The authors have declared that no competing interests exist.
¶ Membership of the Difäm-EPN Minilab Survey Group is provided in the Acknowledgments
Substandard and falsified medical products present a serious threat to public health, especially in low- and middle-income countries. Their identification using pharmacopeial analysis is expensive and requires sophisticated equipment and highly trained personnel. Simple, low-cost technologies are required in addition to full pharmacopeial analysis in order to accomplish widespread routine surveillance for poor-quality medicines in low- and middle-income countries.
Ten faith-based drug supply organizations in seven countries of Africa and Asia were each equipped with a Minilab of the Global Pharma Health Fund (GPHF, Frankfurt, Germany), suitable for the analysis of about 85 different essential medicines by thin-layer chromatography. Each organization was asked to collect approximately 100 medicine samples from private local medicine outlets, especially from the informal sector. The medicine samples were tested locally according to the Minilab protocols. Medicines which failed Minilab testing were subjected to confirmatory analysis in a WHO-prequalified medicine quality control laboratory in Kenya.
Out of 869 medicine samples, 21 were confirmed to be substandard or falsified medical products. Twelve did not contain the stated active pharmaceutical ingredient (API), six contained insufficient amounts of the API, and three showed insufficient dissolution of the API. The highest proportion of substandard and falsified medicines was found in Cameroon (7.1%), followed by the Democratic Republic of Congo (2.7%) and Nigeria (1.1%). Antimalarial medicines were most frequently found to be substandard or falsified (9.5% of all antimalarials). Thin-layer chromatography according to the Minilab protocols was found to be specific and reproducible in the identification of medicines which did not contain the stated API. Since only samples which failed Minilab testing were subjected to confirmatory testing using pharmacopeial methods, this study did not assess the sensitivity of the Minilab methodology in the detection of substandard medicines, and may underestimate the prevalence of poor-quality medicines.
Surveillance for poor-quality medicines can be carried out by local organizations in low- and middle-income countries using a simple, low-cost technology. Such surveillance can identify an important subgroup of the circulating substandard and falsified medical products and can help to prevent them from causing harm in patients. A collaboration of the national drug regulatory authorities with faith-based organizations and other NGOs may therefore represent a promising strategy towards the Sustainable Development Goal of “ensuring access to quality medicines”.
The United Nations have declared access to “safe, effective, quality, and affordable essential medicines” to be one of the Sustainable Development Goals in their 2030 Agenda [
The American Journal of Tropical Medicine & Hygiene has recently devoted a special issue to the “Global Pandemic of Falsified Medicines” [
The quality standards which pharmaceuticals must comply with, as well as the methods to prove their compliance or non-compliance, are defined in the pharmacopeias, such as the International Pharmacopeia, the United States Pharmacopeia and the British Pharmacopeia. Most commonly, the identity and the amount of the active pharmaceutical ingredient (API) are determined using high performance liquid chromatography (HPLC). For solid oral dosage forms, the dissolution of the API which is a necessary precondition for bioavailability and therefore for effectiveness is determined in a special dissolution apparatus using quantification by HPLC or ultraviolet spectroscopy. Other pharmacopeial tests include e.g. tests for uniformity of the dosage units in terms of mass and in terms of content of API, tests for friability (i.e. mechanical stability/durability of tablets) etc. The equipment required for pharmacopeial analysis, especially for HPLC, is expensive and delicate, requiring an appropriate laboratory infrastructure including an electricity supply of constant voltage and regular maintenance by skilled personnel from the manufacturer. Performing the analyses requires highly trained professionals and expensive reagents, standards and organic solvents of high purity. Therefore, in LMICs only a few laboratories exist which can carry out such analyses. Their capacity is usually very limited and does not allow for large-scale routine surveillance of medicines in the health facilities and markets. The costs for pharmacopeial analysis are very high. A recent study reported that the price for the analysis of a single medicine sample, offered by a WHO-prequalified laboratory in South Africa, was on average 1,580 US$ [
Therefore, more affordable methods for the surveillance of medicine quality are desirable. Currently, two such technologies are commercially available and widely used. One is represented by portable Raman (or near-infrared) spectroscopy instruments, such as the hand-held TruScan RM by Thermo Fisher Scientific, Waltham, Massachusetts, USA. Raman spectroscopy requires very little time and therefore only little costs for labour. However, Raman instrumentation still requires a high capital cost compared to thin layer chromatography (see below), and depends on the availability of a library of pre-recorded spectra of authentic medicines [
The Ecumenical Pharmaceutical Network (EPN) is an international faith-based network based in Nairobi, Kenya, which comprises members in 36 countries. It seeks to strengthen the faith-based pharmaceutical sector in developing countries and to improve people’s access to quality pharmaceutical services. One of the EPN member organizations is the German Institute for Medical Mission (Deutsches Institut für Ärztliche Mission; Difäm), a German faith-based NGO promoting health care in developing countries. Among other activities, Difäm aims to strengthen 15 faith-based drug supply organizations (DSOs) in different African and Asian countries who procure drugs and medical supplies for faith-based health care institutions in their respective countries. The total annual turnover of these DSOs is about 90 million US $.
