MANUSCRIPT CITATION

African Neonatal Sepsis Trial (AFRINEST) group, Tshefu A, Lokangaka A, Ngaima S, Engmann C, Esamai F, Gisore P, Ayede AI, Falade AG, Adejuyigbe EA, Anyabolu CH, Wammanda RD, Ejembi CL, Ogala WN, Gram L, Cousens S. Oral amoxicillin compared with injectable procaine benzylpenicillin plus gentamicin for treatment of neonates and young infants with fast breathing when referral is not possible: a randomised, open-label, equivalence trial. Lancet 2015 May 2; 385(9979):1758-66. PMID: 25842223

REVIEWED BY

Dr Shuchita Gupta MD
Pediatrics (Scientist),

Dr Ramesh Agarwal MD DM
Neonatology (Additional Professor)

Newborn Health Knowledge Centre
WHO Collaborating Center for Training and Research in Newborn Health
Division of Neonatology
All India Institute of Medical Sciences (AIIMS), Delhi, India

TYPE OF INVESTIGATION

Treatment

QUESTION

P=In Neonates and young infants (0-59 day old) with respiratory rates >60 breaths/minute and whose families do not accept or cannot access referral level care
I= is oral amoxicillin (twice per day for 7 days)
C= compared to injectable procaine benzylpenicillin plus gentamicin (once per day for 7 days)
O= equally effective
T=by day 8 of treatment

METHODS

• Design: Randomized, open label, equivalence (non-inferiority) trial
• Allocation: For allocation concealment, treatment codes were printed on small pieces of card folded once and sealed in an opaque envelope (2 sets of color-coded envelopes are used, one for each age group)
• Blinding:Unblinded, this was an open label trial
• Follow-up period: on days 4, 8, 11 and 15 after the day of enrollment
• Setting: It was a community based, multicentre trial conducted at 5 sites- 1 in Democratic Republic of Congo (DRC), 1 in Kenya and 3 in Nigeria. Treatment was provided at a health facility or at home.
• Patients: Infants between 0–59 days of age who had tachypnoea (defined as respiratory rate of ≥60 breaths per minute) whose parents did not accept or could not access referral level care and gave consent to participate in the study.
  Infants were excluded if they had any signs of severe infection (defined as poor feeding on observation, movement only when stimulated, severe chest indrawing and axillary temperature ≥38.0°C or <35.5°C), were critically ill (characterized by the presence of any of the following signs: unconsciousness, convulsions, unable to feed at all, apnea, unable to cry, cyanosis, dehydration, bulging fontanel, major congenital malformations inhibiting oral antibiotic intake, active bleeding requiring transfusion, surgical conditions needing hospital referral and persistent vomiting defined as vomiting following 3 attempts to feed the baby within one-half hour), were very low weight (<1500 g at the time of presentation) or had been hospitalized for illness in the last 2 weeks or were previously enrolled in the study.
• Intervention: Infants were randomized individually within each site and age stratum to two groups- intramuscular injections of gentamicin in the range 4.0–7.5 mg/kg/d and procaine penicillin in a dose of 50,000 units/kg once daily for 7 days depending on the age of the young infant (treatment regimen A; control group), and oral amoxicillin in suspension form in a dose of 100 mg/kg/d (for infant ≥2 kg) and 75 mg/kg/d (for infant <2kg), divided in 2 equal doses given twice daily for 7 days (treatment regimen B; intervention group). Randomization was stratified by postnatal age into two groups: 0-6 days and 7-59 days and used a fixed block size of 8. The randomization scheme was computer-generated offsite at WHO using statistical software STATA version 10.0 (STATA Corp., College Station, tX) by a person not involved in the study.
  Injections were given once daily by a health worker at a facility or at home, whereas oral medicines were given at home by the mother under the supervision of community health workers (CHWs). A daily assessment by a health worker (treatment nurses in DRC and Kenya and CHW in Nigeria) was conducted to identify possible treatment failures (primary outcome).
• Outcomes:
  • Primary outcome: The primary outcome was treatment failure by day 8 post enrollment visit.  Treatment failure is defined as any one of the following:
    • Death
    • Clinical deterioration, defined as emergence of any sign of severe infection or critical illness at any time after enrollment (as defined in exclusion criteria), or hospitalization any time after enrollment
    • Persistence of fast breathing on day 4 or recurrence after day 4 up to day 8
    • Development of a serious adverse event (other than death) that is related to the study antibiotics, e.g., organ failure, anaphylactic reaction, severe diarrhea, disseminated and severe rash.
  • Secondary outcomes:
    • Death occurring 9-15 days after enrollment
    • Relapse, defined as fast breathing that disappears on day 8 of enrollment and reemerges between days 9 and 15, or development of any sign of severe infection or critical illness signs between days 9 and 15 after enrollment
    • Adherence to the study therapy between days 1 and 8
• Analysis and sample size: 
  • The authors had initially calculated a sample size of 1150 children per group, based on power 90% based on 95% CI for the risk difference of treatment failure. They had anticipated a failure rate of 8% in the reference group and an absolute equivalent margin of 4% (relative reduction of equivalence margin of 50%), assuming that 85% infants would receive adequate treatment and assessment. However, later the treatment failure rates were found to be higher than expected (approximately 20%), while almost 95% infants were receiving adequate treatment and assessment. It was therefore estimated that the original sample of 1150 infants per group had 80% power to show equivalence with an absolute margin of 5%, assuming failure rate of 20% in the injectable (reference) group. The primary analysis was a per-protocol analysis. The authors included the infants in analysis if they met the pre-defined criteria for per protocol analysis:
    • If outcome was assessed on all days (4, 8, 11, and 15); or on day 4 and any of the days, 11 and 15, AND
    • If they fully/partially adhered to treatment:
      • fully adherent – those who received all antibiotic doses till 7 days of treatment unless treatment failure earlier
      • partially adherent – received all antibiotics till day 3 or by the time of treatment failure, and at least 50% subsequently till day 7 or treatment failure AND did not receive any non-study injectable antibiotic before day 8 (except given for treatment failure, or any non-study oral antibiotic on days 1-3.
  • For the primary outcome of treatment failure, authors calculated the difference in the risk of treatment failure between the two groups, along with a 95% CI for the risk difference. They also performed an intention-to-treat analysis along with the per-protocol analysis and reported unadjusted results of both the analyses. Secondary analysis was done to study adverse events, and tested if the results were modified by variables such as age, sex, weight, maternal age, and maternal education using regression analysis. One planned interim analysis was done by DSMB after 66% of participants were enrolled and treated, which recommended continuation of the study.
• Patient follow-up:  For the primary outcome of treatment failure, authors calculated the difference in the risk of treatment failure between the two groups, along with a 95% CI for the risk difference. They also performed an intention-to-treat analysis along with the per-protocol analysis and reported unadjusted results of both the analyses. Secondary analysis was done to study adverse events, and tested if the results were modified by variables such as age, sex, weight, maternal age, and maternal education using regression analysis. One planned interim analysis was done by DSMB after 66% of participants were enrolled and treated, which recommended continuation of the study.

