Ferrous sulfate tablets
Five (22%) of the 23 GMS samples failed. However, three failures were above and two were below specifications (95-105%, BP), The lowest assay value found was 93.7% and the highest 110.0%. These values do not represent a serious initial quality problem. 31 (41%) of the 76 facility samples failed. However, 29 of these were above the upper limit while only two (3%) were below (Figure 2). There was a difference between manufacturers: all 29 high failures were produced by manufacturer A (representing 54% of the samples of this producer), while both low failures were produced by manufacturer B.
Slow disintegration is known to reduce the bioavailability of ferrous sulfate. Some samples failed to meet the specified times in the disintegration test, but an insufficient number were tested to enable conclusions to be made.
Figure 2: Ferrous sulfate tablets
The mean assay was higher in facility samples (103.5%) than in the GMS sample group (100.6%), The longitudinal series samples showed no loss of potency (mean difference +1.5%, 95% CL -0.7 to 3.6%, n.s.). There was no indication of decreased potency in samples close to the end of their expiry or beyond expiry. These results indicate that the drug is stable under the worst case conditions in Zimbabwe.
The poor compliance (mostly high failures) of ferrous sulfate tablets with specifications probably reflects the rather narrow assay limits (95-105%, BP). Non-compliance with pharmacopoeal specifications cannot be accepted; yet in view of the wide therapeutic index and. the absence of any evidence on loss of active ingredient, these results do not indicate a medically significant problem. The summary results presented in Table 2 indicate only the "low failures". Two (9%) GMS samples and two (3%) facility samples had assay values below 95% with a minimum value of 91.9%.
All five GMS samples were below the lower limit of assay specifications (95-120%, USP). The assay results were low (59.7, 63.4, 68.5, 71.6 and 74.6%) with a mean value of 67.6%, The number of GMS samples is small, but represent all batches received in the period of one year. 46 (90%) of the 51 facility samples were below the lower limit of specifications (Figure 3). The mean assay was 73.5% (95% CL 69.4-77.5%). A wide range was seen with a minimum value 40.2% and a maximum value of 116.7%. Although there is a wide margin in the therapeutic dose of retinol, the frequent low potency represents a significant under-dosing with appreciable risk of clinical failure.
Figure 3: Retinol
20% of facility samples had expended more than half of their three year shelf-life, but no samples were obtained in the last year of shelf-life. No clear association was found between assay and age of facility sample (Figure 3). The 30 GMS/facility sample pairs showed large gains as well as large losses; the mean difference was +8.6% (95% CL -0.8 to 18.1%) for a mean interval of 6.9 months. The high level of measurement error implied by these results does not enable any conclusions regarding stability.
The mean assay results by individual manufacturer are shown in Annex 1.8. Most facilities had product B and only eight samples were obtained from manufacturer A. Similar low potency and failure rates at GMS and facility level were found for both manufacturers. The manufacturers' batch certificates indicate the required level of active ingredient, with one exception (92.8%). GMS samples were taken early in their shelf-life (mean 5.9 months), GMS assay results indicate an apparent great loss of potency (mean GMS assay is 31.2% lower than mean manufacturers' assay). However, the results from facility samples do not show an equivalent rapid loss of potency over this time interval. Some discrepancy is expected to arise when results are obtained from different laboratories; however this large difference is outside the norm.
A pilot study in Sudan 4,5 of two retinol samples taken from a district hospital showed a 12.5% loss of original content when compared with the manufacturer's assay value after an interval of 18 months. A study before and after international shipment conducted by WHO/UNICEF1,2 reported a mean loss of potency of 1.5% (the mean of eight samples) when compared with a 120.8% content in a control kit. This difference was statistically significant but had no medical implications, particularly since the product contained an extra 20% of the active ingredient.
In Zimbabwe, the initial quality of retinol was poor for both manufacturers. The low potency found at health facilities compromises the efficacy of therapy. The assay results are not sufficiently reliable to make conclusions on its stability. Evidence of moderate instability found in other studies could therefore not be confirmed.
Epinephrine injection (adrenaline)
As there was only one batch at GMS, it was not possible to draw a conclusion on its initial quality. The sample of this batch was within specifications. At facility level, 14 (58%) of the 24 samples were expired. Due to a prolonged stockout at GMS, facilities continued to use this expired stock. This group of samples included products from five manufacturers with varying shelf-life. New stocks were received at the end of the study, limiting the number of longitudinal data.
In Annex 1.9, the results of assays on expired samples are reported separately. Figure 4 illustrates all facility assay results including the expired samples. Two (20%) of the 10 "in date" facility samples failed, with one failure below the limit (87%) and one above (124%). Expired samples showed a similar pattern; four (29%) of 14 samples failed; two failures were below the lower limit (85.4, 85.8%) and two above the upper limit (117.0, 134.6%).