Starting from 2010 and supported by funds from the German faith-based charitable organization Bread for the World, Berlin, Germany, successively each DSO was supplied with a GPHF Minilab. By the time the present study was initiated (early 2015), ten faith-based drug supply organization from six African countries and from India had each received a GPHF Minilab and appropriate training, and used the Minilab for basic quality testing of those medicines which they procured for the faith-based health-care services in their respective countries.
In order to document the possibilities and limitations of the use of the GPHF Minilab by local drug supply organizations in Africa and Asia, the current survey was initiated. Each of the ten involved organizations in Africa and Asia was asked to collect 100 medicine samples from private medicine vendors, especially informal (= non-licensed) vendors, to analyse these samples according to the Minilab manuals, and to report the results to Difäm. If samples failed Minilab testing, a confirmatory test according to pharmacopeial procedures was carried out by the WHO-prequalified quality control laboratory of the Mission for Essential Drugs and Supplies (MEDS) in Kenya [
In total, 869 medicine samples were analysed in this study, leading to the identification of 21 medicines which were confirmed to be falsified or substandard.
Medicine samples were collected between April and September 2015.
The ten involved organizations were requested to collect samples of medicines for which analytical protocols existed in the GPHF Minilab manuals [
The involved organizations were asked to collect medicines in the areas where they are located. Two organizations were based in Cameroon, one in the South-West Region and the other one in the North-West-Region. Two partners were located in the eastern part (Bukavu and Bunia) of the Democratic Republic of Congo (DRC). The other African organizations were located in Ghana (Accra), Kenya (Nairobi), Nigeria (Plateau State) and Uganda (Kampala). The two partners from India were based in the regions of Odisha and Tezpur, in the east part of the country.
For the selection of the medicine collection sites, convenience sampling was used. The involved organizations were requested to collect medicine samples in their respective countries from different private drug outlets, preferably from informal (= non-licensed) drug vendors e.g. at local markets or bus stations. Since the types of formal and informal drug vendors are different in each country, no further standardization of the included collection sites was attempted. In the reports supplied by the involved organizations, the terms used to describe the collection sites were heterogeneous, e.g. “drug vendor on the open market”, “pharmacy on black market” or “private shop”. The described collection sites also included outlets with were most likely licensed, e.g. “community pharmacy”. In case of Uganda, apparently also pharmaceutical wholesalers were included.
When collecting medicines from informal (= non-licensed) drug vendors, a mystery shopper approach was used routinely. However, the investigators were free to use an overt approach (i.e. identifying themselves and the purpose of the study) if this was necessary, e.g. due to requirements by the local health authorities. All medicines collected were being paid for. From each sample, 150 tablets/capsules or 50 vials of the same batch were to be collected if available. Medicines were collected in their original packages whenever possible. They were transported immediately to the local laboratory and thereupon stored in a dry, cool place until analysis. If they needed to be forwarded to another organization for retesting or for confirmatory pharmacopeial testing, transport was carried out by a commercial courier service at ambient temperature.
Analysis according to the Minilab protocol was carried out in the laboratories of each of the ten involved faith-based drug supply organizations. For each involved organization a five-day workshop had been held, training several of its coworkers in the procedures of GPHF Minilab analysis. Testing according to pharmacopeial monographs was carried out at the WHO-prequalified medicine quality control laboratory of the Mission for Essential Drugs and Supplies (MEDS) in Nairobi, Kenya [
Physical inspection, thin-layer chromatography (TLC), colour reaction testing and disintegration testing were carried out according to the manuals of the GPHF Minilab [
For instant-release oral dosage forms, disintegration testing was performed using six tablets or capsules. These were kept in water at 37°C with occasional shaking or stirring. Disintegration was required to occur within 30 min. If not all of the tablets disintegrated in this time, the test was repeated (total three times).
