Amoxicillin 3 vs 5 days for chest-indrawing pneumonia in Malawian children

BACKGROUND Evidence supporting duration of antibiotic treatment for children in low-resource African settings with chest-indrawing pneumonia is lacking. METHODS We conducted a double-blind, randomized controlled 2-arm, non-inferiority trial in Lilongwe, Malawi with follow-up for 14 days to determine whether treatment with 3 days of amoxicillin for chest-indrawing pneumonia is less effective than 5 days. HIV-uninfected children aged 2 to 59 months with chest-indrawing pneumonia were randomized to 3-or 5-day amoxicillin twice-daily. Primary endpoint was the proportion of children with treatment failure (TF) by Day 6 with a relative non-inferiority margin of 1.5 times the TF rate in the 5-day amoxicillin group. Planned secondary analyses included TF or relapse by Day 14. RESULTS Between March 29, 2016 and April 1, 2019, 3000 children were randomly assigned to 3-day (n=1497) or 5-day (n=1503) amoxicillin. Children receiving 3-day had a 5.9% (85/1442 with outcome data) TF rate by Day 6, within the non-inferiority margin of those receiving 5-day (5.2% (75/1456) TF rate), with an adjusted absolute difference of 0.75% and 95% confidence interval (CI) -0.92%,2.41%. Among children with known Day 14 outcome, 176/1411 (12.5%) receiving 3-day and 154/1429 (10.8%) receiving 5-day had TF by Day 6 or relapse by Day 14 (absolute difference 1.7%, 95%CI -0.7%,4.1%). There were no unexpected serious adverse events. CONCLUSIONS In HIV-uninfected African children, 3 days of amoxicillin treatment for chestindrawing pneumonia was non-inferior to 5 days. We recommend revisiting antibiotictreatment guidelines applicable to similar pediatric populations. ClinicalTrials.gov registration: NCT02760420.


INTRODUCTION
Approximately 920,000 children die before age 5 from pneumonia annually. 1 There is a critical need to provide greater access to appropriate and effective treatment. Treatment of bacterial pneumonia requires an effective antibiotic used in adequate doses for an appropriate duration. Determining optimal duration of antibiotic therapy is key to ensuring effective treatment while maximizing adherence and minimizing adverse drug effects, costs, and antimicrobial resistance.
A 5-day course of oral amoxicillin at least 40mg/kg/dose twice-daily (80mg/kg/day) is recommended by World Health Organization (WHO) as first-line treatment for chestindrawing pneumonia among immune-competent children <5 years old. 2,3 However, it is unclear whether a 5-day course of amoxicillin is necessary or if a shorter duration of treatment would be as effective. Based on studies of 3-versus 5-day oral antibiotics for fast-breathing pneumonia, WHO recommends a 3-day course of oral amoxicillin for treatment of fast-breathing pneumonia among immune-competent children <5 years old. 2,4-7 A Cochrane review found no qualifying randomized controlled trials comparing 2-3versus 5-day intravenous antibiotics for chest-indrawing or more severe pneumonia. 8 Few data exist to inform optimal duration of treatment for pneumonia, and no study has looked at 3-versus 5-day oral antibiotics for chest-indrawing pneumonia. 9,10 International and national pneumonia treatment guidelines rely on expert opinion and limited and weak evidence. 10,11 In light of the global threat of increasing antimicrobial resistance, evidencebased recommendations are needed for the optimal duration of antibiotic treatment for pneumonia. Given the paucity of African data, African-specific research in malaria-endemic settings is critical to establish optimal management of children with chest-indrawing pneumonia.

Study design
The primary objective of this prospective, double-blind, randomized controlled 2-arm, noninferiority trial was to determine whether treatment with 3-day amoxicillin in HIV-uninfected children 2-59 months of age with chest-indrawing pneumonia in a malaria-endemic region of Malawi is (null hypothesis) substantively less effective than 5-day amoxicillin. An innovative non-inferiority design was formulated based on the belief that 3-day amoxicillin could not be expected to be more beneficial than 5-day amoxicillin with respect to the primary outcome of treatment failure (TF) by Day 6, but might be (alternative hypothesis) only slightly worse than 5-day amoxicillin. 12  ASG and SM designed the study. TM, MP, CN, and AP gathered the data. RS, JH, and SM analyzed the data. All authors vouch for the data and analysis and decided to publish the paper.
ASG, EN and SM wrote the first draft of the paper. There were no confidentiality agreements between funder, sponsor, or any involved institutions.

