieved. Microbiologists who were unaware of the treatment assignment analyzed stool cultures for colonization studies. Therefore, randomization, intervention, and outcome assessments were adequately blinded. A sample size calculation was performed.

Ng SC 2003: Study infants were randomly allocated to the two groups by sealed envelopes. The assignment was known only to a pharmacist, who was blinded to the care of the infants. The placebo was made to look identical to the treatment drug. Thus, blinding of randomization and intervention were achieved. However, it was unclear whether outcome assessments were blinded. A sample size calculation was performed.

Madani 2004: There was no information in the published paper regarding the method of randomization and blinding procedures. No sample size calculation was performed.

Nuntnarumit 2006: Infants were randomized to erythromycin or placebo by opaque sealed envelopes. Randomization was stratified by gestational age (< 32 weeks or ≥ 32weeks), and block of four randomizations were done to ensure balanced allocation. Blinding of randomization, intervention, and outcome assessments were achieved. A sample size calculation was performed.

PREVENTION STUDIES

1. Patient demographics
Oei 2001: Twenty-five infants were enrolled into the erythromycin group, and 25 in the placebo group. One infant in each group was withdrawn prior to the start of the study because of withdrawal of parental consent. Two infants in the placebo group were given cisapride during the trial due to recurrent vomiting and were taken out of the study. One infant in the erythromycin group was transferred back to the referring hospital before full feed was established. Finally, 1 infant in each group died of NEC and sepsis before feeding was established. The author reported data on the remaining 43 infants (22 in the erythromycin group and 21 in the placebo group). There was no significant difference in the baseline characteristics between the two groups.

Stenson 1998: Thirty five infants were enrolled in the erythromycin group and 41 in the control group. All infants were enrolled at birth. The two groups were similar in their demographic characteristics, except that the control group had a much higher male: female ratio than the treatment group (18/17 vs. 30/11, p = 0.05). Infants in the erythromycin group and the control group were started on enteral feedings at a similar median age of 3 (IQR 2 to 4) vs. 3 (IQR 2 to 5) days, p = 0.34.

Patole 2000: Thirty six infants were enrolled in the erythromycin group and 37 in the placebo group. There was no significant difference in the demographic characteristics between the two groups. Infants in the the treatment and placebo groups were started on enteral feedings at a similar median age of 5 (IQR 3 to 7.7) vs. 5 (IQR 3 to 7.5) days, p = 0.84.

2. Time to established full enteral feeds
Using low dose erythromycin (10 mg/kg/day), Oei (Oei 2001) showed that infants in the erythromycin group achieved full feeds significant earlier than the placebo group [mean (SD) days to establish full feeds of 6.0 (2.3) vs. 7.9 (3.5) days, p = 0.04]. However, the other two studies using high dose erythromycin (Stenson 1998; Patole 2000) failed to show any significant difference between the erythromycin and placebo group in days to achieve full feeds [median (IQR) of 8 (5 to 12) vs. 9 (6 to 14) days, p = 0.45] in Stenson 1998 and [median (IQR) of 3.9 (3.4 to 5.9) vs. 4.3 (3.4 to 6.8) days, p = 0.6] in Patole 2000. In addition, Patole (Patole 2000) found no significant difference between the treatment and placebo group in age at full feeds [median (IQR) of 9.5 (7 to 13) vs.11 (7 to 16) days of age, p = 0.9].

Meta-analysis of this primary outcome could not be performed because data in all but one study were reported as median and interquartile range. Studies were also different in the dose and duration of erythromycin used.

3. Duration of TPN

Only one study (Oei 2001) reported on the duration of TPN, which was not different between the erythromycin and placebo group during the study period [mean (SD) of 4.6 (5.1) vs. 4.0 (3.9) days, p = 0.68].

4. Weight gain

Two studies (Oei 2001; Patole 2000) reported on weight gain. In the study by Oei (Oei 2001), there was no difference in the time to regain birth weight between the erythromycin and placebo groups [mean (SD) of 14.9 (2.6) vs.15.3 (6.6) days, p = 0.83]. Similarly, Patole (Patole 2000) did not find a statistically significant difference between the erythromycin and placebo group in median (IQR) weight gain from birth to discharge [425 (162.5 to 1190) vs. 715 (450 to 1117) grams, p = 0.15].

Meta-analysis of this outcome could not be performed because one of the two studies reported this outcome in median and interquartile range, and because of the difference in the way weight gain was defined and reported in the two studies.