For TLC testing [
At the beginning of this survey, samples which did not pass TLC testing were directly forwarded to the laboratory of MEDS, Kenya, for confirmatory testing according to pharmacopeial procedures. Several of these samples were found to pass analysis in the MEDS laboratory. Therefore, the procedure was modified in order to save expenses for the costly pharmacopeial analysis. Samples which did not pass TLC testing by the organization which had collected the sample were subsequently forwarded to another of the involved organizations for retesting. Only if they failed again, the samples were then forwarded to the laboratory of MEDS for pharmacopeial testing. Therefore, results of a second TLC test are only available for a part of the samples and are not included into
Number of: | % of samples confirmed to fail pharmaco-peial tests |
||||||||
---|---|---|---|---|---|---|---|---|---|
sam-ples repor-ted | sam-ples exclu-ded | sam-ples inclu-ded | medi-cines (brands) inclu-ded | bat-ches inclu-ded | sam-ples failing 1st TLC test | sam-ples failing pharma-copeial tests | |||
1 | Cameroon | 111 | 5 | 106 | 86 | 97 | 12 | 9 | 8.5% |
2 | Cameroon | 108 | 2 | 106 | 74 | 105 | 11 | 6 | 5.7% |
3 | DR Congo | 85 | 0 | 85 | 83 | 84 | 8 | 4 | 4.7% |
4 | DR Congo | 98 | 0 | 98 | 67 | 82 | 1 | 1 | 1.0% |
5 | Nigeria | 98 | 3 | 95 | 93 | 95 | 3 | 1 | 1.1% |
6 | Kenya | 94 | 0 | 94 | 78 | 89 | 0 | 0 | 0% |
7 | Uganda | 100 | 69 | 31 | 31 | 31 | 0 | 0 | 0% |
8 | Ghana | 105 | 16 | 89 | 59 | 81 | 0 | 0 | 0% |
9 | India | 101 | 0 | 101 | 64 | 98 | 0 | 0 | 0% |
10 | India | 64 | 0 | 64 | 33 | 64 | 0 | 0 | 0% |
Total | 964 | 95 | 869 | 622 |
816 |
35 | 21 | 2.4% |
# Only samples which failed TLC testing were subjected to confirmatory testing using pharmacopeial methods. It is therefore possible that the true percentage of medicines failing pharmacopeial standards is higher than indicated in the last column of this table.
* Numerical addition of the numbers of medicines and batches collected by the individual organizations would result in 668 medicines and 826 batches. The indicated total numbers of medicines and batches included in this study is slightly smaller since identical medicines (brands) and batches were sometimes collected by different organizations.
This study focused on TLC analysis. If a sample passed TLC testing but failed colour reaction testing, disintegration testing or physical inspection, the result was recorded but the sample was not forwarded to the MEDS laboratory, since funds for the costly pharmacopeial analysis were limited.
In the WHO-prequalified medicine quality control laboratory of MEDS, Kenya, the samples were analysed according to the specifications of the pharmacopeia which was indicated by the manufacturer on the product label. Typically, these tests included: identity; assay for amount of active pharmaceutical ingredients (APIs) declared on the label; dissolution of the APIs in case of solid dosage forms; and uniformity of dosage units by mass and by API content. Reference standards were obtained from the United States Pharmacopeial Convention (
The following pharmacopeial monographs were applied by MEDS:
USP38-NF33 for amoxicillin/clavulanic acid tablets (dissolution testing according to USP36-NF31); azithromycin tablets; chloroquine phosphate tablets; mebendazole tablets; sulfadoxine/pyrimethamine tablets.
BP 2015 for amoxicillin capsules; ampicillin capsules; captopril tablets; clomifene citrate tablets; metronidazole tablets; prednisolone tablets; quinine sulfate tablets.
MEDS in-house methods for dihydroartemisinin/piperaquine tablets (identity, assay, dissolution).
In the laboratory of the Centers for Disease Control and Prevention (CDC), quantitative analysis of dihydroartemisinin and piperaquine in tablets was performed using a modified high performance liquid chromatographic (HPLC) procedure described by Green
In TLC testing, the Rf (= retention factor) is the ratio of the distance travelled by the API divided by the total distance travelled by the mobile phase. Samples were considered non-compliant if the Rf value of the APIs was different by more than 10% from that of the authentic standards, and/or if the intensity of the spot was less than that of a reference containing 80% of the stated amount of the API. Before concluding non-compliance, the, TLC analysis had to be repeated twice starting from another tablet (or capsule/vial). Therefore, a negative outcome had to be observed in three independent experiments. TLC results were recorded in a standardized table. TLC plates were not photographed in some cases but not routinely.
In disintegration testing, non-compliance was concluded if, in three tests with six dosage units each, more than two out of 18 dosage units did not disintegrate in 30 min. The solid oral dosage forms included in this study did not include any slow-release or enteric-coated tablets, i.e. all were expected to disintegrate in 30 min.
