Small sample size
Smaller numbers were obtained for drugs used only at hospital level; nevertheless data were sufficient to enable comment on facility results for all of these drugs. Longitudinal series data were insufficient for amoxicillin capsules, doxycycline capsules, phenoxymethylpenicillin tablets and epinephrine injection.
Age of samples and pipeline times
A fast rate of consumption and turnover was seen for ampicillin capsules, phenoxymethylpenicillin tablets, tetracycline capsules and procaine benzyl-penicillin injection, with fewer than 10% of facility samples in the second part of their shelf-life. This implied the possibility of undetected problems. However, this is considered to be representative of the real situation (with the possible exception of procaine benzylpenicillin injection) where longer storage periods, particularly at facility level, may be expected. A stockout of epinephrine injection at GMS resulted in an unrepresentative picture of facility samples with mostly expired samples.
Measurement error: content of active ingredient
It is difficult to quantify measurement error. The examination of three measurement results for each sample proved a high level of precision. The duplicate assays obtained through pilot testing a few reference samples for initial quality and subsequent re-testing of the same batch as paired samples provided some indication of reproducibility. The gain in active ingredient observed in GMS/facility sample pairs is an obvious artifact which provides an indication of measurement error. A comparison of the results obtained for a small number of samples from an external laboratory provides a measure of reproducibility; this exercise is described in Annex 2, A difference of less than 5% in assay results was not considered significant when comparing individual samples.
Validation of results
The validation of assay results was carried out by the Chemical Laboratory of the Department of Health in Copenhagen, Denmark. The assay value of one CMS sample of retinol was very different from the others (90.6%), but a re-test in Denmark produced a result (63.4%) which was much more consistent with the other results. The mean obtained in the Zimbabwe laboratory was 73.0% (n=5) while the mean obtained from the external laboratory was 65.9% (n=3). In addition, five facility samples were also validated in Denmark. The mean result for these five re-tested samples was 66.6%, The conclusion of ZRDCL (all failures) was confirmed.
Samples of most other drugs were also sent to Denmark for validation. The results were as follows: acetylsalicylic acid tablets ±2%, amoxicilline ±8%, ferrous sulfate tablets ±6%, penicillin tablets ±4%, tetracycline tablets ±6%, epinephrine injection ±5%, ampicillin injection ±2%, benzylpenicillin injection ±4%, ergometrine injection ±15%, and procaine penicillin injection ±19%. In all cases the conclusion of pass/fail was confirmed with a mean discrepancy of 4.3%. In general the validation results were lower than those from Zimbabwe. This implies that the results of this study underestimate rather than overestimate the problem.