The samples did not show consistent low values toward the end of shelf-life. The mean assay for 14 expired samples (102.3%) was not lower than the mean for 10 non-expired samples (102.5%) with 24 months difference in mean age between them. Similarly, no loss of potency was found in four GMS/facility sample pairs, but the mean interval was very short (2.1 months).
Figure 4: Epinephrine injection
A pilot study in Sudan 4,5 tested a single batch of epinephrine injection 18 months after manufacture. A control sample stored by the supplier in the Netherlands showed a low potency (88.6%). A loss of 14% of the control sample value was found in three samples taken from district hospitals at the same time. The conclusion was that epinephrine injection is unstable under tropical conditions.
Another study investigating the effects of the extreme conditions during shipment from Sweden and in-country distribution within Sudan6 reported significant degradation of epinephrine in a preparation of lignocaine and epinephrine injection which was of proven high initial quality. The content of epinephrine fell below the USP limit after 12 months storage in regional stores, and fell to almost 0% after two years storage time. A wide variation was noted (16.9%-60.8%) in the content of vials of the same batch taken from different packs. Some variation was even noticed in vials from the same pack. Two degradation products were apparent at the same time as sodium metabisulfate (an anti-oxidant agent) disappeared from the chromatograph.
In Zimbabwe, the quality of epinephrine injection was found to be variable with similar rates of low And high failures at facilities. It was not possible to distinguish whether low potency was due to poor initial quality or to instability. The minimum and maximum values for active ingredient (85.4 134.6) caused some concern regarding possible clinical consequences. The strong evidence of instability from other studies could not be confirmed. The extent of loss of active ingredient varied according to the setting and formulation of epinephrine.
Two (20%) of 10 GMS samples (age 1-18 months) were below the lower limit (86.8% and 90.9%). One of these failures occurred at two months, suggesting an initial quality problem. One batch was received in GMS with more than half its shelf-life expended. Five (20%) of 24 facility samples were below the lower limit, a failure rate equivalent to that at GMS (Figure 5).
The GMS and field sample groups showed mean assay values (96.7% and 96.2%), but the facility sample group had a higher mean age (14.6 versus 7.7 months). The 13 GMS/facility pairs showed a mean loss (-1.1%), but this was not statistically significant. There was no sign of decreased potency in samples close to the end of shelf-life, and two expired samples were within the narrow specification limits. One sample awaiting destruction at GMS was tested at age 44 months (20 months after expiry) and 103.6% of stated content was found. There are therefore no signs of instability.
A pilot study in Sudan4,5 showed some sign of instability in three samples obtained from district hospitals. However, the potency loss was less than 2% when compared with a control sample stored by the supplier in the Netherlands after an interval of 18 months. Degradation products in these samples increased from 5.1% (reference) to 6.5% (field samples), close to the BP limit, but considered acceptable.
Figure 5: Ampicillin injection
In a study before and after international shipment conducted by WHO/UNICEF1,2, all samples failed to meet BP specifications. However, they found no further loss of potency during transport (mean of eight samples) when compared with the content of a control kit (84.1%.). The study therefore highlighted a problem of poor initial quality. Another aspect of quality reported was the overfilling of vials by 13.8% to compensate for the low potency of active ingredient. Despite this, the total content for four out of eight samples was below BP limits (95%), but all samples were within USP limits (90%).
In Zimbabwe, ampicillin injection showed a problem of low initial potency which has potential clinical consequences. This finding was also found in other studies, A minor stability problem found in Sudan was not confirmed in Zimbabwe and is unlikely to have therapeutic significance as long as the initial quality of the product is assured.
Samples were obtained from three manufacturers. Manufacturer C had completed delivery on the tender of the previous year and only one batch (age 14.5 months) remained in stock at GMS. Two other manufacturers (A and B) were accepted on new tenders. The mean assay for all GMS samples was 82.4%, but there were obvious differences between manufacturers. The single sample from manufacturer C was within specifications even though it had expended half of its two year shelf-life, 62% of A samples failed and all B samples failed. All failures were below the lower limit of specifications (90-110, USP). Two batches from Manufacturer B were received with 50% of shelf-life already expended. The lowest value obtained was 58.4% which was in a sample at the beginning of its shelf-life (age 1.2 months), see Figure 6.
Figure 6: Ergometrine injection (GMS samples, three manufacturers)
Batch certificates were obtained from Manufacturer A: the mean for 21 batches was 104.7% (95% CL 103.5-105.8%). The mean assay for these 21 batches when sampled at GMS was 85.2% (95% CL 80.8-89.5%) with mean age 1.7 months. It is difficult to account for this 20% loss of active ingredient since transport and packaging arrangements were found to be satisfactory and complied with realistic guidelines in this setting. There is clearly a serious problem of initial low potency.