Procedures
On Day 1, eligible children were randomized and enrolled, double-blinded, in a 1:1 ratio to receive either 3-day twice-daily amoxicillin dispersible tablets (DT) followed by 2-day twicedaily placebo DT (intervention) or 5-day twice-daily amoxicillin DT (control). High-dose oral amoxicillin was provided in 250mg DT in 2 divided doses based on age bands (500mg/day for children 2-11 months, 1000mg/day for 12-35 months, and 1,500mg/day for 36-59 months of age), current WHO-recommended therapy for HIV-uninfected children. 2 Study drugs were identical in appearance, smell, taste, dispersion activity and packaging. Randomization was stratified by age groups (2-11, 12-35 and 36-59 months) using blocks of size 2, 4 and 6. Other than unblinded biostatisticians, pharmacists, monitor, and data and safety monitoring board (DSMB) members, everyone else on the study team was blinded to each child's assigned treatment group.
Enrollment was conducted solely at KCH initially (phase 1), and then transitioned (September 20, 2016) to BDH (phase 2) after KCH introduced user fees which reduced patient volumes. BDH enrollees were transferred to KCH for additional evaluation and admission. To maximize safety, most enrollees were hospitalized for 2 days and discharged on Day 3 if no TF criteria (Table 1) were present.
Enrolled children were evaluated on Days 2 (while hospitalized), 4, 6, and 14 in clinic or home.
During follow-up, all children were assessed for TF or relapse and study drug adherence at all scheduled and unscheduled visits. Most TF or relapse cases were hospitalized and treated with intravenous antibiotics. Once on intravenous or other second-line antibiotics, the child was considered non-adherent to randomized treatment.

Outcomes
The primary endpoint was the proportion of children with Day 6 TF (Panel). Secondary endpoints included proportions of children with relapse (Days 7-14 among children without TF before or on Day 6), and with Day 6 TF or relapse by Day 14. Four of 6 prespecified subgroups are reported with respect to TF by age groups, malnutrition, malaria, and very fast breathing for age. Prespecified subgroups of low oxygen saturation (n=10) and wheeze (n=49) are not reported due to small numbers.
All adverse events were assessed and managed per KCH standard clinical practice, documented, and followed and treated until resolution or stabilization. All serious adverse events were reported to the study safety team for review within 24 hours.