5. Adverse effects associated with the use of erythromycin
Two studies (Oei 2001; Patole 2000) reported on adverse effects associated with the use of erythromycin, but none was noted in either the erythromycin or placebo group during the study period.

6. NEC
All three studies reported on NEC. In the study by Oei (Oei 2001), one infant in each of the erythromycin and placebo group developed NEC. However, they were excluded from further analysis in the study. NEC was defined by pneumatosis intestinalis on radiograph or by findings on laparotomy in the two remaining studies (Stenson 1998; Patole 2000). In the study by Patole (Patole 2000), none of the infants developed NEC during the study period. In the study by Stenson (Stenson 1998), no significant difference in the number of infants with NEC was noted between the erythromycin and placebo group [2 (6%) vs. 4 (10%), p = 0.41].

Meta-analyses of this outcome from the two studies that reported on this outcome (Stenson 1998; Patole 2000) showed no difference in the incidence of NEC between erythromycin and placebo group. (typical risk ratio 0.59; 95% confidence interval 0.11 to 3.01; typical risk difference -0.02; 95% confidence interval -0.09 to 0.05).

7. Duration of hospitalization

One study (Patole 2000) reported on this outcome and found no difference in duration of hospitalization [43 (30.5-57) vs. 46 (23.7-69) days, p=0.89] between the erythromycin and placebo group.

8. Other outcomes
Studies by Oei (Oei 2001) and Stenson (Stenson 1998) reported on the incidence of feeding intolerance, although the definition of feeding tolerance differed in the two studies. Oei (Oei 2001) reported a significantly lower number of gastric aspirates (> 30% of previous feed) in the erythromycin group compared with the placebo group [mean (SD) of 1.1 (1.9) vs. 3.6 (2.4), p = 0.0007]. Stenson (Stenson 1998) reported that feeding tolerance [defined by the median net enteral balance (nasogastric feed volume minus nasogastric feed residual volume)], was not different between the erythromycin and control group at any day during the 14 days study period.

Oei 2001 reported no difference in the number of vomiting episodes between the erythromycin and placebo group [mean (SD) of 4.7 (6.4) vs. 3.9 (5.8) episodes, p = 0.66] and Stenson (Stenson 1998) did not find any difference in the number of infants who vomited [12 (34%) vs. 10 (24%) episodes, p = 0.59] between the erythromycin and placebo group during the study period.

Stenson (Stenson 1998) reported no statistically significant difference between the erythromycin and placebo group in the number of glycerin suppositories used [15 (43%) vs. 24 (58%), p = 0.17].

Stenson (Stenson 1998) reported no difference in mortality rate between the erythromycin and placebo group [7 (20%) vs. 8 (20%), p = 0.92].

TREATMENT STUDIES

1. Patient demographics
Cairns 2002: Thirty-two infants were enrolled in the erythromycin group and 28 in the placebo group. There was no significant difference in the baseline characteristics between the two groups. Infants in the erythromycin and placebo group entered the study at similar mean postmenstrual age of 29.4 (SD 1.7) vs. 29.1 (SD 1.9) weeks. Infants in the two groups did not differ in the number of days feeds were attempted before entering the study [8.1 (SD 4.7) vs. 8.6 (SD 6.6) days]. Eighty-five percent of all study infants were fed with breast milk.

ElHennawy 2003: Fifteen infants were enrolled in the erythromycin group and 12 in the placebo group. There was no significant difference in the baseline characteristics between the two groups. Infants in the erythromycin and placebo groups were started on enteral feedings at a similar mean age [8 (SD 12) vs. 5 (SD 4) days, p > 0.05], and the mean age at study entry [24 (SD 13) vs. 26 (SD 13) days, p > 0.05) ] and mean feeding volume at study entry [36 (SD 39) vs. 28 (SD 21) ml/kg/day, p > 0.05] were also similar.

Aly 2007: Sixty infants meeting the inclusion criteria were enrolled. Eleven infants (five in erythromycin group and 6 in placebo group) died before reaching full feeds. Among those who completed the study, 25 (13 infants ≤ 32 weeks and 12 infants > 32 weeks) were given erythromycin and 24 (12 infants ≤ 32 weeks and 12 infants > 32 weeks) placebo. Infants in the erythromycin and placebo group did not differ in baseline characteristics. Median age at enrollment between the erythromycin and the placebo group was similar [2 (IQR 2 to 24) vs. 2 (IQR 2 to 10) days, p = NS].

