Estimating the prevalence of poor-quality anti-TB medicines: a neglected risk for global TB control and resistance

Search strategy

A systematic review was conducted of scientific and lay reports on anti-TB medicine quality, using Web of Science, Medline, Pubmed, Scopus, Google, Google Scholar, WHO, US Pharmacopeia (USP), and Medicines Regulatory Agencies (MRA) websites.

Search included any published report, without an inclusion start date, up to 31 October 2021 in English, French and Spanish. Search terms included the terms ‘falsified’, ‘substandard’, ‘counterfeit’ and other commonly used terms such as ‘fake’ or ‘spurious’ or ‘quality’, together with ‘anti-TB’ or ‘TB medicines’ (online supplemental table 1). Manual searches of the reference lists in the included articles were also performed.

Eligibility criteria

The inclusion criteria were of any study describing in vivo or in vitro tests to determine anti-TB medicine quality, assays to determine quality, articles about seizures, recalls and confiscations of anti-TB medicines; case reports or articles describing adverse events or patients not responding to anti-TB treatments where quality was questioned.

We included studies that surveyed the quality of TB drugs in one or more locations (hereafter ‘quality surveys’), that compared the pharmaceutical equivalence of different brands of a given TB drug (hereafter ‘bioequivalence studies’), studies describing techniques to determine the quality of TB drugs (hereafter ‘laboratory assembled collections’), studies aiming to predict the stability profile of a medicine (hereafter ‘accelerated stability assays’) and recalls, seizures or case reports by national medicines regulatory authorities (NMRA) (hereafter ‘case reports/NMRA seizures’).

Studies with results from several countries or locations are included under each specific country/location. A ‘data point’ is defined as a specific location where medicines were collected for quality analysis, at a given time during a given study.

No time restrictions were applied for the systematic search of the literature.

General discussions over sampling methodology and pharmaceutical legislation (eg, on the regulatory framework surrounding SF TB treatment) and reviews of the literature on various aspects of TB medicines were also included in the database for reference but are not included in the analysis of this review.

The exclusion criteria were studies with results for a whole region or a whole class of TB drugs, without specific country or location data. We also did not include reports for ciprofloxacin and co-amoxiclav as, although there are many reports of SF products, they are included in the parallel review of the quality of antibiotics.39 We did not include reports of quackery and use of SF TB medicines without an objective evidence base of efficacy. Neither the quality of BCG vaccines or diagnostic tests for TB were included.

Key variables and definitions

Anti-TB medicine quality failure rates were extracted into a database from each report, including data on packaging, content of API, disintegration, dissolution, microbiology and accelerated stability results when available. Poor quality medicines classified as falsified, substandard or degraded by the reports’ author’s definitions were tabulated and interpreted in relation to current WHO definitions.40 Samples that failed chemical assays, but without detection of wrong active ingredients and without packaging analysis, are classified as substandard or falsified (SorF) and not as either falsified or substandard as this distinction cannot be reliably made without reference to the packaging or regulatory status of the sample.41 However, samples that contained the wrong API or no API but without packaging analysis were assumed to be falsified. There is a small risk of misclassification of such samples, as falsified when they are actually substandard, due to gross manufacturing errors.42 43 Samples that did not fail chemical and packaging tests (when these were done) are considered as good quality.

Data collection

Information within each publication was manually extracted and entered in a database constructed using MS Access 2010, developed in collaboration with the informatics team at IDDO (https://www.iddo.org/mq-scientific-literature-surveyor). Information extracted included: publication type (eg, report, original research article), year of publication, publisher, sampling type, location (country and city, where available) and type of outlet where samples were collected, total number of samples collected, API/API combination name, number of samples failing medicine quality test(s) and the techniques used to analyse samples.

Analysis and reporting

Data were extracted and statistical analyses were conducted in Microsoft Excel 365 and Stata (V.11.2, Stata Corp, College Station, Texas, USA). This review follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.

Risk of bias assessment

Since there are no standardised methods to assess equivalence, analysis technique, lay press and case reports publications, their risk of bias was not addressed. Publication bias should be taken into consideration.

Patient and public involvement

Patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of our research.

Ethics approval

No ethical approval was sought for this study and no patient identifiable data were used.

Types and timing of publications

A total of 530 publications were screened by title and abstract, 507 gathered through electronic searches and 23 additional publications identified through manual reference screening and other sources such as lay press, NGOs, WHO and US Pharmacopeia Medicines Quality Database reports. See figure 1, PRISMA statement and flow chart of papers in the online supplemental material.

