Quality of oxytocin available in low‐ and middle‐income countries: a systematic review of the literature

Oxytocin is the drug of choice for preventing and treating postpartum haemorrhage, an important cause of maternal death. Oxytocin is widely available in low and middle‐income countries (LMIC) but there are concerns about its quality.


Introduction
Every year, an estimated 94 000 women die due to bleeding complications related to pregnancy and childbirth. 1 Post-partum haemorrhage (PPH) remains the leading direct cause of maternal mortality worldwide, representing 27.1% of all maternal deaths. 1 The use of uterotonics plays an important role in the prevention and treatment of PPH. Oxytocin is recommended by the World Health Organization (WHO) as the uterotonic drug of choice for the prevention and treatment of PPH. 2  Women and Children (UNCoLSC) list. Oxytocin is widely available in low-and middle-income countries (LMIC) but there are many concerns about its quality. 3 This can be due to low manufacturing quality or inadequate transport and storage conditions along the supply chain, or both. 4 Oxytocin needs to be stored under refrigeration (2-8°C) as much as possible, although short periods of unrefrigerated transport not exceeding 1 month at 30°C or 2 weeks at 40°C, are acceptable. 5 Oxytocin should also not be frozen, although recent evidence suggests that multiple cycles of freezing and thawing do not significantly change oxytocin content. 6 Although several studies and reports have assessed the quality of oxytocin in LMIC, their findings have not been analysed or synthesised. The objective of this systematic review was to identify, critically appraise and synthesise the findings of studies on the quality of oxytocin commercially available in LMIC countries.

Methods
This review followed the recommendations of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) and the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) statements. 7, 8 We did not register a protocol for this systematic review.
The quality of medicines includes many aspects directly related to the product itself (e.g. appearance, volume, pH, pharmaceutical content, sterility, related substances, contaminants, solubility), as well as aspects related to packaging, labelling, transportation and storage conditions of the product, including light, humidity and temperature of the area where the product is kept. We report failure rates of oxytocin samples according to the parameters assessed by the authors of each included study. However, we focus on two main aspects of oxytocin quality, namely active pharmacological ingredient (API) and sterility, essential attributes of this injectable thermolabile drug, that are also the most frequently reported in primary studies on medicine quality. According to the International Pharmacopoeia, oxytocin injections should contain not less than 90% and not more than 110% of the amount of oxytocin (C 43 H 66 N 12 O 12 S 2 ) stated on the label and should be sterile. 9 These limits are also used by the British Pharmacopoeia and US Pharmacopeia. Therefore, for all these pharmacopoeias as well as for regulatory authorities and manufacturers that registered these limits, products containing less than 90.0% or over 110% oxytocin would be considered substandard and not to be used in treatment.
For the purpose of this review, oxytocin samples were considered of inadequate quality if the level of API did not reach 90% of the stated content (low fail), or was higher than 110% (high fails). Although the administration of large doses of highly concentrated oxytocin with large vol-umes (usually over 3-4 l) of intravenous liquids devoid of electrolytes (e.g. 5% dextrose) can in theory lead to severe adverse effects due to hyponatraemia, these events are rare. 10 Therefore, the main focus of our review was on the rate of oxytocin samples with less than the stated amount (low fails), since it is logical to assume that the injection of products containing less than the minimum recommended active ingredients can compromise the effectiveness of the drug in the prevention or treatment of PPH. Sterility of an injectable product is also an important quality parameter and we report the results of this test for all of the samples, if provided by the authors, in isolation or combined with the results of oxytocin quantification assays. Samples with adequate oxytocin content (i.e. API 90-110%) but that failed sterility tests were considered of poor quality.
All studies that provided assays of injectable oxytocin quantification (API) and/or sterility, using any valid laboratory method were eligible for inclusion. Studies that also reported the results of other tests (e.g. pH, contaminants, impurities, packaging status or storage temperature) were eligible for inclusion as long as they also provided separate information on oxytocin quantification and/or sterility tests. Studies assessing samples collected from the public or private sector, at facilities (e.g. hospitals, clinics, local medical stores or pharmacies, informal vendors or markets) or central distributors (e.g. warehouses, major distributors or central medical stores) were eligible for inclusion. Outlets run by nongovernmental organisations (NGO) were categorised as private.
Editorials, reviews or commentary papers were excluded from this review. We did not include studies on analytical methods for the identification of oxytocin, or studies that only assessed storage conditions or simulation studies, which created artificial conditions to test the effects of heat, light or other factors on oxytocin quality.