Ng PC 2001: Twenty-seven infants were enrolled into the erythromycin group, and 29 in the placebo group. There was no significant difference in the baseline characteristics between the two groups. Infants in the erythromycin group and placebo group were started on enteral feeding at a similar median age [median (IQR) of 6 (4 to 11) vs. 8 (5 to 11) days, p = NS) ]. The volume of feeds at the time of study entry was also similar [median (IQR) of 43 (17 to 58) vs. 51 (39 to 68) ml/kg/day, p = NS]. The percentage of infants on breast milk, formula, or a combination of both were also similar [breast milk 19 vs. 10%, p = NS; formula 30 vs. 28%, p = NS; mixed 51 vs. 62%, p = NS].

Ng SC 2003: Thirteen infants were enrolled in the erythromycin group and 11 in the placebo group. Infants in the two groups did not differ significantly in baseline characteristics. However, a statistically insignificant difference was noted in the mean BW of infants in the erythromycin group compared to the placebo group [mean (SD) of 806 (216) vs. 981 (285) g, p = 0.18], which might be attributed to the trend toward larger number of small-for-gestational age infants in the erythromycin group [4/13 (31%) vs. 1/11 (9%), p = 0.22]. Infants in the erythromycin group and those in the placebo group were similar in age at initiation of feeds [mean (SD) of 5.5 (3.6) vs. 6.3 (3.7) days, p = 0.90], age at study entry [mean (SD) of 19.7 (9.0) vs. 17.3 (5.3) days, p = 0.77], gestational age at study entry [mean (SD) of 29.9 (1.4) vs. 30.0 (2.6) weeks, p = 0.88], and volume of feeds at study entry [median (SEM) of 19.6 (4.1) vs. 19.6 (2.8) ml/kg/day, p = 0.42].

Madani 2004: Twenty-nine infants were enrolled in the erythromycin group, and 28 in the control group. Infants in the two groups did not differ in baseline characteristics. Infants in the erythromycin group and the control group were started on enteral feeds at similar age [mean of 3.8 vs. 3.0 days, p = 0.27].

Nuntnarumit 2006: Forty-six infants meeting the inclusion criteria were enrolled. That included 42 infants < 32 weeks and four infants ≥ 32 weeks. Twenty-three infants each were allocated to the erythromycin and the placebo group. There was no significant difference in the baseline characteristics between the two groups. Infants in the erythromycin and placebo groups were enrolled at a similar age [median (IQR) of 7 (6 - 8) days vs. 6 (6-8) days, p = 0.41]. The volume of feeds at study entry, although not statistically significant, was higher in the erythromycin group compared to the placebo group [median (IQR) of 25 (9 to 40) vs. 15 (7 to 31) ml/kg/d, p = 0.17]. A similar, statistically insignificant trend was seen in the percentage of infants fed breast milk, where the percentage was higher in the erythromycin group compared to the placebo group (39% vs. 17%, p = 0.19). However, the type of feeds was deemed an insignificant independent variable by a Cox regression analysis model.

2. Time to establish full enteral feeds
The three studies that used low dose erythromycin (Cairns 2002; ElHennawy 2003; Aly 2007) failed to show any significant difference between erythromycin and placebo in the times to establish full feeds in preterm infants <32 weeks' gestation with feeding intolerance. In the study by Cairns (Cairns 2002), no significant difference in times to full enteral feeds between the erythromycin and the placebo group was found [mean (SD) of 13.6 (10.1) vs. 16.4 (7.1) days, mean difference adjusted for strata was -2.7 days; 95% confidence interval (CI) -7.5 to 2.2, p = 0.27]. The study by ElHennawy (ElHennawy 2003) found no significant difference in the number of days to reach full enteral feeds [mean (SD) of 31 (15) vs. 36 (16) days, p > 0.05] and full oral feeds [mean (SD) of 54 (27) vs. 57 (25) days, p > 0.05 ] between the erythromycin and the placebo group. In the study by Aly (Aly 2007), erythromycin use was not associated with an earlier achievement of full feeds in infants born ≤ 32 weeks' gestation [mean (SD) 18.7 (8.1) vs. 20.7 (6.1) days, p=0.5]. While the first two studies (Cairns 2002; ElHennawy 2003) included only preterm infants < 32 weeks' gestation, the study by Aly (Aly 2007) is the only study that included infants > 32 weeks' gestation, and, in a subgroup analysis, found that erythromycin was associated with a significantly earlier achievement of full enteral feeds compared to placebo [mean (SD) of 10.5 (4.1) vs. 16.3 (5.7) days, p = 0.01] in that gestational age group.

