We carried out a study to determine the effect of zinc as adjunct therapy on clinical recovery, case fatality and adverse events in children with severe pneumonia in Uganda's national referral hospital. There are two key findings in this study: overall, zinc supplementation in these children significantly decreased case fatality, but did not reduce the time to normalization of the parameters for disease severity.
The children who received the placebo were three times more likely to die as compared to those who received zinc. The number needed to treat was 13. This means that we need to treat 13 children with zinc as adjunct therapy to avert one death. Since the cost of zinc for one child was 630 Ugandan shillings (0.3 USD), we needed 8,190 Ugandan shillings (or 4 USD) to prevent one death. On sub-group analysis the difference in case fatality attributable to the protective effect of zinc therapy was greater among HIV-infected children than among HIV uninfected children. This is not surprising since zinc is known to improve immune response. For example, zinc supplementation might increase phagocytosis  and zinc deficiency predisposes to apoptosis of T lymphocytes in HIV-infected patients . In fact, it has now been established that zinc deficiency compromises immunity through a number of mechanisms, such as T cell dysfunction and dysregulation of intracellular killing .
We found that 11.5% of the patients with positive blood cultures and receiving the placebo died, while none of the blood culture positive patients receiving zinc died. Case fatality among the blood culture-negative patients receiving the placebo was 18.8% compared to 6.7% of the blood culture-negative patients receiving zinc. Despite the small numbers, zinc reduced case fatality regardless of blood culture results. We did not find a similar study with which to compare our results. In a related study, Cole and Bose  used a CRP of > 40 mg/L as a proxy for bacterial pneumonia. They found that zinc therapy for children with severe pneumonia was associated with prolonged hospitalization. They, however, did not study the effect of zinc on case fatality.
Girls were two times more likely to die than boys. The reason for this is not clear. One study from India found pneumonia case fatality was about '50% higher in girls than in boys' . Indeed, excess case fatality rates among girls has been reported by other workers from India who speculated that female children are less likely to be taken for vaccination and medical treatment .
In our study, the serum zinc concentrations were low. They are lower than those previously reported in Ugandan children [6, 7]. The very low zinc levels could be explained by pre-existing zinc deficiency, redistribution of zinc to the liver in response to pro-inflammatory cytokines, and low albumin concentrations . Children with zinc deficiency have increased susceptibility to bacterial disease and are more likely to die . Animal studies have arrived at similar conclusions .
Studies on the effect of zinc supplementation on pneumonia have not reported the prevalence of HIV in the patients studied [10, 20]. None the less, the prevalence of HIV in children in our study was high. In a way our HIV-infected children with severe pneumonia suffered from a double disadvantage: HIV related immunodeficiency and that induced by zinc deficiency.
It appears that initial supplementation with zinc modulates this double disadvantage and significantly reduces case fatality. However, the 'modulation' does not appear sufficient to shorten recovery from the pneumonia itself in these HAART naïve, HIV infected children. It is also possible that zinc increases production of pro-inflammatory cytokines which would be expected to exacerbate the situation [25, 26]. However, this may not be sufficient to cancel out the benefit of the zinc on case fatality.
In our study, there was no significant difference in the measures of clinical improvement between children receiving the placebo and those receiving zinc, implying that the zinc might have no appreciable role in shortening the duration of severe pneumonia even though it significantly reduces case fatality. What is clear, though, is that we have un-earthed a very interesting, yet contradictory phenomenon, that seems to be related to HIV infection and severe zinc deficiency. It calls for further studies on the interaction among HIV, zinc and severe pneumonia.
Hitherto, zinc had only been conclusively linked to reduction of pneumonia incidence and case fatality when used for long term supplementation and not as adjunct therapy . In a study by Brooks et al. in Bangladesh, weekly zinc supplementation for 12 months reduced the incidence of pneumonia. Of the 14 deaths in the Bangladeshi placebo group, 10 were pneumonia-related while there was no pneumonia-related death in the zinc group .
Conditions associated with decreased immunity, such as malnutrition, partial immunization and HIV infection, were more likely to be associated with case fatality. For populations where children are deficient in zinc, severe prolonged infections deplete the zinc even further and decrease resistance to infection .
Time to normalization of the parameters for disease severity
The second finding in this study is that zinc supplementation as adjunct therapy for pneumonia did not reduce the time to normalization of the parameters for disease severity. Indeed, the median time for normalization of the respiratory rate, oxygen saturation and temperature in the zinc and placebo groups were not significantly different. This is consistent with results of two Indian and one Australian studies [9, 29, 30].
However, a Bangladesh study demonstrated a significant difference in time to normalization of respiratory rate (40 to 48 hours in the zinc and placebo groups) . Indeed, in a recent study from Nepal, it was shown that zinc adjunct therapy did not accelerate recovery from non-severe pneumonia  and called for larger studies, especially of severe pneumonia. So regarding the effect of zinc on normalization of parameters of disease severity in severe pneumonia, the jury is still out.
There are a few limitations to our study. The median age at death for children with pneumonia in Uganda is 16 months with an interquartile range of 8 to 23 months according to a study by Kallander . Therefore, we enrolled children aged 6 to 59 months into the study, which encompasses this age group. However, one would expect children below the age of six months to have the highest risk of case fatality from pneumonia. So excluding them might have affected our results.
We attempted to get blood culture on most of our patients but did this on only 184/352 (52.3%) of the children. Of these, 45/184 (24.5%) were culture positive. Secondly, we did not measure serum C-reactive protein (CRP), an acute phase protein. We followed up patients for only seven days, yet it is possible that a sizeable proportion had pneumonia for a longer period. In the sample size calculation for case fatality, we arbitrarily chose an HR of 1.49. As it turns out, this was over ambitious and represents an over-optimistic effect.
It might have been unrealistic to expect a reduction of the case fatality from 20% in the placebo group to 9.3% in the zinc group. In the calculation of the sample size, we assumed a case fatality of 20% in the placebo group. This was based on results of previous studies by our team  where we demonstrated a case fatality of 20% in children with severe pneumonia. Another one of our studies showed a case fatality of 15.5% among children with severe pneumonia in the same hospital .
Of course, these are high case fatalities but were not un-expected, especially since the children were ill with malnutrition and/or HIV.
In our study, 20/351 (4.5%) children died before achieving normalization of the respiratory rate, temperature and oxygen saturation (the numbers and proportions are much smaller when each of these outcomes is considered independently) and were thus lost to follow-up. This loss may have caused bias but, because the loss was minimal, it is unlikely to have been large enough to alter the results significantly. Indeed, these high case fatalities reflect the importance of pneumonia as a cause of child deaths in the community.
In a recent study, Kallander and others in the Iganga Demographic Study Site in eastern Uganda found that pneumonia accounted for 27% of child deaths . We are heartened that the overall case fatality in the current study was only 28/352 (8.0%), dropping to 4.0% in the zinc supplemented group.
Unlike a similar study, which followed up patients for two weeks , we followed up patients for a maximum of seven days based on our experience of severe pneumonia patients most of whom spend less than seven days [2, 31] in the hospital. The relatively short time of follow-up might have reduced our power to detect a difference in the duration of hospitalization.