Drug resistance of Mycobacterium tuberculosis in Malawi: a cross-sectional survey

Abstract Objective To document the prevalence of multidrug resistance among people newly diagnosed with – and those retreated for – tuberculosis in Malawi. Methods We conducted a nationally representative survey of people with sputum-smear-positive tuberculosis between 2010 and 2011. For all consenting participants, we collected demographic and clinical data, two sputum samples and tested for human immunodeficiency virus (HIV).The samples underwent resistance testing at the Central Reference Laboratory in Lilongwe, Malawi. All Mycobacterium tuberculosis isolates found to be multidrug-resistant were retested for resistance to first-line drugs – and tested for resistance to second-line drugs – at a Supranational Tuberculosis Reference Laboratory in South Africa. Findings Overall, M. tuberculosis was isolated from 1777 (83.8%) of the 2120 smear-positive tuberculosis patients. Multidrug resistance was identified in five (0.4%) of 1196 isolates from new cases and 28 (4.8%) of 581 isolates from people undergoing retreatment. Of the 31 isolates from retreatment cases who had previously failed treatment, nine (29.0%) showed multidrug resistance. Although resistance to second-line drugs was found, no cases of extensive drug-resistant tuberculosis were detected. HIV testing of people from whom M. tuberculosis isolates were obtained showed that 577 (48.2%) of people newly diagnosed and 386 (66.4%) of people undergoing retreatment were positive. Conclusion The prevalence of multidrug resistance among people with smear-positive tuberculosis was low for sub-Saharan Africa – probably reflecting the strength of Malawi’s tuberculosis control programme. The relatively high prevalence of such resistance observed among those with previous treatment failure may highlight a need for a change in the national policy for retreating this subgroup of people with tuberculosis.


Introduction
Although the World Health Organization (WHO) has monitored the emergence of drug resistance of Mycobacterium tuberculosis since 1994, 1 there have been few national surveys of such resistance in sub-Saharan Africa. 2 In 2012, it was estimated that about 1.9% of people newly diagnosed and 9.4% of those undergoing retreatment in Africa had multidrug-resistant (MDR) tuberculosis. 3 The prevalence of MDR tuberculosis in Africa varies between countries 4 and might be generally increasing. 3,5 Over several years, attempts have been made -at the Central Reference Laboratory in Lilongwe -to isolate M. tuberculosis from all smear-positive patients undergoing retreatment in Malawi to investigate drug susceptibility. In 2008, about 8% of people investigated in this manner were found to have MDR tuberculosis (James Mpunga, Malawi National Tuberculosis Control Programme, personal communication, 2008) -although most of the samples came from urban centres and the laboratory's attempts to isolate M. tuberculosisM. tuberculosis often failed. 6 The only published data on MDR tuberculosis in Malawi indicated that just 0.5% of people newly diagnosed with tuberculosis and 0.9% of people being retreated in Karonga district had MDR tuberculosis in 1996-1998. 7 In 2007, the nationally recommended treatment regimen for people newly diagnosed with tuberculosis in Malawi changed. The initial supervised treatment remained the same -i.e. daily isoniazid, rifampicin, pyrazinamide and ethambutol for 2 months -but the unsupervised continuation phase changed from 6 months of isoniazid and ethambutol to 4 months of isoniazid and rifampicin. 8,9 There are four problems since this change that need monitoring. The first is that poor adherence during this currently-recommended continuation phase could lead to the emergence of MDR tuberculosis. Another problem is that nothing is known about the resistance of Malawian isolates of M. tuberculosis to the second-line drugs that began to be used routinely in Malawi in 2007. A third problem is the high prevalence of human immunodeficiency virus (HIV) infection among people with tuberculosis. 10 In 2010, 63% of Malawian tuberculosis patients tested for HIV were found positive. 4 Finally, the national prevalence of drugresistant tuberculosis may be affected by migration of people from neighbouring countries, where such outbreaks have occurred. 11 Given these issues, we conducted a national survey of resistance to anti-tuberculosis drugs in Malawi.
Objective To document the prevalence of multidrug resistance among people newly diagnosed with -and those retreated for -tuberculosis in Malawi. Methods We conducted a nationally representative survey of people with sputum-smear-positive tuberculosis between 2010 and 2011. For all consenting participants, we collected demographic and clinical data, two sputum samples and tested for human immunodeficiency virus (HIV).The samples underwent resistance testing at the Central Reference Laboratory in Lilongwe, Malawi. All Mycobacterium tuberculosis isolates found to be multidrug-resistant were retested for resistance to first-line drugs -and tested for resistance to second-line drugs -at a Supranational Tuberculosis Reference Laboratory in South Africa. Findings Overall, M. tuberculosis was isolated from 1777 (83.8%) of the 2120 smear-positive tuberculosis patients. Multidrug resistance was identified in five (0.4%) of 1196 isolates from new cases and 28 (4.8%) of 581 isolates from people undergoing retreatment. Of the 31 isolates from retreatment cases who had previously failed treatment, nine (29.0%) showed multidrug resistance. Although resistance to second-line drugs was found, no cases of extensive drug-resistant tuberculosis were detected. HIV testing of people from whom M. tuberculosis isolates were obtained showed that 577 (48.2%) of people newly diagnosed and 386 (66.4%) of people undergoing retreatment were positive. Conclusion The prevalence of multidrug resistance among people with smear-positive tuberculosis was low for sub-Saharan Africaprobably reflecting the strength of Malawi's tuberculosis control programme. The relatively high prevalence of such resistance observed among those with previous treatment failure may highlight a need for a change in the national policy for retreating this subgroup of people with tuberculosis.