For studies using high dose erythromycin (Ng PC 2001; Ng SC 2003; Madani 2004; Nuntnarumit 2006), the treatment effect of erythromycin compared to placebo on time taken to establish full feeds in preterm infants was not consistently demonstrated. In the study by Ng PC (Ng PC 2001), infants in the erythromycin group took a significantly shorter time to achieve 50% of enteral feeds [median (IQR) of 3.5 (2 to 7) vs. 6 (4 to 11.5) days, p < 0.05], 75% of enteral feeds [median(IQR) of 8.5 (6 to 19) vs. 13 (9 to 22) days, p < 0.05 ] and full enteral feeds [median (IQR) of 13.5 (8 to 22) vs. 25 (16 to 33) days, p < 0.0001] than those in the placebo group. The study by Nuntnarumit (Nuntnarumit 2006) found similar results, with infants in the erythromycin group taking a significantly shorter time to achieve full feeds than those in the placebo group [median (IQR) 7 (6 to 9) days vs. 13 (9 to 15) days, p < 0.001]. However, the study by Ng SC (Ng SC 2003) found no statistically significant effect of erythromycin in the number of days to achieve full enteral feeds (130 ml/kg/d) compared with placebo [mean (SEM) of 24.9 (2.9) vs. 30.8 (4.1) days, p = 0.17]. The age at full feeds were not different between groups as well [mean (SD) of 46.6 (18) vs. 52.1 (17.5) days, p = 0.37]. The study by Madani (Madani 2004), the only study among the four with subgroup analyses for infants < and ≥ 32 weeks' gestation, found that erythromycin was associated with a significantly earlier achievement of full enteral feeds compared to placebo only in those born ≥ 32 weeks' gestation [mean (SD) of 9.2 (1.5) vs. 13.5 (6.3) days, p = 0.03]. The studies by Ng PC, Ng SC, and Nuntnatumit (Ng PC 2001; Ng SC 2003; Nuntnarumit 2006) included preterm infants with a wide range of gestational age, although the majority of subjects in these studies were < 32 weeks' gestation at birth. Contrary to these three studies, Madani (Madani 2004) found that erythromycin was not associated with an earlier achievement of full feeds in preterm infants < 32 weeks groups [mean (SD) of 18.3 (11.4) vs. 17.6 (6.9) days, p = 0.84].

Meta-analysis of the primary outcome was not appropriate because all seven studies defined feeding intolerance differently; therefore, the patients chosen for each study were inherently different and the primary outcome measure cannot be combined for analysis.

3. Duration of TPN

Three studies (Aly 2007; Ng SC 2003; Nuntnarumit 2006) reported on this outcome. Aly (Aly 2007) used low dose erythromycin and found that erythromycin was associated with a shorter duration of TPN but only in infants > 32 weeks' gestation [mean (SD) of 5.5 (3.3) vs. 9.4 (4.8) days in those > 32 weeks, p = 0.03 and 12.6 (6.3) vs. 12.6 (5.4) days in those ≤ 32 weeks, p = 0.9]. The other two studies (Ng SC 2003; Nuntnarumit 2006) that used high dose erythromycin and studied infants mostly < 32 weeks found no difference in this outcome between groups [mean (SD) of 39.4 (13.8) vs. 43.3 (18.3) days (p=0.74) in Ng SC (Ng SC 2003); and median (IQR) of 13 (11 to 15) vs. 17 (13 to 25) days (p = 0.03) in Nuntnatumit (Nuntnarumit 2006)].

Meta-analysis of this secondary outcome was not performed because the three studies defined feeding intolerance differently, so comparing the duration of TPN in the three studies or combining such data for meta-analysis was not appropriate.

4. TPN-related cholestasis

Four studies (Aly 2007; Ng PC 2001; Ng SC 2003; Nuntnarumit 2006) reported on cholestatic jaundice as an outcome. However, only in one study (Ng PC 2001) was cholestatic jaundice defined (as conjugated bilirubin level >34 mmol/L). None of the studies noted any significant difference in the incidence of cholestatic jaundice between the erythromycin and the placebo group [2 (6.7%) vs. 4 (13.3%) (p=NS) in Aly (Aly 2007); 5 (19%) vs. 10 (35%) with conjugated bilirubin > 34 mmol/L (p = NS) in Ng PC (Ng PC 2001); 4 (31%) vs. 7 (64%) (p = 0.113) in Ng SC (Ng SC 2003); 3 (13%) vs. 2 (9%) (p = 1.00) in Nuntnarumit (Nuntnarumit 2006)].