Statistical analysis
A relative non-inferiority margin of 1.5 times the TF rate in the 5-day amoxicillin group was chosen based on an anticipated TF rate in the 5-day group of 8%. This non-inferiority margin, 50% higher TF rate in the 3-day compared to the 5-day group, was chosen after extensive discussions among the investigators and with external experts regarding what TF rate might be acceptable to clinicians for the 3-day compared to the 5-day group, considering the anticipated potential TF rate in the 5-day group and potential for enrollment into the study. Initially adjusting for 2 formal interim analyses (with O'Brien-Fleming boundary for early noninferiority 13 and Pocock boundary for early inferiority 14 ), enrolling 2,000 children (1,000 per group) provided 88.1% power if the TF rate was equal in both groups at 8%, and 64.8% power if the TF rate was 4% in both groups. A potential increase in sample size was considered during planning of the study in case the overall TF rate was much lower than the anticipated 8%. After the second formal interim analysis, it was clear that the overall TF rate was less than 6%. To maintain a power (with equal TF rates) of 80% or higher, the maximum sample size was increased to 3,000 children (1,500 per group), and a third formal interim analysis was performed after a little more than 2,000 children were enrolled. The decision to increase the sample size was made by blinded study investigators after consultation with the funding agency. With increase in maximum sample size (and assuming equal TF rates in each group), the study had 84.8% and 89.8% power for 5% and 6% TF rates, respectively. Power calculations took into account a drop-out rate of 5% and assumed a 1-sided alpha of 0.025 for a test of a difference in proportions. Primary analyses were performed based on the intent-to-treat principle of complete cases using linear regression adjusted for age groups, study phase and sex, and using robust standard errors based on the Huber-White sandwich estimator. 15,16 Justified because the sample size was sufficiently large, linear regression was used for this binary outcome to model differences in rates. 17 Estimates for treatment differences for prespecified subgroups are reported with individual 95% confidence intervals (CIs) without adjustment for multiple comparisons. No post-hoc subgroup analyses were performed. The independent DSMB considered formal stopping boundaries during their interim reviews, but decided not to follow them, but rather, treat them as guiding only. Thus, the primary analysis was not adjusted for interim monitoring. Sensitivity analyses were performed using multiple imputations and tipping point analyses. 18 For multiple imputations, a hot-deck approach (20 imputations) was used considering a match on at least 3 of the following 5 factors: age (2-11, 12-35, 36-50 months), sex, mother's education (none, primary, secondary/tertiary), number of children in the home (1, 2, 3, 4+), and number of amoxicillin doses taken (≤4, 5-7, 8-9, 10).
Analyses of secondary endpoints used robust standard errors unadjusted for interim analyses or other factors. Of 6 prespecified subgroup analyses, 4 are reported.
Prior to Day 4, both the 3-and 5-day groups were receiving amoxicillin, and as such, we would expect the TF rate prior to Day 4 to be the same. The TF rates prior to Day 4 in the 3-and 5-day groups were 2.3% (33/1442) and 2.3% (33/1456), respectively (post-hoc descriptive unadjusted). During Days 4 and 5, the 3-day group was receiving placebo whereas the 5-day group continued to receive amoxicillin. The TF rates for Days 4 through 6 in the 3-and 5-day groups were 3.6% (52/1442) and 2.9% (42/1456), respectively.
When considering both TF before or by Day 6 and relapse by Day 14, 176/1411 (12.5%) in the 3-day group and 154/1429 (10.8%) in the 5-day group met criteria (absolute difference of 1.7%, 95%CI -0.7%,4.1%). Additional secondary outcomes results are detailed in Table 3. The TF rate was generally consistent across prespecified subgroups defined by age groups, malnutrition, malaria, and very fast breathing for age. Most 95% CIs for the subgroups did not exclude a 1.5 non-inferiority margin and any adjustment for multiple comparisons would have resulted in all 95% CIs including the non-inferiority margin. The amount of missing primary outcome data was low (overall n=102, 3.4%; n=55 and n=47 in the 3-and 5-day groups respectively). Estimates derived from multiple imputations for missing outcome data were similar to the complete case analysis. When considering a tipping point analysis, we failed to conclude non-inferiority only if there were at least 3 additional children with TF among children in the 3-day group compared to children in the 5-day group among those who have missing data. If the same TF rates observed for the complete data applied to the missing data, the expected average difference is 1.2 individuals (55*5.9% -47*5.2% = 1.2). As such, we would have needed to observe a larger difference among the missing data (e.g., 3 out of 55, and 0 out of 47) in order to fail to conclude non-inferiority. If primary results would have been adjusted for sequential monitoring, the conclusion of non-inferiority remains the same.
The percent of children with at least 1 serious adverse event between enrollment and Day 14 was 9.8% in the 3-day group compared to 8.8% in the 5-day group (Table 4). There was 1 (0.1%) death due to pneumonia in the 3-day group, and 2 (0.1%) deaths, 1 due to pneumonia and 1 due to acute gastroenteritis, in the 5-day group.
Caregiver-reported adherence was high with 91.6% reporting adherence with all doses in the 3-day group and 91.8% reporting adherence with all doses in the 5-day group.