Quality at facilities
12 (15%) of all facility samples were expired. These expired samples were excluded from the summary results presented. 48 (72%) of 67 non-expired facility samples failed, which is higher than at GMS level. The mean assay for all unexpired facility samples was 73.6% with mean age 13.5 months. The minimum assay result of 20.5% obtained in one sample represented a complete loss of therapeutic value. However there were also circumstances in which the potency was well maintained: 19 (28%) samples passed, all of which were in the second year of shelf-life.
No clear association is seen between assay and age in the facility sample group as a whole. However, it should be realized that this was a very heterogenous group with different manufacturers and different batches, see Figure 7.
Observations on stock management and expiration at the facility level
Ergometrine appeared to be the drug most affected by inventory control inefficiencies and logistic difficulties. Demand for this item was also more variable than for other drugs: the number of maternity cases was changeable and use depended on the presence of nurses with maternity training. Being the exceptional drug item requiring refrigeration, it was more often without a stock card and irregularly ordered. The mean facility storage time was 5.6 months but varied greatly. However, a high awareness was observed in relation to the storage requirements with refrigeration being the rule, and removal from outer packaging was only observed in a single instance.
Figure 7: Ergometrine injection, facility samples (three manufacturers)
Four samples were found (apparently still in use) approximately two years over the labelled expiry. A total of 12 expired samples showed a mean assay of 65% (53-77%) at mean age 32.5 months. This mean is 9% lower than the mean for unexpired facility samples. However, two expired samples were within specification limits which reflected a similar variability to that seen in unexpired facility samples.
The longitudinal series data indicated a substantial difference between GMS/facility potency. The mean loss of -17.1% for a mean sample interval of 4.8 months was strongly significant (p<0.0005). Figure 8 illustrates the loss/gain for 50 individual sample pairs according to the sampling interval, Ergometrine injection is very unstable and there is a correlation with sampling interval (correlation test: p<0.005, r=0.399, N=52).
Assessment of outcome by manufacturer
As already described, initial quality was poor for two manufacturers, but the only batch from Manufacturer C passed. This difference was confirmed at facility level: of the 19 samples that passed, 17 were from Manufacturer C. Figures 9, 10 and 11 show the longitudinal series data for each manufacturer. The figures illustrate the data separately for each manufacturer, showing both GMS assay and facility assay (at corresponding age) in the same chart. The decrease in potency was visually apparent for each manufacturer. The mean loss in GMS/facility sample pairs was similar for all three manufacturers. The loss of potency is significant for each of the manufacturers individually.
Figure 8: Ergometrine injection
Figure 9: Ergometrine injection, GMS and facility samples, per manufacturer
Figure 10: Ergometrine injection, GMS and facility samples, per manufacturer
Figure 11: Ergometrine injection, GMS and facility samples, per manufacturer
Other factors associated with low potency
The wide range of assay values suggested that a number of factors may have accounted for the decreased potency of ergometrine injection at facility level. 10 (15%) of the 67 facility samples were found stored at room temperature (unrefrigerated) at time of sample collection. There was a marginally significant higher relative risk of failure in this group compared with the refrigerated samples. There was no proven association with any other single risk factors. However, the manufacturer and age of the sample are major variables (and determinants of outcome at facility level) which may mask the effects of other factors.
Experience from other country studies
A report7 combining data from a series of field studies that collected 49 unexpired ergometrine injection samples from 34 different health facilities in six developing countries (Bangladesh, Democratic Yemen, Gambia, Malawi, Sudan and Zimbabwe) showed only 15 (31%) within BP/USP assay limits. 12 (24%) had levels of active ingredient between 80-89%, and as many as 15 (31%) samples had less than 60% potency. Defective products were reported in each country, representing six manufacturers and indicating a widespread problem at the level of the end-user. The problem is not confined to developing countries with one of four locally manufactured products taken from different large hospital pharmacies in the Netherlands failing to comply with USP/BP standards.
A pilot study in Sudan 4,5 investigated one batch manufactured in France with a two year shelf-life and concluded that ergometrine is unstable under tropical conditions. Three field samples collected from district hospitals showed a consistent low potency with a mean of 45%. A control sample stored by the supplier for a period of 18 months in the Netherlands measured 62%, This control sample had been stored at room temperature.
Another study investigating the effects of the extreme conditions during shipment from Sweden and in-country distribution within Sudan 6 showed no significant degradation in a product of known high initial quality between Europe and Port Sudan. The content dropped below 90% after overland transport from the port to central stores in Khartoum. The potency declined to 58% when sampled from central and regional stores after 25 months. A colour change was noted after less than 12 months. The content of degradation products increased to 20%, 10 times the USP maximum limit.