The lack of information in how TPN cholestasis was defined made it inappropriate to combine this outcome for meta-analysis.

5. Weight gain

The four studies that reported on weight gain (ElHennawy 2003; Aly 2007; Ng SC 2003; Nuntnarumit 2006) showed no significant difference between infants in the erythromycin group versus those in the placebo group. Using low dose erythromycin, the study by ElHennawy (ElHennawy 2003) showed no significant difference in weight at study day eight between the erythromycin and placebo group [mean (SD) of 1625 (430) vs. 1611 (476) grams, p > 0.05]; similarly, in the study by Aly (Aly 2007), the erythromycin and placebo groups did not differ in weight gain since enrollment for study infants in either gestational age strata (≤ 32 or > 32 weeks). Using high dose erythromycin, Ng SC (Ng SC 2003) showed no difference between the erythromycin and placebo group in age when birth weight was regained [mean (SD) of 12.8 (4.4) vs. 16.8 (6.2), p = 0.11]; similarly, the study by Nuntnarumit (Nuntnarumit 2006) showed no difference between groups in days to regain birth weight [median (IQR) of 11 (10 to 14) vs. 12 (11 to 15) days, p = 0.49] and in the weight at discharge [median (IQR) of 2170 (1987 to 2587) vs. 2560 (2130 to 3600) grams, p = 0.06].

Meta-analysis of this outcome was not possible because weight gain was measured differently in the four studies. In addition, patients in the four studies were different because feeding intolerance was defined differently in these studies.

6. Adverse effects associated with the use of erythromycin
The six studies (ElHennawy 2003; Aly 2007; Ng PC 2001; Ng SC 2003; Madani 2004; Nuntnarumit 2006) that reported on adverse effects related to the use of erythromycin found no difference in incidence between the erythromycin and placebo groups. The study by Ng PC (Ng PC 2001) specifically reported no difference between erythromycin and placebo group in the QT interval during treatment [median (IQR) of 0.36 (0.35 to 0.4) vs. 0.38 (0.35 to 038) ms, p = NS]. No case of hypertrophic pyloric stenosis was reported in any of the studies.

In terms of septicemia, four studies (Aly 2007; Ng PC 2001; Madani 2004; Nuntnarumit 2006) reported on this outcome and found no significant difference between erythromycin and placebo group in incidence of sepsis 11 (36.7%) vs.15 (50%) of study infants (p = NS) in Aly (Aly 2007); 11 vs. 9 episodes of bacterial and fungal septicemia in Ng PC (Ng PC 2001); 5 (17%) vs. 7(25%) infants (p = NS) in Madani (Madani 2004); 3 (13%) vs. 4 (17%) of study infants (p = 1.00) in Nuntnarumit (Nuntnarumit 2006)]. Meta-analysis of this outcome from the four trials showed no difference in the incidence of septicemia (typical risk ratio 0.83; 95% confidence interval 0.47 to 1.45; typical risk difference -0.04; 95% confidence interval -0.17 to 0.08)

7. NEC

Five studies (Aly 2007; Ng PC 2001; Ng SC 2003; Madani 2004; Nuntnarumit 2006) reported on the incidence of NEC, but only one study (Nuntnarumit 2006) clearly defined NEC (as > Bell stage I). None of the five studies reported a significant difference in the incidence of NEC between the erythromycin and placebo group 3 (10%) vs. 4 (13.3%) (p = NS) in Aly (Aly 2007); 2 (7%) vs. 3 (11%), (p = NS) in Madani (Madani 2004); 1 (4%) vs. 4 (13%), (p = 0.61) in Nuntnarumit (Nuntnarumit 2006)]. The study by Ng PC (Ng PC 2001) reported no NEC events and Ng SC (Ng SC 2003) reported one infant in the placebo group who developed NEC one month after full feeds were attained.

Meta-analysis of this outcome was not appropriate. The inclusion criteria of feeding intolerance was defined differently in all five studies, thus the risk of developing NEC in these patients might also be different.