Process of study identification and data extraction
We searched the literature for all published studies or reports that described the quality of oxytocin samples collected in LMIC countries, regardless of sample size or date of publication. The following electronic databases were searched, from inception up to March 2015, without language restrictions: EMBASE, PUBMED/MEDLINE, African Index Medicus (AIM), International Pharmaceutical Abstracts (IPhA), LILACS, IBECS and Cochrane. The search strategy was created with the help of a librarian experienced in systematic reviews and included the terms 'oxytocin' AND 'post-partum haemorrhage" and synonyms (see Supplementary material, Appendix S1, for the full search strategy). The search was complemented by reviewing the list of references of all articles selected for full text reading and by looking for unpublished studies or reports through contacts with investigators who had conducted or were conducting studies on oxytocin quality and with WHO staff.
Citations were downloaded using a reference manager software (ENDNOTE version X7; Thomson Reuters, New York, NY, USA) into a single file and duplicates were excluded. The titles and abstracts of all unique citations were screened and those that were potentially eligible were selected for full text reading. The full text articles were obtained and assessed for inclusion based on the aforementioned selection criteria. The following data were extracted from included studies: geographical location and year of sample collection, types of outlets where samples were collected (public versus private sector, central versus facility level), origin of the drugs (manufacturer and country), types of tests performed, number of sample assayed and number of samples failed according to the authors' criteria, number of samples failed due to inadequate (low or high) API and to lack of sterility. Two independent reviewers (MRT and CG) performed the process of study selection and data extraction; disagreements were discussed until consensus was reached.

Assessment of the quality of the studies
We assessed the methodological quality of all included studies using the criteria proposed by Almuzaini et al. 11 based on the methodology section of the MEDQUARG (Medicine Quality Assessment Reporting Guidelines) checklist for reports of surveys of medicine quality. 12 Basically, each study was graded on 12 domains: (1) survey details, (2) definitions of substandard medicines, (3) types of outlets sampled, (4) sampling design and sample size calculation, (5) type and dosage of units collected per outlet, (6) random sampling, (7) details of samplers, (8) packaging assessment, (9) statistical methods, (10) chemical analysis description, (11) method validation details and (12) blinding of chemical assessors (see details in Supplementary material, Appendix S3). Two independent reviewers (MRT and CG) graded the quality of each study, compared their scores and discussed discrepancies until consensus was reached on a final grade for each study. There were no established cut-offs, so studies with a grade of 6 or more were categorised as being of good quality, as proposed by a previous review. 11

Data synthesis and statistical analyses
We first present the total rate of failed samples in general, as reported by the authors. We decided not to pool the proportion of failed samples into a meta-analysis because this definition was heterogeneous in the primary studies. We then present the prevalence of samples failed because of inadequate amount of active ingredient (API <90% or >110%), the prevalence of low fails (API <90%) and high fails (API >110%) per study and according to geographical location (continents) and type of outlets where the samples were collected. We performed subgroup analyses of sam-ples with low API content collected in public versus private sector outlets, as well as central versus facility-level settings. The countries providing samples were classified according to the World Bank as low income (LIC), lower middle income (LMIC), upper middle income (UMIC) countries. 13 The chi-square and two-sided Fisher exact tests were used for statistical analyses. P < 0.05 was considered significant.

