Page 80 - 80 guidelines for the treatment of malaria_opt

nd
Guidelines for the treatment of malaria – 2 edition

1. DESIgN LIMITATIONS: is the design of the trials flawed for a set outcome?
For hard outcomes, such as death, the panel considered the adequacy of the allocation
concealment.
For softer outcomes, such as vomiting, the panel also considered blinding.
Sensitivity analyses, excluding poor quality trials, enabled the panel to make judgements
about whether variations in methodological quality between trials constituted a limitation
or not, and, if so, if this was serious or very serious.

2. INCONSISTENCy: is the effect size similar across trials for a set outcome?
The panel considered the 95% confidence intervals for each trial reporting an outcome
and whether they overlapped or not.
2
The panel also considered the I value and viewed the forest plot to make a judgement.
3. INDIrECTNESS: is there sufficient evidence for your population and drugs
of interest for a set outcome?
The panel considered the trial populations, the population of interest and whether they
would expect differences in the performance of either the intervention or the control in
different settings.
If the panel only had partial information on the performance outcome in the population
of interest, then drugs were downgraded by 1 (serious indirectness).
If the panel only had limited information on the performance outcome in the population
of interest, then drugs were downgraded by 2 (very serious indirectness).

4. IMPrECISION: are there sufficient data and or clear relative effects for a set
outcome?
The panel agreed that the borders of appreciable benefit and harm were a relative risk
value of 0.75 < RR > 1.25.
If the 95% CI for the pooled estimate included appreciable benefit (or appreciable harm)
and no significant difference between the intervention and control group, then the panel
downgraded by 1 (serious imprecision).
If the 95% CI for the pooled estimate included appreciable benefit and appreciable harm
with the intervention, then the panel downgraded by 2 (very serious imprecision).

The GRADE quality rating for RCT evidence may, therefore, be “downgraded” by one,
two or three levels from high to moderate to low or very low quality, and it is rated on
the following four-point scale:

66