Study setting and design
We engaged all of Malawi's 48 tuberculosis registration centres to conduct a prospective, cross-sectional survey. The centres were grouped into three zones -northern, central and southern -for phased sample collection.

Data collection and management
Health workers in each registration centre formed a recruitment team and attended a three-day training course about the survey protocol. They subsequently collected data on each consenting smear-positive tuberculosis patient, including the patient's age, sex, level of education, occupation, marital status and HIV status -if known -and details of any previous tuberculosis treatment. After each patient was asked if they had received tuberculosis treatment, the patient's medical records at the health facility of recruitment were checked for evidence of such treatment.
Following national policy in Malawi, 8

Participants and case definitions
Using the definitions recommended by WHO, 12 new cases were defined as people who had never been treated for tuberculosis -or had previously received anti-tuberculosis medications for less than one month -and retreatment cases were defined as those who had previously received tuberculosis treatment for at least one month. Retreatment cases were grouped according to the outcome of previous treatment: cured, completed, defaulted or failed. A patient was defined as cured when the person was smear-negative at, or one month before, treatment completion and on at least one previous occasion. A completed treatment was defined as a patient who completed treatment but without smear microscopy proof of cure. Persons who had treatment interruption for two consecutive months or more were grouped as defaulted. Those who remained smear-positive when tested five or six months after initiation of their previous treatment were defined as treatment failures.
For our survey, sputum samples were collected from each newly-diagnosed person with sputum-smearpositive tuberculosis seen at a registration centre in the northern, central and southern zones in May-July 2010, August-October 2010 and November 2010-January 2011, respectively. Sputum samples were also collected from each person with smear-positive tuberculosis undergoing retreatment at any registration centre between February 2010 and March 2011.

Drug resistance definition
Isolates of M. tuberculosis were defined as MDR if they were at least resistant to isoniazid and rifampicin, and extremely drug resistant (XDR) if they were also resistant to an injectable drug and a quinolone of the second-line medications.

Sample size projections
Assuming that 1.8% and 20% of the people newly diagnosed would have MDR tuberculosis and be lost to follow-up, respectively, we estimated that we needed to enrol 1260 new cases to estimate the prevalence of MDR tuberculosis among such cases with a precision of ± 1%. Similarly, assuming that 5.0% and 20% of our retreatment sample would have MDR tuberculosis and be lost to follow-up, respectively, we estimated that we would have to enrol 770 people undergoing retreatment to estimate the prevalence of MDR tuberculosis with a precision of ± 2.0%.

Laboratory procedures
Prior to enrolment, each participant had been found positive for tuberculosis by the microscopic examination of three smears of sputum. 12,13 Each month, a random selection of sputum smears from the registration centres -five from each health centre and 25 from each district hospital -was re-examined by a visiting laboratory supervisor. Concordance between the registration centres' results and the supervisor's remained above 96% during our survey.
For our survey, two additional sputum samples were collected -under supervision and approximately one hour apart -from each enrolled patient and stored at 2-8 °C in the registration centre. Efforts were made to ensure that these samples were collected before antituberculosis treatment was commenced. The samples were transported to the Central Reference Laboratory, in cooler boxes, by bus or in a district health vehicle or study team vehicle.
Once a sample had reached the laboratory, it was decontaminated and further homogenized. 14 Part of the pellet produced by centrifuging the sample was smeared, stained with auramine phenol stain and then checked for acidfast bacilli. Another part was inoculated into two tubes of Lowenstein-Jensen medium -one containing glycerol and the other containing sodium pyruvatewhich were examined for growth weekly for up to 8 weeks. Each contaminated culture was discarded and replaced with a new culture that was set up using another part of the relevant pellet -which had been kept in a refrigerator. The Capilia tuberculosis test 15 was used to identify isolates belonging to the M. tuberculosis complex. Indirect susceptibility testing to isoniazid, rifampicin, ethambutol and streptomycin was performed, on one isolate per participant, using the proportion method on Lowenstein-Jenson medium. 16 All isolates defined as MDR tuberculosis were sent to the South African Medical Research Council's Supranational Reference Laboratory in Pretoria. There, they were retested for their susceptibility to first-line drugsusing a line probe assay and automated liquid culture 17,18 -and tested for their susceptibility to the second-line drugs amikacin, kanamycin, capreomycin, ofloxacin and ethionamide -using automated liquid culture.