Request permission

After removal of duplicate reports and articles excluded for eligibility, 278 full-text publications were reviewed and 162 were included. Other articles were excluded as they were describing surveys, techniques to determine the quality of drugs and seizures of other classes of medicines (n=116).

A third of the articles were primary research peer-reviewed papers (55, 33.9%), followed by peer-reviewed reviews or compilations of articles (40, 24.7%) and those describing laboratory assembled collections for TB medicines (24, 14.8%). Reports from international organisations accounted for 19 (11.7%) of the publications. Other types of articles included short communications (12, 7.4%), lay press articles (9, 5.5%), two non-peer-reviewed articles (1.23%) and one public alert (0.6%) (figure 1). No PhD theses about TB medicine quality problems were found.

The first publication we found reporting poor-quality TB medicines was dated from 1906,44 describing fake anti-TB nostrums, with no other publications found until 1968.45 46

Most of the publications were published since 1999 (154, 95.1%) with 17 (10.5%) reports being published in the year 2013. The most recent papers were from 2020 (2, 1.2%) (figure 2). Although reports were published in 70 different journals, 26 (16.5%) were published in the International Journal of TB and Lung Disease (https://www.theunion.org/what-we-do/journals/ijtld).

Request permission

Of the 162 publications included in the database, 65 (40.1%) described one or more TB quality surveys in a specific location or region with enough information to yield an estimate of the local prevalence of poor-quality TB medicines; 231 records/data points of medicine samples identified in these 65 publications were included in the analysis.

Of the 65 publications, 7 publications were removed from the analysis as they included an unspecific number of samples; 6 had unknown numbers of samples collected47–52 and 1 publication reported the diversion of 2 million doses that were not analysed,53 leaving a total of 58 publications included in the quantitative analysis. Publications were then categorised by sampling methodology and study type. Sampling categories were not mutually exclusive. As a few studies were included in multiple sampling categories, the number of publications adds to more than 100% (n=67).

Half of the reports described quality surveys (33, 49.2%), followed by bioequivalence studies (17, 25.3%). Ten (14.9%) studies described accelerated biostability assays. Other studies included in the quantitative analysis were five laboratory assembled collection reports and two case reports/NMRA seizures (7.5% and 2.9%, respectively).

All publications have been mapped on the Infectious Diseases Data Observatory Medicine Quality Surveyor system and data are downloadable here38 (https://www.iddo.org/mqsurveyor/%23anti-TB).

Type of medicines and geographical data

Medicines collected included 17 different APIs and 7 different coformulations. The most frequently assessed were rifampicin monotherapy in 45 studies (19.5%) and isoniazid monotherapy in 33 (14.3%), rifampicin-isoniazid-pyrazinamide-ethambutol FDC in 28 studies (12.1%) and rifampicin-isoniazid in 20 (8.6%). The median (IQR) number of samples collected per study was 12 (1–478). The content of the APIs was measured in 52 (80.0%) studies.

In total, 7682 samples were collected in 22 countries. Of those, 6255 were from quality surveys in 17 countries and 1086 samples were from equivalence studies in 6 countries.

The majority of data points were in India with 92 (41.4%), 20 (9.0%) in Kenya and 16 (7.2%) in Kazakhstan (table 1). The country where samples were collected was not specified for 985 (12.8%) samples and 11 (4.9%) data points (table 1).

Of the 7682 samples collected, 1170 (15.2%) failed at least one quality test, 6.7% (78) were classified as falsified, 22.6% (265) as substandard and 70.7% (827) as SorF since packaging analysis was not performed (table 2 and online supplemental table 2).

In quality surveys 879 (14.1%) samples failed at least one quality test out of 6255 collected, 136 (12.5%) samples out of 1086 in bioequivalence studies and 87 (36.9%) samples of 236 in accelerated biostability studies.

The highest proportion of substandard anti-TB medicines was observed in the Americas, although it only included one study of the bioavailability of co-formulated rifampicin-isoniazid-pyrazinamide medicines in Mexico and one accelerated biostability study of doripenem, imipenem-cilastatin and meropenem in the USA. Asia had the largest number of samples collected, 5758 samples in 11 countries with an overall failure rate of 15.2%. Failure rate ranged from 7.7% in Myanmar to 33.5% in Vietnam and 35.2% in Indonesia. Total samples collected in Africa accounted for 749 in five countries, with a 13.1% failure rate; Nigeria had the highest failure rate of 45.8%.

