Study design and setting
This was a randomized double blind placebo-controlled clinical trial. The study was conducted in one pediatric ward at Mulago, Uganda's teaching and national referral hospital. It receives patients from the capital city, Kampala, and referrals from the rest of the country.
The primary outcome measures were the time taken for normalization of respiratory rate, time taken for normalization of temperature and time taken for oxygen saturation to normalize (92% or more), while breathing room air. The oxygen saturation cut off was set at 92% according to a study in the same hospital by Tumwesigye [11
]. The secondary outcome measure was
the proportion of children who died in each arm of the study.
Adverse effects due to zinc adjunct therapy, such as diarrhea and vomiting, were recorded. All children were observed for adverse events for one hour after zinc administration. The caretakers were subsequently asked (daily) about adverse events, such as vomiting.
Severe pneumonia was defined as the presence of cough or difficult breathing with fast breathing (tachypnoea) and chest in-drawing. Tachypnoea was defined as a respiratory rate of 50 or more breaths per minute in a child 6 to 12 months of age. For those aged 12 or more months it was defined as 40 or more breaths per minute. In any case, the child had to be calm. Normalization of oxygen saturation was regarded as oxygen saturation above 92% while breathing room air for more than 15 minutes, and maintaining above 92% on subsequent readings. Children, who were not hypoxic on enrolment, were assigned 'zero' as the time to normalization of oxygen saturation.
The respiratory rate was regarded as normal if it was consistently below 50 breaths per minute in infants and 40 breaths per minute in children above 12 months of age for more than 24 hours. Temperature was recorded as normal if below 37.5°C and maintained below this level for more than 24 hours. Two investigators agreed on the presence of lower chest retractions or chest in-drawing. If they disagreed, a third investigator was called to observe and decide whether chest in-drawing was present. When two of the three agreed, then the decision was that chest in-drawing was present.
For time to normalization of the respiratory rate, a sample size of 313 was calculated. We assumed 90% power, a two-sided 0.05 type I error and a hazard ratio of 1.49. We assumed a median time to normalization of respiratory rate of 73 hours (based on an Indian study) [9
]. Allowing for 10% attrition, the minimum total sample size was 344.
For time to normalization of the temperature, a sample size of 312 was calculated. We assumed 90% power, a two-sided 0.05 type I error and hazard ratio of 1.49; and a median time to normalization of the temperature of 72.3 hours (based on the same Indian study) [9
]. Allowing for 10% attrition the minimum sample size was 343. The sample size for time to normalization of oxygen saturation was similar to that of time to normalization of the temperature. For case fatality as a secondary outcome, the minimum sample size of 378 (189 in each group) was based on 20% case fatality in children admitted with severe pneumonia to Mulago Hospital [12
]. We assumed that zinc supplementation would reduce the case fatality from 20% to 9.3% with 80% power and a type 1 error of 0.05. Allowing for 10% attrition, the minimum total sample size would be 416 (208 in each group).
Participants and recruitment
We enrolled children, aged 6 to 59 months, who were admitted to the Mulago Hospital pediatric emergency ward with severe pneumonia, and whose parents or caretakers gave informed written consent. The children were followed up for one week. The definition of severe pneumonia was based on WHO criteria [13
Children with known heart conditions, obstructive air way disease and those on medication with zinc supplements were excluded. Children with a current episode of wheeze were also excluded, irrespective of previous wheezing attacks. The exclusions were based on caretaker history, physical examination findings and previous medical records. Heart disease was defined as the presence of tachypnoea associated with a grade 3 or 4 heart murmur or medical records confirming heart disease. Obstructive airway disease was defined as recurrent wheezing attacks associated with tachypnoea or chest in-drawing confirmed on two or more previous examinations by medical records. Current treatment with zinc was confirmed by reviewing medical records.
For all enrolled children, two of us (MGS and CKM) took a detailed history and clinical examination.
A pulse-oximeter (model Ohmeda Biox 3700 Ohmeda (GE Healthcare), Pollards Wood Chalfont St. Giles, Bucks HP8 4SP United Kingdom.) with an in-built automatic calibration was used to measure oxygen saturation. Blood smears for malarial parasites and HIV serostatus of the children were also assessed. Additional consent was sought for HIV testing. Only 311/352 or (88.4%) consented to HIV testing.
Randomization, blinding and treatment allocation
Eligible children were enrolled in serial order and randomized to receive either oral zinc or placebo.
The zinc and placebo tablets were packaged and labeled with study identification numbers by a pharmacist not directly involved in the care of the patients. Macleod Pharmaceutical Ltd, Mumbai, India, manufactured the zinc and placebo tablets. Both zinc and placebo tablets were similar in color, shape and size. The investigators, study nurses and parents/caretakers were masked to the treatment and only the pharmacist had access to the study code. On receiving an eligible child, the study physicians informed one of the investigators (pediatrician) who in turn informed the pharmacist who had access to the randomization code. The pharmacist then prepared the dose of zinc/placebo according to the age of the participant. He packaged the zinc or placebo in a sealed opaque envelope, which he labeled with the study identification number.
The envelopes were then delivered to one of the investigators' offices from where the nurse took them to the ward. The nurse gave treatment according to the envelopes. Zinc/placebo administration was directly supervised by the nurse within 30 minutes after giving antibiotics.
Children less than 12 months received 10 mg of zinc (as zinc gluconate) or the placebo while those aged 12 months or more received 20 mg of zinc or placebo once daily for seven days as we anticipated children with pneumonia to have recovered [14
] by Day 7. Dosing was closely related to the recommended dietary allowance (RDA) of zinc [15
]. The water soluble tablets were crushed into powder and placed into the mouth of those children who could not take them directly. If vomiting occurred within half an hour, the dose was repeated.