MAIN RESULTS

The study randomized 2333 infants between 0-59 days of age (2196 included in per-protocol analysis) to oral amoxicillin twice a day for 7 days (intervention group) or intramuscular injection of procaine penicillin once a day for 7 days (control group).

For the primary outcome of treatment failure by day 8 after enrollment, a pooled analysis across the three study sites showed the incidence to be 19.0% in the intervention, and 22.0% in the control arm. The failure was primarily due to persistence of fast breathing on day 5 (16.0% in the intervention versus 18.0% in the control arm). Though there was no significant difference between the groups (RD -2.6%; 95% CI: -6.0% to 0.8%), the lower limit of the CI for RD was outside the equivalency margin of ±5% (as it was -6.0%, i.e., below -5.0%). This indicates that the new treatment, i.e., oral amoxicillin may be equivalent to the control treatment i.e., intramuscular penicillin and gentamicin, though it is not statistically conclusive. It is important to note here that more infants in the intervention group adhered to their allocated treatment compared to infant in the control group (98 vs. 91%), which could have resulted in less treatment failure in the intervention arm.

Data expressed as number (percentage)
*Relapse percentage based on number of infants who did not fail in the first week (n=827 in injectable, and n=914 in oral amoxicillin group)

For the secondary outcome of death within 15 days of enrolment, the pooled analysis across the three sites showed the incidence to be <1% in both groups (with 4 deaths in each group). There was no significant difference between the groups (RD 0.0%; 95% CI: -0.5% to 0.5%) and the RD and its 95% CI were within the pre-specified equivalency margin of ±5%. The two regimens were therefore statistically equivalent with respect to this outcome.

For the secondary outcome of relapse during second week, pooled analysis across the three sites showed the incidence of relapse to be 2.0% in the 914 infants in intervention group; and 2.0% in the 827 control group, denominator being infants who did not fail treatment in the first week after enrolment. There was no significant difference between the groups (RD 0.2%; 95% CI         -1.2% to 1.6%). Since RD and its 95% CI lie within the pre-specified equivalency margin of ±5%, it indicates that the two regimens are statistically equivalent with respect to this outcome.

The authors also report unadjusted results of the intention-to-treat analysis, which corroborates the results as obtained on the per-protocol analysis. For the primary outcome of treatment failure by day 8 after enrollment, a pooled analysis across the three study sites showed the incidence of treatment failure to be 19.0% ((244 of 1170 infants) in the intervention, and 21.0% (222 of 1163 infants) in the control arm (RD -1.8%; 95% CI: -5.0% to 1.5%). This again indicates that the new treatment (oral amoxicillin) was equivalent to the control treatment (intramuscular penicillin and gentamicin).

There is some debate among experts with respect to whether per-protocol or intention-to-treat analysis is better employed in inferiority and/or equivalence studies. In this regard, it is now considered acceptable that a two-sided confidence interval approach be used, and results of both analysis be reported.^1,2  Equivalence and/or non-inferiority is considered to be established if results of the two analysis corroborate. The authors of the present paper have reported the results of both analyses and have used the two-sided confidence interval approach, which corroborate. Therefore, it is reasonable to accept that the new treatment is equivalent to the control treatment.