In the laboratory of MEDS, definition of non-compliance followed the specifications of the respective pharmacopeial monograph. Before concluding non-compliance, the analysis was repeated by another investigator in that laboratory. Dihydroartemisinin/piperaquine tablets, which were investigated according to MEDS in-house methods, were considered non-compliant if the content of either or both APIs deviated by more than 5% from the stated amount, or if less than 70% of either or both APIs dissolved in dissolution medium in 60 min.
For the conversion of cost estimates from Euro to US$ the exchange rate of 1st April 2015 was used (1 Euro = 1.0772 US$)
The ten faith-based drug supply organization from six African countries and from India who participated in the present survey (
As shown in
Screening by thin layer chromatography (TLC) led to the identification of 20 samples which were confirmed to be substandard or falsified medical products. One further sample with insufficient dissolution of the active pharmaceutical ingredient was discovered from disintegration testing (see text).
Of the 964 samples, 95 had to be excluded from data analysis. For 58 of them, no TLC results were reported. Out of these 58 cases, 54 represented medicines for which no protocol for TLC analysis was available in the manuals of the GPHF Minilab and therefore TLC analysis could not be carried out. For 4 of these 58 cases, TLC results were missing for unknown reasons.
Twenty samples were excluded from data analysis in this survey because TLC results were reported despite the fact that no Minilab protocol existed for the respective active pharmaceutical ingredient at the time of this survey (e.g. samples of amlodipine, nifedipine and ibuprofen). Seventeen further samples were excluded since TLC results were reported but no Minilab protocol existed for the respective dosage form (e.g. samples of amoxicillin suspension, ciprofloxacin i.v. infusion and metronidazole i.v. infusion). Most of these 37 TLC reports came from two of the involved organizations. Although it cannot be excluded that the investigators in these organizations developed and used their own protocols for TLC analysis, a likely explanation is that, after the sample passed visual inspection, the result “complies” was simply entered into all fields of the reporting table in these cases. This notion is supported by the observation that for 14 of the above mentioned 37 samples, the result of the disintegration test was reported as “complies”, despite the fact that a disintegration test was impossible for the respective dosage forms (e.g. oral suspensions or i.v. solutions). All 14 of these cases came from the same organization, indicating a misunderstanding of the correct procedures in the laboratory of this organization.
As shown in
No entry in a cell signifies zero.
Organization and country of collection | Drug class (WHO ATC code) | ||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
J01 Antibacterials for systemic use | P01 Antiprotozoals | N02, M01 Analgesics | P02 Anthelminthics | J02 Antimycobacterials | A10 Drugs used in diabetes | R03 Drugs for obstructive airway diseases | C07 Beta blocking agents | A07 Antifungals for dermatological use | C03 Diuretics | H07 Corticosteroids for systemic use | G03 Sex hormones, modulators of genital system | C09 Agents acting on the renin-angiotensin system | Total | ||
1 | Cameroon | 49 | 14 | 4 | 7 | 10 | 5 | 6 | 4 | 3 | 4 | 106 | |||
2 | Cameroon | 52 | 29 | 6 | 4 | 1 | 6 | 4 | 4 | 106 | |||||
3 | DR Congo | 47 | 4 | 15 | 10 | 1 | 1 | 6 | 1 | 85 | |||||
4 | DR Congo | 45 | 8 | 17 | 12 | 2 | 9 | 5 | 98 | ||||||
5 | Nigeria | 47 | 17 | 6 | 6 | 2 | 12 | 5 | 95 | ||||||
6 | Kenya | 46 | 13 | 6 | 7 | 4 | 4 | 2 | 6 | 2 | 2 | 2 | 94 | ||
7 | Uganda | 15 | 3 | 4 | 1 | 1 | 2 | 1 | 2 | 1 | 1 | 31 | |||
8 | Ghana | 35 | 10 | 8 | 6 | 8 | 6 | 1 | 6 | 9 | 89 | ||||
9 | India | 55 | 7 | 13 | 9 | 6 | 11 | 101 | |||||||
10 | India | 35 | 4 | 1 | 17 | 5 | 2 | 64 | |||||||
No. of samples per drug class | 426 | 105 | 83 | 63 | 32 | 38 | 30 | 25 | 21 | 21 | 11 | 10 | 4 | 869 | |
No. failing pharma-copeial analysis | 4 | 10 | 2 | 1 | 3 | 1 | 21 |
Generic medicines sold under their international non-proprietary names (INN) constituted 310 (35.6%) of the samples. So-called branded generics, sold under a brand name given by the respective manufacturer, represented 538 (61.9%). Only 21 (2.4%) were originator brand medicines (as judged from the label claim), sold under the brand name given by the manufacturer who prepared the original new drug application for the respective API.
The medicines analysed in this study came from more than 290 manufacturers in 32 different countries, as judged from their labels.