Laboratory and radiological investigations
Blood for full blood count and HIV serology was taken by venipuncture from the cubital fossa or dorsum of the hand in 5 ml EDTA vacutainer tubes (Becton Dickinson, Franklin Lakes, NJ, USA). HIV sero-status of study patients below 18 months of age was confirmed using DNA PCR. Blood for serum zinc was collected using trace element-free vacutainers (Becton Dickinson) and storage tubes.
Just before the zinc assays were carried out, we diluted the serum samples five-fold using de-ionized water according to the manufacturer of the test kit [16
]. Colorimetric zinc determination was done using a Quantichrom™ zinc assay kit (BioAssay Systems · 3191 Corporate Place, Hayward, CA 94545, USA). In this assay, serum zinc reacts with a chromogen to give a colored complex whose color intensity is directly proportional to the zinc concentration read at 425 nm. Internal control was used, using pooled serum and every batch produced a value of 45.2 ± 4.5 μg/dL. Unfortunately, no external controls were available.
We transferred 50 μL of water, zinc standard (10 μM), sample and sample Blank (50 μL sample + 2 μL EDTA) into wells of a clear bottom 96-well plate according to the manufacturer's manual [16
]. We then added 200 μL working reagent, tapping the plate gently in order to mix. This was incubated for 30 minutes at room temperature and the optical density was read at 420 to 426 nm (peak absorbance at 425 nm).
The concentration of zinc was then calculated using the following equation:
OD is the optical density at 425 nm. Where n is the dilution factor (n = 5 for serum or plasma samples). The conversion was ''1 μM zinc equals 6.5 μg/dL.''
Treatment of participants
Children received either intravenous chloramphenicol[Ciron Drugs &Pharceuticals Ltd, MIDC, Tarapur Boosar Thane India] 25 mg/kg every six hours or ceftriaxone [Hoffmann Roche Ltd Basel, Switzerland] 75 mg/kg once daily for seven days. Second line antibiotics: gentamicin (2.5 mg/kg every 12 h) and cloxacilin [Flamingo Pharmaceuticals Ltd, Indus, India]. (50 mg/kg every 6 h) were only given if there was no response after two days of treatment. On enrolment, 50/352 (14.2%) children were unable to feed orally and fed by naso-gastric tube on milk, porridge and locally available soups. Children with a peripheral oxygen saturation of less than 92% were given oxygen by nasal prongs or catheter.
For patients on oxygen, we disconnected the oxygen supply for five minutes before the oxygen saturation was measured. We counted respirations using a timer for 60 seconds when the chest and abdomen were exposed according to WHO Integrated Management of Childhood Illness (IMCI) guidelines [13
]. The count was repeated and the average was taken as the true value if the difference between the two counts was not more than five breaths per minute. If the difference was more than five breaths per minute, a third count was done and the average of the closest two respiratory rates was taken as the true value.
The study team was comprised of three doctors and two nurses (clinic team) and two laboratory personnel. The clinic team had prior experience in assessing and managing children with severe pneumonia. They were trained on the study procedures and assessment of clinical signs before commencement of the study to ensure reproducibility and consistency.
A three-member independent team was comprised of a pediatrician, an epidemiologist and a pharmacist, who regularly and closely monitored the data. Interim analysis was done when we had accrued half the sample size. All children received vitamin A as part of treatment for pneumonia as recommended by WHO [13
]. Fluid therapy was given according to hydration status.
All children were followed up until discharge, death or a maximum of seven days, whichever came first. Patients were reviewed every six hours for the first 48 hours and then every 12 hours. The clinical review included the respiratory rate, presence of chest in-drawing, oxygen saturation, auscultation findings, fever, cyanosis and mental state. All children were monitored for vomiting and diarrhea using information from the caretaker or by observation. Criteria for discharge included: resolution of fever, persistently normal respiratory rate, and oxygen saturation for more than 48 hours and disappearance of chest in-drawing.
Data management and analysis
• All the data were double-entered into the EPI-INFO 6.04 computer software package (Center for Disease Control and Prevention, Atlanta, Georgia, USAand analyzed using SPSS(SPSS Inc., Chicago Ill.)
= and Stata version 9.2 (StataCorp LP, College Station, TX, USA). Analysis was by intention to treat. The Student's t-test was used for comparing normally distributed continuous data, and the Mann Whitney test for skewed continuous data. The Chi square or Fisher's exact test (where appropriate) was used to assess differences in categorical data between the zinc and placebo arms. Each child's observation started on enrolment and ended on Day 7 or earlier, at the time of death for those who died. Participants who died before normalization of the primary study outcome, such as the respiratory rate, contributed person time of observation. Those who died after the normalization of the primary study outcome, such as the respiratory rate, were not censored for that outcome.
We used risk ratios (RR) to measure the association between case fatality and the intervention. We used multivariable generalized linear model (GLM) regression analysis with a log link to estimate the adjusted RR of the intervention on case fatality.
Initially, we estimated the effect of wasting (WHZ), percentage oxygen saturation, HIV status, age, relationship with the caretaker and blood culture results on case fatality in crude analyses. However, only variables that were associated with case fatality with a P-value < 0.25 or whose addition resulted in a 10% difference in the parameter estimates, were included in the final multivariable model.
A cross product term (for HIV status and zinc) was included in one of the multivariable models to test for interaction on the multiplicative scale. We examined interaction on the additive scale by estimating homogeneity of stratum specific risk differences. Cox proportional hazards models were used to compare time to normalization of respiratory rate, temperature and oxygen saturation in the two arms.
Permission to conduct the study was obtained from Makerere University Faculty of Medicine and Mulago Hospital Ethics committees.