8. Duration of hospitalization

Five studies (Aly 2007; Ng PC 2001; Ng SC 2003; Madani 2004; Nuntnarumit 2006) reported on this outcome. Four of the five studies [all except Madani (Madani 2004)] failed to show any significant difference in length of hospital stay between the erythromycin and placebo group 33.2 (14) vs. 29.2 (13.8) days, p = 0.48 for GA ≤ 32 weeks strata; 18.3 (13.9) vs. 27.8 (10.4) days, p = 0.07 for GA > 32 weeks strata in Aly (Aly 2007); median (IQR) of 73 (64 to 97) vs. 86 (64 to 109) days (p = NS) in Ng PC (Ng PC 2001); mean (SD) of 98.3 (35.9) vs. 99.6 (58.6) days (p = 0.68) in Ng SC (Ng SC 2003); median (IQR) of 46 (24 to 74) vs. 60 (43 to 89) days (p=0.07) in Nuntnarumit (Nuntnarumit 2006). In the study by Madani (Madani 2004), sub-group analysis of the two gestational age strata showed that infants ≥ 32 weeks' gestation who received erythromycin were discharged at a significantly earlier age [mean (SD) of 12.9 (3.2) vs. 22.3 (8.9) days, p = 0.003]. However, for infants < 32 weeks' gestation, those who received erythromycin and those in the control group were discharged at a similar age [mean (SD) of 28.3 (10.8) vs. 26.4 (11.1) days, p = 0.64].

Meta-analysis of this outcome was not appropriate because the inclusion criteria of feeding intolerance was defined differently in all five studies, so the factors contributing to the hospital length of stay might also be different.

9. Other outcomes

Gastric residuals or feeds held

Three studies reported gastric residuals as a secondary outcome (ElHennawy 2003;Aly 2007; Nuntnarumit 2006). Using low dose erythromycin, ElHennawy (ElHennawy 2003) reported no significant difference between the erythromycin and placebo group in number of gastric residuals over 1 ml 2 ½ hrs post-feed [mean (SD) of 16 (8) vs. 21 (11) episodes, p > 0.05], and the number of hours where feeding was held [mean (SD) of 42 (44) vs. 37 (38) hours, p > 0.05 ]. However, using a similar dose of erythromycin, Aly (Aly 2007) found fewer episodes of significant gastric residuals (> 30% of previous 6 hour feed volume) in those treated with erythromycin but only in those born > 32 weeks [median (IQR) of 0 (0 to6) vs. 2 (0 to 8) episodes, p = 0.03] and not in those ≤ 32 weeks [median (IQR) of 2 (0 to 6) vs. 2.5 (0 to 8) episodes, p = 0.77]. Using high dose erythromycin, Nuntnarumit (Nuntnarumit 2006) reported a composite outcome of the number of feeds held/significant gastric residuals (>50% of feed) and found a significantly smaller number in the erythromycin group compared with the placebo group [median (IQR) 1 (0 to 2) vs. 9 (2 to 13), p < 0.001].

Physiologic outcomes

The study by ElHennaway (ElHennawy 2003) reported gastrointestinal physiologic outcomes. Using low dose erythromycin, the study showed no significant difference between the erythromycin and placebo group in rate of gastric emptying [mean (SD) of 47 (23) vs. 32 (20)% of a feed, 20 minutes after feeding, p = NS; 71 (20) vs. 64 (26)% of a feed, one hr after feeding, p = NS ], and whole gut transit time [mean (SD) of 83 (48) vs. 78 (77) p = NS. Antroduodenal motor contractions showed similar characteristics in both groups. Although the number of antral motor contractions was higher in the erythromycin group [numbers not shown, p < 0.02], they were similar after drug administration and after feeding [numbers not shown, p = NS]. Within the erythromycin group, the number of antral contractions varied significantly (ANOVA for repeated measures F = 10.0; p < 0.001).

The study by Ng SC (Ng SC 2003) evaluated the effects of high dose erythromycin on gastroesophageal reflux and showed no difference in the reflux indices before and after the study between the erythromycin and placebo groups [mean (SD) of 7.3 (15.5) vs. 13.6 (17.3)% before, p = 0.72 and 4.3 (7.1) vs. 0.3 (0.6)% after, p = 0.07]; the percentage of infants with a reduction in reflux in each group as a result of the study drug also did not differ significantly [23% vs. 36%, p = 0.57].

At present, administration of erythromycin in low or high doses cannot be recommended for the treatment of feeding intolerance, or for prophylaxis against feeding intolerance in preterm infants.

Further studies are needed to determine whether erythromycin in lower or higher doses is effective as a prokinetic agent in preterm infants > 32 weeks' GA with feeding intolerance or at risk of developing feeding intolerance.

Risk of bias table

Risk of bias table

Risk of bias table

Risk of bias table

Risk of bias table

Risk of bias table

Risk of bias table

Risk of bias table

Risk of bias table

Risk of bias table