A study before and after shipment conducted by WHO/UNICEF1,2 reported a mean loss of potency of 5.8% over a period of two months (10 measurements in each of eight test kits), expressed as a percentage of content of control value. Another noteworthy finding was that the initial quality (control kits) was below the lower specification limits with mean 87.1% (ranging from 84-89%). The test kits showed a wider variation with minimum value 49.5% and maximum value of 88%.
Simulation study on injectable oxytocics
This study7,8 comparing three drugs (ergometrine injection, methylergometrine injection and oxytocin injection represented by 11 brands) addressed specific questions on stability arising from results from field studies. No difference was found between methylergometrine and ergometrine other than the differences between brands: the loss of active ingredient when kept under refrigeration for one year ranged from 0 to 14% among the eight brands (mean 4-5%). The mean loss increased to 25% when stored at 30°C in the dark. Under exposure of light at 21-25°C, the loss was over 90% in one year. Oxytocin was found to be far more stable than (methyl) ergometrine, particularly with respect to light. Storage for 12 months at 30°C resulted in a 14% loss (range 9-19%).
An interesting finding was the correlation between discoloration and substandard content of active ingredient 9. The validation of a visual method, whereby discoloration was identified by comparing the ampoule contents with water in clear glass tubes against a well-lit white background, showed a 97-100% sensitivity for detection of sub-standard quality with about 15% false positives. This has practical applications at the level of the user and also provides a fast (and financially friendly) method which can be used at distribution level.
Overall conclusions and recommendations
The Zimbabwe study confirms, on the basis of a larger sample, the problems of poor initial quality and instability already known to affect ergometrine injection. The influence of the manufacturer of the product was shown to be the principal factor associated with the actual potency and therapeutic efficacy of ergometrine injection at end-user level The prequalification of suppliers is extremely important for this drug. Routine testing of initial quality is recommended wherever facilities exist. However, all studies to date prove that the quality of ergometrine injection cannot be assumed to be satisfactory in any circumstances. Promoting the visual method of quality assessment will enable more frequent checks and can ensure a certain level of confidence before issuing the stock. However, the reported variation between individual ampoules in a pack makes sampling a more critical and complex issue. This leads to the conclusion that an alternative formulation (such as dry powder for reconstitution) or an alternative oxytocic drug may be the only ways to ensure an efficacious product, irrespective of setting. The use of oxytocin injection has been recommended on the grounds of fewer side effects and better stability 10,11,12,13. However, this recommendation is not universally accepted, and there is little experience with the use of oxytocin injection at the PHC level.
Penicillin procaine injection
All 25 GMS samples were within specification limits with a mean assay value of 101.0% (95% CL 99.3-102.8%). There is therefore no problem with initial quality. Penicillin procaine had been out of stock for some time just before the collection of facility samples began. The consequence was that only two samples older than one year were obtained. The age of 72 facility samples therefore varied from 1.6-15.2 months with a mean of 6.0 months. Three (4%) of 72 samples were below the minimum specification (72.2, 83.0, 85.0%). All three failures were from the same batch, the same geographical area and were transported by rail (Figure 12).
Figure 12. Procain benzylpenicillin injection
As there was no problem of poor initial quality, the facility sample failures must be attributed to the instability of the penicillin. 42 GMS/facility sample pairs showed a loss of potency (mean -3.6%) which was statistically significant (p < 0.001). The extent of loss was modest, but the sample interval was only 43 months (Figure 3 in Annex 1.13).
The only formulation of penicillin procaine registered and distributed in Zimbabwe is a suspension, which has a greater likelihood of instability when compared to a dry powder product. The manufacturer recommends cool storage (8-15 degrees), but this is not practically feasible at facility level. Anecdotal reports of quality problems in samples stored at facility level were common prior to this study. Quality problems were visually identified by "yellow discoloration" and "caking" - meaning the powder in the vial had settled and could not be suspended by shaking. Experience also indicates that the problem was related to duration and temperatures of storage.
Experience from other country studies
A pilot longitudinal study in Sudan4,5 which included fortified procaine benzylpenicillin injection (dry powder form) reported no signs of instability. The study compared three field samples from district hospitals with a control batch stored by the supplier in the Netherlands.
Another study investigating the effects of the extreme conditions during shipment from Sweden and in-country distribution within Sudan6 concluded that procaine benzylpenicillin injection (dry powder) was "reasonably stable", reporting only a slight increase in degradation products with time.
There was no problem with initial quality, but evidence of moderate instability suggested that longer periods of storage at facility level would result in a loss of potency which is clinically significant. A dry powder formulation, proven to be stable, seems more appropriate in the Zimbabwe setting.