Results
A total of 2611 unique citations were identified through various sources, 22 studies were selected for full text reading, 14 were excluded for different reasons (see Supplementary material, Appendix S2) and eight records were included in the review 5,14-20 ( Figure 1). Two were articles published in peer-reviewed journals 17,18 five were unpublished reports available online in various institutional sites 3,5,14,19,20 and one was a congress abstract. 16 We obtained additional (unpublished) details from the authors of five reports. 5,14,16,19,20 The eight included studies reported on the collection and analyses of 559 samples from 15 different countries: five studies included only LMIC, 14,[16][17][18]20 one study only a LIC, 5 one study an UMIC 19 and one study included LIC and LMIC 15 (Table 1).
The final methodological quality scores of the studies ranged from 2 to 9, out of a maximum of 12 (mean 5.8 AE 2.4, standard deviation) and four of the eight studies were considered of good methodological quality (scores ≥6). 14, [16][17][18] Almost all studies scored highly on the domains related to the description of the type of outlet sampled (domain 3), the location and timing of the study (domain 1) and the chemical analyses performed (domain 10) but none provided details about the methods used to validate the chemical tests performed (domain 11). The domains with the lowest scores were those related to the description of statistical techniques used for data analyses (domain 9), blinding of chemical assessors (domain 12) and number of dosage units obtained from individual outlet (domain 5) (see Supplementary material, Appendix S3).
The eight studies assayed a total of 559 oxytocin samples, ranging from 5 5 to 193 18 samples per study (median 34 samples per study, interquartile range 7.5-124.7) ( Table 2). Ghana was the country with the largest individual number of samples (n = 215), 14,17 followed by India 18 (n = 193), while Kenya, Madagascar and Burkina Faso contributed with only one sample each. 15 The manufacturers of 559 samples included in the eight studies were from 13 different countries, with several different manufacturers in each country. The three countries which manufactured the largest numbers of oxytocin samples were India, 3 total samples each, respectively (see Supplementary material, Appendix S4). Most of the samples were collected at facility level settings (91%, n = 509) and from the private sector (57.4%, n = 321). Five of the studies 3,14,16,19,20 performed additional tests besides oxytocin quantification assays, including tests for sterility, contaminants or strange particle matters and pH ( Table 2, and see Supplementary material, Appendix S5).
The median prevalence of oxytocin samples that did not pass quality testing, as defined by various parameters adopted by the original authors, was 45.6% (range 0-80%, interquartile range 8.8-66.7%). The median prevalence of oxytocin with inadequate API content was 36.0% (range 0-80%, interquartile range 8.8-60.7%, n = 223) and most failures (n = 204) were due to low (<90%) API content ( Table 2). Only 19 of the 559 samples (3.4%) failed because of excessively high (>110%) oxytocin content. The study with the highest rate of high fails was conducted in Zimbabwe and included only five samples of oxytocin collected in 1992, which were imported from unspecified countries; four of these (80%) failed due to API content >110%. 5 Two more recent studies, from India 18 and Indonesia, 16 also reported a small number of oxytocin samples with API >110% (n = 12 of 193 and n = 3 of 110, respectively); the medicine in both studies was manufactured by local companies (see Supplementary material, Appendix S5).
The concentration of oxytocin in the 204 samples with low API ranged from 0% 14,16 to 89.8%. 15 Most (48/69) of the failed samples from India had between 76 and 89% content 18 and most (19/35) of the low fail samples collected in Ghana by Stanton et al. 17 in 2010 had 60-89% API. The eight failed samples from the multi-country study had oxytocin concentrations between 50 and 89%. 15 There were only two samples, one from Indonesia 16 and one from Ghana, 14 with 0% API, which is an indication of possible counterfeit drug. These same two studies did not provide the exact oxytocin concentration for their other low-fail samples (n = 12 and n = 93, respectively). Table 3 presents the proportion of failed samples according to their provenance. The proportion of low fails (API <90%) was significantly higher in the private than in the public sector samples (34.0% versus 25.9%, respectively, P = 0.032). Similarly, the proportion of low fails was significantly higher in samples from facilities than from central level settings (37.9% versus 22.0%, P = 0.030). There is no information suggesting that the oxytocin samples collected at facility level were from the passed the oxytocin quantification assay (i.e. API between 90 and 110%). The proportion of failed samples was significantly higher in Africa than in Asia or Latin America (57.5% versus 22.3% versus 0%, respectively, P < 0.0001).
Only three studies provided the results of both API content and sterility tests for their samples. The six oxytocin samples collected in Guatemala 20 and the eight samples from Peru 19 passed both tests. The third study 14 assayed a total of 169 samples in Ghana and 94 failed due to low API. Due to resource constraints, the authors randomly selected 40 samples (out of the 169) and performed sterility tests only in this subset. The investigators report that 97.5% of these samples failed either the API or sterility or both tests.

Main findings
This review found a relatively small number of studies that assessed the quality of oxytocin samples in LMIC. The analyses of these samples reveal a high prevalence of poor quality oxytocin in LMIC countries, especially in Africa and Asia, and the main problem is insufficient amount of active ingredient. Over one-third of the 559 samples tested in the eight studies included in this review had oxytocin content below the minimum (90%) stated amount, indicating substandard medicine and two samples had no active ingredient, possibly corresponding to counterfeit drugs. The proportion of low fails was highest in Africa, where almost 60% of the samples assayed had <90% of the oxytocin quantity stated in the ampoules. Samples collected in the private sector and in endpoint facilities had higher proportions of low API content than those collected in the public sector or in central facilities. Few studies, involving a small number of samples, performed sterility tests and reported a high rate of samples failing either the API or sterility or both tests.

