We are sorry that we caused concern with our statement "Mortality was
lower in the children who took half the recommended dose of vitamin A
compared with the full dose at both six months (mortality rate ratio 0.69,
95% confidence interval 0.36 to 1.35) and nine months (0.62, 0.36 to 1.06)
of follow-up." Language conventions may differ between countries or
professions. We could have written that “mortality was lower, but not
statistically so..”, “mortality tended to be lower..”, or “there was no
statistically significant difference in mortality..”. However, this should
be evident from the confidence interval reported in the paper.
The study was a randomised study. We tested the distribution of a large
range of potential confounding factors such as maternal age and education,
socioeconomic indicators and nutritional status, and they were equally
distributed between groups (Table 1 in the paper). Hence, it is unlikely
that there are underlying differences in malaria, diet etc. between the
two randomisation groups at baseline which explain the results.
As stated in the paper, we chose to conduct an explorative study and
include all eligible children during a combined vitamin A and oral polio
vaccine campaign; we had no expectation of finding significant differences
in mortality. The reason we did find significant differences in mortality
was the surprising sex-differential effects of the two doses of vitamin A
resulting in significantly decreased mortality among girls who received
the low dose of vitamin A.
Based on our findings it is easy to calculate which number of children
should be enrolled should one wish to conduct a new study in order to test
the hypothesis that the half dose is better that the full dose. We have
pointed out that the finding of sex-differential effects and the resulting
significant beneficial effect of half versus full dose in girls was the
result of a post-hoc analysis. We welcome any attempt to confirm or refute
our findings in other settings.
Rapid Response:
The confidence limits of language
We are sorry that we caused concern with our statement "Mortality was
lower in the children who took half the recommended dose of vitamin A
compared with the full dose at both six months (mortality rate ratio 0.69,
95% confidence interval 0.36 to 1.35) and nine months (0.62, 0.36 to 1.06)
of follow-up." Language conventions may differ between countries or
professions. We could have written that “mortality was lower, but not
statistically so..”, “mortality tended to be lower..”, or “there was no
statistically significant difference in mortality..”. However, this should
be evident from the confidence interval reported in the paper.
The study was a randomised study. We tested the distribution of a large
range of potential confounding factors such as maternal age and education,
socioeconomic indicators and nutritional status, and they were equally
distributed between groups (Table 1 in the paper). Hence, it is unlikely
that there are underlying differences in malaria, diet etc. between the
two randomisation groups at baseline which explain the results.
As stated in the paper, we chose to conduct an explorative study and
include all eligible children during a combined vitamin A and oral polio
vaccine campaign; we had no expectation of finding significant differences
in mortality. The reason we did find significant differences in mortality
was the surprising sex-differential effects of the two doses of vitamin A
resulting in significantly decreased mortality among girls who received
the low dose of vitamin A.
Based on our findings it is easy to calculate which number of children
should be enrolled should one wish to conduct a new study in order to test
the hypothesis that the half dose is better that the full dose. We have
pointed out that the finding of sex-differential effects and the resulting
significant beneficial effect of half versus full dose in girls was the
result of a post-hoc analysis. We welcome any attempt to confirm or refute
our findings in other settings.
Competing interests:
None declared
Competing interests: No competing interests