Thirteen publications specified if the failing samples had lower or higher amounts of the active substance in relation to predefined cut-offs. For rifampicin, 132 samples (30.2%) had lower amounts of API, followed by cycloserine (54, 48.6%), pyrazinamide (25, 38.5%), isoniazid (16, 11.3%), levofloxacin (9, 32.1%), ethambutol (7, 6.8%) and kanamycin (4, 100%). Higher amounts of API were found in 20 (4.6%) rifampicin samples, 11 (10.7%) ethambutol, 7 (4.9%) isoniazid and 5 (7.7%) pyrazinamide samples.

High performance liquid chromatography (HPLC) was conducted in 27 (51.9 %) studies, either alone or coupled with UV-spectrophotometry, calorimetry (DSC) or Fourier transform infrared spectroscopy. Three bioavailability studies conducted HPLC and measured drug levels (or its metabolites) in urine rather than serum.

Minilab thin layer chromatography (TLC) was performed in 7 (10.8%), either alone or coupled with an additional technique such as Raman, HPLC or LC with UV spectrophotometry.

Four studies (6.1%) used ultraviolet spectrophotometry alone to determine the content of the API.

Other methods used to analyse the quality of the medicines included: fluorimetry, direct analysis in real time coupled to a compact single quadrupole mass spectrometer (DART-QDa-MS-DART-QTOF), an ultra-fast UPLC-UV method and liquid chromatography, coupled with mass spectrometry (LC-MS). Ten studies (15.4%) did not specify the tests conducted to determine the API content.

Dissolution testing was included in 25 studies (38.5%), and of those studies only 5 included disintegration testing.

Seventeen (26.2%) studies included physical inspection and packaging inspection was specifically mentioned in only seven studies (10.8%).

A third of the publications (25, 38.5%) specified the cut-off followed to assess if the medicine met quality standards. Five publications mentioned using an 85%–115% API cut-off, four used a <80% limit, four publications 90%–110% and another two a 95%–105% cut-off. Ten (15.4%) reports specified using pharmacopoeial methods, three of them mentioned using the British Pharmacopoeia, another three the USP and one using the Indian Pharmacopoeia. The other 30 (46.1%) reports did not specify which pharmacopoeial method was followed.

Quality surveys

Of the 33 publications describing surveys in 17 countries, 28 (84.8%) were convenience surveys and 5 (15.2%) used random sample location selection. Surveys accounted for 141 (61.0%) individual medicine samples of the total 231 data points.

India yielded the largest number of reports with 14 (34.1%), followed by Kenya (5, 12.2%) and Indonesia, Kazakhstan, Nigeria and Vietnam with 2 (4.8%) reports available for each country. Other countries that had one (2.4%) publication describing anti-TB medicine quality surveys included: Armenia, Azerbaijan, Botswana, Cambodia, Mexico, Myanmar, Republic of Belarus, Thailand, Ukraine, Uzbekistan and Zambia. Three surveys (9.5%) did not state the country of sample collection (table 1).

A third of surveys (10, 29.4%) stated that they sampled only private outlets, 20.6% (7) specified that samples were collected in public hospitals and 26.5% (9) reports did not specify which results referred to private or public outlets. Three (8.8%) surveys obtained the samples from the distribution agent or wholesale pharmacy, two (5.9%) from the manufacturer and one (2.9%) report sourced the medicines from WHO and another one from a private clinic. The type of outlets from which medicines were sampled was not specified in one survey report (2.9%).

The highest reported failure rate was for streptomycin (54.5%), followed by co-formulated isoniazid-ethambutol (40.0%) and co-formulated rifampicin-isoniazid-ethambutol (37.7%). Medicines with higher failure rates also included cycloserine (36.0%), amikacin (25.4%) and ofloxacin (21.4%). Levofloxacin, rifampicin and ethambutol monotherapy had similar failure rates (16.8, 16.6 and 16.1%, respectively). Only meropenem, moxifloxacin, norfloxacin, prothionamide and co-formulated rifampicin-isoniazid-pyrazinamide-ethambutol-pyridoxine had zero failure rates (table 2).

We found no data points for bedaquiline, clofazimine, delamanid, linezolid, p-aminosalycilic acid, pretomanid and terizidone.

Bioequivalence studies

Seventeen (25.4%) publications described bioequivalence studies including 39 (16.9%) individual medicine data points in the database.

Twelve studies described bioavailability of anti-TB medicines in vivo, three in vitro alone and two studies included both in vivo and in vitro methodology. Four of the bioavailability studies compared analysis of urinary excretion data versus plasma concentration time profiles.

Of the bioavailability studies seven described the API content, and one included accelerated stability assays.

Bioavailability data points included 19 from India (48.7%), followed by 13 (33.3%) from South Africa, 2 (5.1%) from Indonesia, and Mexico and Nigeria had 1 (2.6%) each. The UK had three data points (7.7%) (table 1).