Strengths and limitations
This is the first systematic review on the quality of injectable oxytocin in LMIC. Bias was reduced by using a strict methodology including duplicate study selection, data extraction and quality assessment. We used a broad search strategy without language restriction and we tried to identify unpublished reports striving to reduce publication bias. However, although most (six of the eight) of our included studies were not papers published in peer-reviewed journals but reports, as the concern for oxytocin quality is growing, there might exist additional unpublished studies that were not identified and therefore not included in this review. We may also have missed studies with completely favourable results (all samples of good quality) because they may have more difficulty in getting published (publication bias). Finally, it is possible that we did not have access to studies    ).
****Low fail: API <90% of stated content.   with very unfavourable results (high rates of failed samples), which may have not been made publically available by national authorities. An additional limitation of our review is that half of the included studies were of poor methodological quality, mainly due to lack of information on key details (e.g. methods used to validate the chemical tests used or statistical analyses). This seems to be indicative of poor reporting quality and does not necessarily reflect poor methodological quality of the studies. Our focus on two specific aspects of quality, namely API content and sterility, is also a potential additional limitation of this review. Although other aspects of product quality (e.g. pH, solubility, contaminants, labelling) are part of standard processes for medicine quality assessment, these parameters are seldom reported by most primary studies.

Interpretation
This systematic review could not identify the specific causes for the low quality of the oxytocin samples assessed. Since most of the samples from the eight studies were collected at periphery-level facilities, it is difficult to conclude whether the low quality of the oxytocin was due to problems at the manufacturer level, inadequate storage conditions during distribution or at the facilities, or a combination of these problems. None of the studies provided information on how long the ampoules had been stored in the central or facilitylevel settings and under what conditions of temperature they had been stored over time. As product degradation is related to the time spent under suboptimal conditions, this information is highly relevant. Further research is needed to assess the factors and potential causes that may affect the quality of the oxytocin currently available in LMIC countries. Future studies conducted at the country of origin should assess the manufacturing quality of oxytocin and product regulation by the regulatory authorities. We also need studies that follow the product along the supply chain to identify whether temperature changes affect the quality of oxytocin. A study involving both of these assessments is currently being conducted in Ghana, in collaboration with the WHO.
One approach to identify the oxytocin ampoules that were exposed to high temperatures is to include temperature-time indicators (TTI) in the oxytocin ampoules. TTIs indicate excessive heat exposure over time and therefore allow health workers to make better-informed decisions on the extent to which heat has affected the product and whether to use the oxytocin ampoules or not. TTIs mimic the degradation of the pharmaceutical product against time and temperature with a safety margin, and are calibrated against the stability characteristics of a product. Vaccine vial monitors (VVM = TTI for vaccines) have been used for all WHO prequalified vaccines since 1996 and are included among the minimum requirements in all interna-tional tenders by all United Nations procurement agencies and any member states. 21,22 A simulation study involving TTIs with oxytocin used VVM type 30 (meaning that VVM will reach its discard point in 30 days if kept at 37°C constant temperature), found that a considerable number of oxytocin samples were rejected although they were still within acceptable specifications. 23 However, we must note that VVM is not a potency indicator and is designed to warn health workers on the impact of time and temperature on the product with a reasonable safety margin. In addition, when it is decided to have TTIs for oxytocin, the temperature sensitivity of the product should be studied to match the best TTI category-which may not correspond to VVM30. 24 All included studies except one 5 assayed samples collected since 2010, showing that the quality of oxytocin is increasingly becoming a global concern. If the samples obtained in the studies included in this review are representative of what is currently available in LMIC countries, for every three women who receive oxytocin in these settings, at least one will receive an injection that contains less than the expected amount of the medicine and will therefore be denied the effects of this potentially life-saving medicine.
Most of the samples collected in the two largest studies conducted in Africa 14,17 were from non-registered manufacturers. Moreover, none of the samples analysed in all the included studies are WHO-prequalified oxytocin products. The registration of a product in a country ensures the safety, efficacy and the quality of the drugs, and the WHO prequalification process helps to identify quality drugs from countries. However, to achieve prequalification, producers need to improve the manufacturing process and consequently increase the price of the product. Therefore, as highlighted in the Jhipiego business case 4 governments must incentivise manufacturers to produce high-quality oxytocin without losing competitively against non-quality-assured drugs. On the other hand, having oxytocin included in the WHO prequalification list is helpful to procurement agencies to identify quality products. Globally, approximately 100 million doses per year are used for prevention and treatment of PPH and there are at least 300 different oxytocin products manufactured, creating a market that is difficult to regulate. 4

Conclusion
There is a widespread problem with the quality of oxytocin samples from LMIC countries, especially in Africa and Asia. This can potentially be a contributing factor to the high rates of maternal morbidity and mortality in these settings. Future studies are needed to identify the exact factors involved in the low quality of oxytocin in LMIC countries and to plan appropriate interventions to improve this situation.