Organization and country of collection | Stated country of origin | Total | |||||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
India | China | Ghana | Kenya | Nigeria | Uganda | DR Congo | Benin | Cameroon | South Africa | UK | Germany | France | Netherlands | Cyprus | Italy | Spain | Switzerland | Belgium | Poland | Others |
|||
1 | Cameroon | 50 | 33 | 5 | 3 | 1 | 5 | 2 | 3 | 1 | 1 | 2 | 106 | ||||||||||
2 | Cameroon | 60 | 14 | 19 | 1 | 1 | 1 | 3 | 4 | 1 | 2 | 106 | |||||||||||
3 | DR Congo | 25 | 19 | 20 | 6 | 9 | 2 | 1 | 1 | 2 | 85 | ||||||||||||
4 | DR Congo | 47 | 36 | 1 | 4 | 1 | 7 | 2 | 98 | ||||||||||||||
5 | Nigeria | 34 | 23 | 31 | 1 | 1 | 5 | 95 | |||||||||||||||
6 | Kenya | 30 | 7 | 36 | 1 | 2 | 6 | 2 | 2 | 8 | 94 | ||||||||||||
7 | Uganda | 14 | 1 | 2 | 6 | 2 | 2 | 1 | 3 | 31 | |||||||||||||
8 | Ghana | 8 | 70 | 6 | 2 | 2 | 1 | 89 | |||||||||||||||
9 | India | 101 | 101 | ||||||||||||||||||||
10 | India | 64 | 64 | ||||||||||||||||||||
No. of samples per country of origin | 433 | 133 | 70 | 59 | 55 | 17 | 9 | 4 | 2 | 2 | 20 | 12 | 8 | 7 | 5 | 2 | 2 | 2 | 1 | 1 | 25 | 869 | |
No. failing pharmacopeial analysis | 3 | 5 | 4 | 1 | 2 | 1 | 3 | 2 | 21 |
No entry in a cell signifies zero.
* Others = 11 other countries, listed in
Out of the 869 samples included in the data analysis, 35 (4.0%) were reported to fail TLC testing according to the Minilab protocols, carried out by the organization who collected that sample (
Out of these 35 samples, 19 were retested in the laboratory of another of the involved organization. Eight of these samples passed the TLC retesting and were not investigated further. Of the remaining 27 samples, two could not be further investigated since too few tablets or vials were available. One sample was lost to follow-up for unknown reasons.
Therefore, 24 samples were tested in the WHO-prequalified laboratory of MEDS, Kenya, or in the laboratory of CDC in Atlanta, GA, USA. Four of these were found to comply with the pharmacopeial specifications. The remaining 20 samples were confirmed not to comply with the pharmacopeial specifications, and details of these samples are presented in
Sam-ple no. | Country and organization of collection | Active pharmaceutical ingredient | Medicine name | Batch No. | Stated manufacturer | Stated country of origin | Confirmation of negative TLC result from: | Reason for non-compliance | |
---|---|---|---|---|---|---|---|---|---|
1 | DR Congo | 3 | Amoxicillin | Amoxyverse 250 mg cps | 021645 | Universe Pharmaceutical Ltd | Kenya | Analysis by MEDS | Absence of stated API |
2 | DR Congo | 3 | Ampicillin | Ampiverse 250 mg cps | 021645 | Universe Pharmaceutical Ltd | Kenya | Analysis by MEDS | Absence of stated API |
3 | Cameroon | 1 | Amoxicillin/ clavulanic acid | Augmentin 500mg/125mg tbl | 448653 | GlaxoSmithKline | United Kingdom | Analysis by MEDS | Absence of stated API |
4 | Cameroon | 1 | Dihydroartemisinin/ piperaquine | Duo-Cotecxin 40/320 mg tbl | 031331 | Zhejiang Holley Nanhu Pharmaceutical Co. Ltd | China | Analysis by CDC | Absence of stated API |
5 | Cameroon | 1 | Dihydroartemisinin/ piperaquine | Duo-Cotecxin 40/320 mg tbl | 031331 | Zhejiang Holley Nanhu Pharmaceutical Co. Ltd | China | Analysis by CDC | Absence of stated API |
6 | Cameroon | 1 | Dihydroartemisinin/ piperaquine | Duo-Cotecxin 40/320 mg tbl | 010906 | Zhejiang Holley Nanhu Pharmaceutical Co. Ltd | China | Analysis by CDC and by MEDS | Absence of stated API |
7 | Cameroon | 2 | Dihydroartemisinin/ piperaquine | Duo-Cotecxin 40/320 mg tbl | 111132 | Zhejiang Holley Nanhu Pharmaceutical Co. Ltd | China | Analysis by CDC | Absence of stated API |
8 | Cameroon | 2 | Sulfadoxine/ pyrimethamine | Maloxine 500/25 mg tbl | TE-3293 | Gracure Pharmaceuticals Ltd | India | Analysis by MEDS | Absence of stated API |