CONCLUSIONS

The authors conclude that oral amoxicillin is as effective as injectable therapy with gentamicin and procaine penicillin for fast breathing on an outpatient basis where referral is nor accepted or possible.

COMMENTARY

A possible serious bacterial infection (pSBI) in neonates and young infants (0-59 days old) is presently recognized through a set of well documented clinical signs. WHO currently recommends that young infants with any of these signs be considered to have pSBI and should be referred for hospitalization and treatment with injectable penicillin plus gentamicin for 7-10 days.^3,4  Some of such infants, however, might not be taken to the referral facility for many reasons, including parental inability and/or refusal.

Of this category of infants, those presenting with fast breathing alone (suggestive of pneumonia) might be at lower risk of serious adverse events compared to those with other signs.⁵  There is also some evidence (though from non-randomized studies), that oral antibiotics might be effective in treating neonates in community settings.⁶  However, no study had earlier compared oral with injectable antibiotics. The present study demonstrates that among young infants with fast breathing alone (and therefore with pSBI, though probably less severe), and whose parents do not accept or cannot access referral care, treatment with oral amoxicillin twice a day for 7 days is as effective as injectable penicillin-gentamicin once a day for 7 days. The study provides evidence that there is no difference in clinical efficacy (treatment failure), serious adverse events or deaths on comparing the oral and injectable treatments. The adherence to oral treatment regimen was in fact, better compared to the injectable therapy.

The present study was a large, community-based randomized trial from diverse settings in east, central and West Africa, and had 80% power to show equivalence between the oral and injectable treatments with an absolute margin of 5% failure rate. The authors used per-protocol analysis as primary analysis, which is desirable in an equivalence trial. It had high rates of follow up, with good adherence to treatment in both randomization groups. The study was implemented within existing health systems and used trained community health workers, which is a good reference model for implementation in similar settings.

The obvious benefits of oral treatment, which is as efficacious as injectable treatment, lie in its simplicity, feasibility, and better acceptability to the families, especially when they refuse referral. Oral administration as opposed to injections is also less painful for infants undergoing treatment. Potentially serious side effects of a drug like gentamicin (renal and oto-toxicity) may also be averted. These advantages will lead to higher rates of compliance and lesser untoward events. Oral treatment is cost-effective and can easily be made accessible in community settings, especially if resource constrained. It can potentially reduce not only the economic burden on families and health systems, but also prevent mortality due to less delay in care seeking and treatment in places where referral systems are not strong or easily accessible.

There are, however, few important points to consider with respect to the present study. First, is that the present study addresses only the subgroup of infants with fast breathing whose parents do not accept or cannot access referral care. When possible, health care settings should follow the guidelines by WHO and refer infants even with fast breathing alone to health facility for injectable antibiotics.Only the subset of infants whose parents do not accept or cannot access referral care should be considered for oral antibiotics, as suggested by the present study. Second, almost one-fifth of infants in both groups failed treatment with either regimen, primarily due to persisting fast breathing by day 4. Thus, possibility of the illness being viral in origin emerges as a significant concern to be addressed. Third, the definition of treatment failure in both groups remains subject to assessor bias in an open label trial, as also pointed out by the study authors. Except for death, all other criteria were subjective clinical signs. Important implementation considerations for the study are that it would be important to ensure that recognition of the infants with fast breathing alone is robust, i.e., health workers are sufficiently trained to be able to differentiate it from signs serious illness (other signs are not missed). Adequate supply of oral antibiotics should be ensured at the peripheral health level, and once oral treatment has been initiated, a regular follow up is maintained to detect any worsening (non-response) at the earliest. It should be ensured that parents are not falsely reassured by the oral treatment.

To conclude, the study suggests that oral amoxicillin is an equally effective alternative to injectable treatment in young infants who cannot access referral care in a low-resource community setting, even though the evidence may not be considered conclusive from a statistical point of view.

REFERENCES:

1. Piaggio G, Elbourne DR, Pocock SJ, Evans SJ, Altman DG. Reporting of noninferiority and equivalence randomized trials: extension of the CONSORT 2010 statement. Jama 2012;308:2594-604.
2. Schumi J, Wittes JT. Through the looking glass: understanding non-inferiority. Trials 2011;12:106.
3. World Health Organization. Integrated Management of Childhood Illness (IMCI), chart booklet. Geneva: WHO; 2014 (http://apps.who.int/iris/bitstream/10665/104772/16/9789241506823_Chartbook_eng.pdf,accessed 21 Sep 2015
4. World Health Organization. Pocket book of hospital care for children, second edition. Guidelines for the management of common childhood illnesses. Geneva: WHO; 2013 (http://www.who.int/maternal_child_adolescent/documents/child_hospital_care/en , accessed 21 Sep 2015).
5. Clinical signs that predict severe illness in children under age 2 months: a multicentre study. Lancet 2008;371:135-42.
6. Zaidi AK, Ganatra HA, Syed S, et al. Effect of case management on neonatal mortality due to sepsis and pneumonia. BMC public health 2011;11 Suppl 3:S13.