DISCUSSION
We evaluated 3-versus 5-day oral amoxicillin treatment among 3000 HIV-uninfected children aged 2-59 months presenting with WHO-defined chest-indrawing pneumonia in a malariaendemic region of Malawi. Our results demonstrated that those children who received 3-day amoxicillin had a non-inferior TF rate on or before Day 6 compared to those who received 5-day amoxicillin. By Day 14, non-inferiority appeared to continue.
This study suggests that 3-day amoxicillin is not substantively worse than 5-day amoxicillin for treatment of chest-indrawing pneumonia among HIV-uninfected children. Keeping in mind both individual and health system benefits of a shorter course of antibiotic therapy, and that WHO already recommends 3-day amoxicillin for treatment of fast-breathing pneumonia 2,5-7 it appears that 3-day amoxicillin for children with chest-indrawing pneumonia might be sufficient.
Currently, WHO recommends a 5-day course of twice-daily high-dose oral amoxicillin to treat chest-indrawing in a child with cough or difficulty breathing. 2,19 However, the findings of this study may allow for harmonization and simplification of treatment courses for both fastbreathing and chest-indrawing pneumonia to be 3 days among HIV-uninfected children. A study from Pakistan found that in cases of chest-indrawing pneumonia without underlying complications, home treatment with a short-course of high-dose oral amoxicillin was preferable to parenteral treatment because of the associated reduction in referral, admission, and treatment costs. 19 Home treatment of chest-indrawing pneumonia with oral amoxicillin is effective across communities and geographic regions. 20-22 In contrast to low-resource settings, in high-resource settings, criteria for diagnosing pneumonia often require chest radiographic confirmation, especially in hospitalized children. 23 Yet, little evidence exists to dictate treatment duration. 11 Of note, in a very small study from Israel, a 3-day course of oral high-dose amoxicillin was associated with a high TF rate of 40% (4/10) among children with radiographconfirmed pneumonia. 24 Poor adherence to antibiotics has been associated with TF in WHO-defined clinical pneumonia. 25,26 Improving adherence with shorter course treatment could improve outcomes in children with chest-indrawing pneumonia while also minimizing adverse drug effects, costs, and the emergence of antimicrobial resistance. 7,25,26 Limitations Limitations in our study included strict inclusion and exclusion criteria, absence of laboratory or radiology testing, and close monitoring and follow-up, which limits the generalizability of our results to routine programmatic care settings. Notably, severe disease was excluded which limits applicability. Pneumonia is frequently considered a single entity, rather than a clinical syndrome encompassing several underlying factors. This makes interpretation of results challenging. Without etiological information, we could only note the effect of the intervention on the clinical syndrome of pneumonia, which is an approach consistent with non-trial conditions relevant to pediatric care in low-resource settings.
Follow-up care and monitoring of enrolled children generally exceeded local standards of care and thus, the TF rate may have been influenced by both the high quality of care provided and by the high awareness and vigilance for identifying TF. It may be that those identified as failing treatment would have recovered without a longer course of antibiotics had we taken a watchful waiting approach and not intervened with antibiotic treatment. However, opportunities for follow-up and access to care are often issues in low-resource settings. In addition, treatment approaches vary widely between countries and regions. Routine pediatric HIV testing included in this study protocol, while recommended, is not rigorously implemented during routine care in low-resource HIV-endemic settings. 27 In areas where pneumococcal immunization coverage is lower, HIV endemicity is high, or where severe acute malnutrition or other predisposing conditions for bacterial disease is common, it may be reasonable to expect a higher TF rate among those not treated with a longer course of antibiotics. As such, our results might not be generalizable across different regions, settings, or non-trial conditions. Specifically, the percentage of TF or relapse observed in this study might be underestimating true TF and relapse rates experienced during non-trial conditions.

CONCLUSIONS
Despite pneumonia being a common and deadly illness, optimal duration of antibiotic treatment for community-acquired pediatric pneumonia has not yet been established. In this population in Malawi, 3-day was non-inferior to 5-day amoxicillin treatment among HIV-   1 Children may be ineligible for more than one reason.
2 Missing follow-up data may be due to missed visits or visits occurring outside visit windows.
3 Missing follow-up data n's do not add up because some children had missing follow-up data for either Day 2 or Day 4 or both, but had outcome data available for Day 6.   Occurring any time after study drug is administered to child up to 14 days after enrollment.
2 Children may have more than 1 serious and/or non-serious adverse event. 3 37 occurred on or prior to Day 6 and were treatment failures while the remaining occurred after Day 6 and thus were considered relapses. 4 The chest radiograph-confirmed pneumonia serious adverse events did not demonstrate fast breathing, chest indrawing, or any danger signs; however, pneumonia was diagnosed through positive chest radiographs.