9 | Cameroon | 2 | Sulfadoxine/ pyrimethamine | Maloxine 500/25 mg tbl | EM-304 | Shreechem Laboratories | India | Analysis by MEDS | Absence of stated API |
10 | Nigeria | 5 | Quinine sulfate | Quinine sulfate 300 mg tbl | 38763 | Remedica Ltd | Cyprus | WHO alert No. 132/2014 | Absence of stated API |
11 | Cameroon | 2 | Quinine sulfate | Quinine sulfate 300 mg tbl | 38763 | Remedica Ltd | Cyprus | WHO alert No. 132/2015 | Absence of stated API |
12 | Cameroon | 2 | Quinine sulfate | Quinine sulfate 500 mg tbl | 10H05 | Novadina Pharmaceutical Ltd | UK | Analysis by MEDS and WHO alert No. 4/2016 | Absence of stated API |
13 | DR Congo | 4 | Chloroquine | Chloroquine Rene | 00312 | Rene Industries Ltd. | Uganda | Analysis by MEDS | Assay: API only 7.1% of stated amount |
14 | Cameroon | 1 | Clomifene | Clomid 50 mg tbl | 7648 | Patheon France S.A,. for Aventis | France | Analysis by MEDS | Assay: API only 8.2% of stated amount |
15 | Cameroon | 1 | Clomifene | Clomid 50 mg tbl | 7648 | Patheon France S.A,. for Aventis | France | (same batch as sample no. 13) | (same batch as sample no. 13) |
16 | Cameroon | 2 | Clomifene | Clomid 50 mg tbl | 7648 | Patheon France S.A,. for Aventis | France | (same batch as sample no. 13) | (same batch as sample no. 13) |
17 | Cameroon | 1 | Captopril | Captopril 25 mg tbl | TT13152 | Tuton Pharmaceuticals | India | Analysis by MEDS | Assay: API only 50.0% of stated amount |
18 | Cameroon | 1 | Prednisolone | Jpsone 5 mg tbl | 130721 | Jiangxi Xi'er Kangtai Pharm. Co. Ltd, for Klusyl Internat. Co. Ltd | China | Analysis by MEDS | Assay: API only 84.0% of stated amount |
19 | DR Congo | 3 | Mebendazole | Natoa 100 mg tbl | 63498 | Laboratory & Allied Ltd. | Kenya | Analysis by MEDS | Dissolution: 7.9% of stated amount |
20 | DR Congo | 3 | Mebendazole | Wormex 100 mg tbl | L1333 | Mac' S Pharmaceuticals Ltd | Kenya | Analysis by MEDS | Dissolution: 9.3% of stated amount |
21 | Cameroon | 1 | Azithromycin | Azithromycin 500 mg tbl | 131082 | KIP Hamburg GmbH | Germany | Analysis by MEDS | Dissolution: 31.6% of stated amount |
One additional sample was tested in the laboratory of MEDS although it had passed TLC testing, but it had failed the disintegration test according to the Minilab protocol. It was confirmed not to comply with the pharmacopeial specifications, and is listed as sample no. 21 in
As shown in
For samples no. 10 and 11 shown in
Three of the samples listed in
The national drug regulatory agencies and the WHO Medical Product Alert System were informed about these findings, and warnings about several of these substandard and falsified medicines have been published on the website of the Global Pharma Health Fund [
In the present study, TLC testing using the GPHF Minilab was found to be specific and reproducible for the identification of medicines which do not contain the stated API. Sixteen samples were reported not to contain the stated API in the initial TLC analysis, and indeed fourteen of those were confirmed in pharmacopeial analysis to contain either no API or less than 10% of the stated amount of the API. For one further sample, absence of the stated API was confirmed by TLC retesting in the laboratory of another organizations, but the sample was subsequently lost to follow-up for unknown reasons (as mentioned above). Only for one of these 16 samples, retesting by TLC by another organization showed that the sample passed the retest.
Likewise, the two samples for which the collecting organization reported incorrect Rf values in TLC analysis (which implies absence of the API with the correct Rf value), were confirmed in pharmacopeial analysis not to contain the stated API in one case, and less than 10% of the stated amount of the API in the other case.