Of the 1086 samples analysed, 136 (12.5%) failed the bioavailability tests conducted. Of the failing samples, 100 (73.5%) were rifampicin monotherapy tablets and 53 of those were samples classified as substandard with poor bioavailability collected in South Africa. Other poor-quality samples included 21 samples of rifampicin-isoniazid-pyrazinamide FDC, 8 samples of rifampicin-isoniazid FDC, 5 ofloxacin samples and 2 rifampicin-isoniazid-pyrazinamide-ethambutol FDC samples (table 2).

Bioavailability studies included samples obtained from the manufacturing company (18, 46.2%) or a distributor (13, 33.3%). Samples collected from pharmacies, either public hospitals or private pharmacies and clinics accounted for 10.3% (4). Four data points (10.3%) did not specify sample origin.

Accelerated bioavailability studies

Ten publications described the stability of anti-TB medicines under accelerated conditions, with eight in India, one in Thailand and one in the USA.

All except two studies were conducted between 2003 and 2005. One study conducted in 196845 studied the stability of cycloserine (149 samples) and one in 2011 studied the stability of doripenem, imipenem and meropenem.54

The overall failure rate was 36.9%, varying from 77.8% failure rate for both rifampicin-isoniazid-ethambutol FDC and isoniazid-ethambutol FDC, respectively, 66.7% for rifampicin-isoniazid-pyrazinamide FDC, and 50% for both ethambutol in single form and rifampicin-isoniazid FDC. Rifampicin-isoniazid-pyrazinamide-ethambutol FDC had a similar failure rate to cycloserine (39.1% and 38.3%, respectively). Other medicines that had zero failure rate included doripenem, meropenem, co-formulated imipenen-cilastatin, rifampicin, isoniazid, and pyrazinamide monotherapy.

Research on methods assessing TB drug quality

Although 25 publications were found describing different techniques for analysis of anti-TB monotherapy and FDC tablets and capsules, only 9 were eligible to be included in the quantitative analysis, as others were validation methods and techniques with no specific medicine quality data included (online supplemental table 3).

Four publications described in vitro methodology for assessing the bioavailability and bioequivalence of TB drugs, and the effect of food on rifampicin. One publication used a calorimetric method to screen for the presence of the less active diastereomeric (R,S) forms of ethambutol dihydrochloride55 and one described a chemiluminescent method for pyrazinamide.56 Seven publications described chromatographic techniques including rapid screening of pharmaceuticals by TLC. One publication described the colorimetric analysis of 14 anti-TB FDCs.57

One publication described a disintegration method for isoniazid58 and another three described dissolution assays. Two articles described paper analytical devices for fast field screening of antimicrobial medicines and anti-TB pharmaceuticals.59 60 Five reports described spectroscopic techniques from low-cost optical spectrometers61 to near-infrared spectroscopy and multivariate calibration to measure the dissolution62 (online supplemental table 3).

Case reports

A 1906 publication from The New York Times described fake TB ‘nostrums’.44 Three reports described the presence of TB medicines as cryptic wrong ingredients in falsified medicines.63–65 Two publications described the analysis of 192 blister packs labelled as antimalarial artemisinin-based combination therapy seized in sub-Saharan Africa.63 64 Sixty-five (33.8%) of those samples contained either ciprofloxacin alone or a combination of chloramphenicol, ciprofloxacin, sildenafil and gatifloxacin (table 3).

One lay press report described the discovery of vials labelled as ceftazidime (Fortum) injection reused from hospital waste and adulterated with streptomycin.65

In a reported 2009 recall from the Government of Canada, bottles of the medicine Rofact (rifampin), were erroneously filled with clonazepam 0.5 mg.52

Other recalls/case reports/NMRA seizures included a lay press report describing the diversion of TB medicines financed by The Global Fund sold in local markets in Côte d’Ivoire.53 One publication reporting ‘over concentration’ of TB drugs in Chad (unpublished results from Réseau Médicaments et Developpement).48 Antib-4 (pyrazinamide-ethambutol-isoniazid- rifampicin) and Ebsar (isoniazid-rifampicin) manufactured in India by Rusan Pharma were recalled in South Africa because of the presence of lower active ingredients than specified.49 A report from the South-East Asian FIP-WHO Forum of Pharmaceutical Associations described falsified Reinex 450 and Clombutol (anti-TB medicines) in India in 2003.66 Manufacturing of TB medicines was suspended at Svizera Labs, a major supplier to developing countries, in Mumbai, India, because of unreliable manufacturing standards67 (table 3).