In summary, out of the 35 samples reported to fail initial TLC testing, 18 failed because the API with the correct Rf value could not be detected. Sixteen of these (
Seventeen samples were reported to fail initial TLC testing due to various other reasons, such as appearance of additional spots. Only four of those were subsequently confirmed not to comply with pharmacopeial specifications (
The focus of the present study was on TLC testing. Nevertheless, disintegration testing was carried out for the solid oral dosage forms according to the Minilab protocol [
Notably, out of the thirty-five samples which failed initial TLC testing fourteen also failed disintegration testing, showing that poor-quality medicines frequently show multiple quality deficiencies.
Until recently, the Minilab protocols comprised besides TLC testing also colour reaction tests for the identity of the APIs [
The Minilab protocol comprises, as a first step of medicine quality investigation, a physical inspection of dosage forms and packaging materials. Visual examination of the packaging sometimes allows the immediate identification of falsified medicines from incorrect or inconsistent labelling. This may be exemplified by the two samples depicted in
Left: Falsified Clomid tablets. Note the misspelling “Citrate de clomifère” instead of “Citrate de clomifène”. Right: Falsified Azithromycin tablets. The indicated manufacturer “KIP Hamburg GmbH Germany” does not exist. Further details of these two falsified medicines are given in
Furthermore, visual and physical examination can detect failures in the appearance of the dosage forms (e.g. erosions, discolorations). However, as noticed in previous studies [
For the present survey, each of the ten involved organizations was provided with an extra budget of 1,600 US$ (for purchasing and transport costs) in order to collect and analyse approximately 100 medicine samples and to report the results to Difäm. The laboratory of the Mission for Essential Drugs and Supplies (MEDS), Nairobi, Kenya carried out confirmatory analyses for this study at a reduced price, (390–580 US$ per sample; average 450 US$ per sample), depending on the active pharmaceutical ingredient and on the tests required by the respective pharmaceutical monograph. In total, 18 pharmacopeial tests were carried out by MEDS specifically for this survey, amounting to 8,100 US$ in total. Four further samples were tested free of charge in the laboratory of the Centers for Disease Control and Prevention (CDC), Atlanta, GA, USA (courtesy of Michael D. Green).
Therefore, the total external budget support required for this survey may be estimated as approximately 43,850 US$, or 50 US$ for each medicine sample included into the data analysis. The costs for the personnel of the ten involved drug supply organizations in Africa and Asia, and of the coordinating organization in Germany, has to be considered in addition.
Even at the reduced price of 450 US $ per sample offered by MEDS for this study, the full pharmacopeial analysis of all 869 samples included into the data analysis would have costed 391,050 US $, increasing the external budget support required for this study approximately 10-fold. As mentioned in the Introduction section, a recent study reported that the price for pharmacopeial analysis of various essential medicines quoted by a WHO-prequalified laboratory in South Africa was, on average, 1,580 US $ per sample [
In the present survey, 869 medicine samples from seven countries were investigated. This is a comparatively large study. Nayyar
In the present survey, 21 medicines (i.e. 2.4% of the 869 samples) were confirmed to be substandard or falsified medical products (
For poor-quality medicines, WHO has discontinued the use of the misleading term “counterfeit medicines” which merely denotes an infringement of a registered trade mark, but does not consider medicine quality [
If a medicine sample contains no or only a very small amount of the stated API, the likelihood that this grave mistake in production and quality control occurred without intent is very low. Therefore, the samples no. 1–16 in
Samples 17–20 failed to meet their quality standards but there is no indication whether this was due to deliberate/fraudulent intent. Unless proof of deliberate falsification is provided, these samples may therefore be classified as substandard medical products.
Within the seven countries included in this survey, the highest proportion of substandard and falsified medicines was found in Cameroon (total 15 out of 212 samples = 7.1%), followed by the Democratic Republic of Congo (2.7%) and Nigeria (1.1%). As mentioned, the true number of poor-quality medicines is likely to be even higher than detected in this study.
Comparing different therapeutic categories, the highest percentage of substandard and falsified medicines was found in antimalarial medicines (10 out of 105 samples, i.e. 9.5%). In Cameroon 8 out of 43 antimalarials (18.6%) did not contain the active pharmaceutical ingredient(s) and were therefore regarded as falsified. This figure from Cameroon is similar to the previously reported figure of 20% falsified antimalarials given by Nayyar
In the literature, the data on the prevalence of falsified medicines are highly conflicting. Some very credible studies report much lower figures than reported by Nayyar
The prevalence of substandard and falsified medicines is known to be very different between different countries and regions. E.g. the above mentioned study by WHO [
Regarding the stated origin of the medicines, the highest proportion of poor-quality medicines was found within the samples supposedly manufactured in Europe (13.3%), followed by samples stated to originate from China (3.8%) and from sub-Saharan Africa (1.8%) (
In this survey, the percentage of substandard and falsified medicines was nearly identical in generic medicines and in so-called “branded generic medicines” (1.9% and 2.0%, respectively). In contrast, within the samples claiming to be originator brand medicines, 4 of 21 (19%) were substandard or falsified, probably indicating a high propensity of originator brand medicines to become victims of criminal falsification. For patients in developing countries, the purchase of medicines labelled as originator brand medicines may therefore not present a useful strategy in order to avoid the risk of receiving substandard and falsified medicines.
The present study confirms previous reports [
When a medicine sample is reported to fail TLC analysis, photographic documentation of the TLC plate under appropriate detection (usually by UV light) provides more solid evidence than a narrative report alone. Photography of the TLC plates has been exemplified in a recent study [
Nearly all substandard and falsified medicines identified in the present study (
Insufficient dissolution of the API is a frequent and serious quality problem of medicines in developing countries [
Besides the GPHF Minilab, another tested and commercially available low-cost technology for medicine quality analysis is Raman (and near-infrared) spectroscopy. As mentioned in the Introduction, the identification of falsified medicines with Raman spectroscopy depends on the availability of pre-recorded reference spectra of authentic samples for each and every preparation which is to be investigated. The present survey included 622 medicines from more than 290 manufacturers in 32 countries. It would have been impossible to obtain a library of reference spectra for all these medicines before this study. Therefore, for medicine quality surveys including a large number of different medicines in developing countries, the GPHF Minilab may currently be the only commercially available ready-to-use low-cost technology.
The central aim of the present survey was to investigate the possibilities and limitations of a surveillance of medicine quality in developing countries using the low-cost GPHF Minilab technology, carried out by the professional staff of local drug supply organizations. Inevitably, the present survey has limitations. The simple analytical methods of the GPHF Minilab are able to detect only a certain part of all possible (and relevant) quality deficiencies of medicines. Since only samples which failed Minilab testing were subjected to confirmatory testing using pharmacopeial methods, this study did not assess the sensitivity of the Minilab methodology in the detection of substandard medicines. Furthermore, the design of this multi-country study did not rigorously standardize the type of medicines to be included, or the type and location of the collection sites. Convenience sampling rather than random sampling was used in the selection of the collection sites. Therefore, in several aspects the present survey deviated from current guidelines on the conduct of surveys of the quality of medicines [
In the present survey, 869 medicine samples were analysed with the GPHF Minilab, and in collaboration with a WHO-prequalified quality control laboratory, 21 samples were unequivocally confirmed to represent substandard or falsified medicines. Since only samples which failed Minilab testing were subjected to confirmatory testing using pharmacopeial methods, it is possible that additional substandard and falsified medicines were present and could not be detected with the applied methodology. Nevertheless, this study shows that surveillance for medicines which contain no or only very small amounts of the active pharmaceutical ingredient can be established at very moderate expense by local drug supply organizations using the GPHF Minilab. Such an approach can identify an important subgroup of the substandard and falsified medicines which are in circulation in developing countries. A collaboration of the national drug regulatory authorities with faith-based drug supply organizations and other NGOs, using simple low-cost analytical technologies, may represent a promising, cost-effective and as yet underutilized strategy in order to identify substandard and falsified medicines in developing countries and to prevent them from reaching the patient, in accordance with the Sustainable Development Goal to “ensure access to quality medicines”.
(DOCX)
(DOCX)
(DOCX)
The authors wish to thank Rüdiger Kilian for training of the staff of the involved organizations, Adeline Mensen for help in the compilation of the results, Michael Deats and Pernette Bourdillon Esteve (WHO, Geneva) for advice and support, and Richard W.O. Jähnke (GPHF, Frankfurt, Germany) for guidance on Minilab analysis, Michael D. Green (Centers for Disease Control and Prevention [CDC], Atlanta, GA, USA) for analysis of four samples of dihydroartemisinin/piperaquine, Gisela Schneider (Difäm) for constant support, and many other important collaborators who are not mentioned here.
The Difäm-EPN Minilab Survey Group comprised the following persons: Tambo A. Cletus, Ndze Edward Ngah, Fidelis Nyaah, Manyi Pattinora, Stephen Bonnah, Frederick Sowah, Ronald Kimaali, Daisy Isa, Tarak Banerjee, B. Pondo Valentin, Richard Neci, Koshy George, Liza Kamei, Stephen Kigera, Joseph Thuranira and Albert Petersen. Stephen Kigera and Joseph Thuranira were responsible for the pharmacopeial analysis at the WHO-prequalified medicine quality control laboratory of the Mission for Essential Medicines and Supplies in Nairobi, Kenya.
We acknowledge support by Deutsche Forschungsgemeinschaft and Open Access Publishing Fund of University of Tübingen.