Statistical analysis
For our final analysis, we excluded those cases from which M. tuberculosis was not isolated in culture. Categorical and

Results
During the study period, 2120 smearpositive individuals consented to participate. Five were excluded as their baseline data were missing, another 1347 were classified as newly diagnosed with tuberculosis and the remaining 768 were classified as retreatment cases (Fig. 1 Compared with the new cases, people undergoing retreatment were more frequently found to be culture-negative or to be culture-positive for mycobacteria other than M. tuberculosis.
Of 86 treatment failures, 31 samples were culture-positive for M. tuberculosis, six were culture-positive for other mycobacteria and 49 were culture-negative.
The median transit time of all samples, from collection to arrival at the Central Reference Laboratory was 4 days (interquartile range, IQR: 2-7 days). Transit time had no apparent effect on the probability that a sample would be found culture-positive for M. tuberculosis (P = 0.71).

Culture-positive tuberculosis
Culture-positive individuals in both new and retreatment groups were similar in terms of their sociodemographic characteristics (Table 1).
Overall, 66.4% (386) of the retreatment cases and 48.2% (577) of the new cases were known or found to be infected with HIV, demonstrating a significantly higher HIV prevalence among people retreated (P < 0.01). The retreatment cases reported that they had received tuberculosis treatment between 1978 and 2010 with a median of 2.4 years (IQR: 1.1-5.9 years) before their enrolment. Just 31 (5.3%) of the culture-positive retreatment cases had failed their previous treatment (Table 1).
Among the 1196 M. tuberculosis isolates from new cases, ethambutol, isoniazid, rifampicin and streptomycin resistance was present in 0.5%, 3.2%, 0.8% and 4.2%, respectively ( Table 2). The corresponding values for the 581 isolates from the retreatment cases were all higher (Table 2).
Five (0.4%) of the 1196 new cases had MDR tuberculosis (Table 2 and Fig. 2). Other types of resistance (monoresistance or any combination of drug resistance excluding MDR tuberculosis) were identified in 75 (6.3%) of the new cases but the remaining 1116 (93.3%) M. tuberculosis isolates from new cases were found to be sensitive to all four first-line drugs. Twenty-eight (4.8%) of the 581 M. tuberculosis isolates from retreatment cases showed multidrug resistance ( Table 2). Other types of resistance were identified in 83 (14.3%) of the retreatment cases but the remaining 470 (80.9%) M. tuberculosis isolates from retreatment cases were found to be sensitive to all four first-line drugs ( Table 2 and Fig. 2).
In the multivariate analysis, sex, age and HIV status were not found to be significantly associated with MDR tuberculosis among new or retreatment cases. There was also no evidence of a significant association between region of residence and MDR tuberculosis. All of the 28 retreatment cases with MDR tuberculosis had received treatment in the previous five years -23 (82%) within the previous two years. MDR tuberculosis in people undergoing retreatment was found to be significantly and inversely associated with time since previous treatment (adjusted IRR: 0.7, 95% confidence interval, CI: 0.5-0.9). Previous treatment failure -but no other previous treatment outcome -was strongly associated with MDR tuberculosis (adjusted IRR: 3.7, 95% CI: 1.6-8.4). Of the 31 treatment failures, nine (29.0%) cultured multi-drug resistant M. tuberculosis.
Of the 33 isolates of M. tuberculosis found to be multidrug-resistant in Malawi, 30 successfully underwent retesting in South Africa, and 11 of these were sensitive to either isoniazid or rifampicin or both of these drugs. If the results from South Africa are used as the gold standard, this indicates a 36.7% false-positive rate (11/30 in Table 3). When a random sample of 106 isolates of M. tuberculosis found not to be multidrug-resistant in Malawi were retested in South Africa, one was identified as MDR tuberculosis -giving a 0.9% false-negative rate (1/106 in Table 3).
The 20 isolates found to show multidrug resistance in South Africa were re-cultured in South Africa and tested for resistance to several second-line drugs. Although 18 of these isolates were successfully re-cultured and tested, none showed extensive drug resistance (Table 4).

Discussion
This is the first national survey of anti-tuberculosis drug resistance done in Malawi. We found the prevalence of MDR tuberculosis among people a Cured defined as a smear-positive patient who was smear-negative at, or one month before, treatment completion and on at least one previous occasion. 12 b Treatment completed defined as a patient who completed treatment but without smear microscopy proof of cure. 12 c Defaulted defined as treatment interruption for two consecutive months or more. 12 d Failed defined as remaining smear-positive when tested five or six months after initiation of previous treatment. 12 e Smear scores indicate the density of acid-fast bacilli seen on a sputum smear. 13 Note: Data are missing for some characteristics. The sum of the percentages for some characteristics may not equal 100 due to rounding. times recorded during our survey were disappointing, long transit times were not associated with isolation failures. Mycobacteria could not be grown from 49 of 86 samples from treatment failures, probably because the bacilli in the 49 samples were dead. Mycobacteria other than M. tuberculosis were cultured from six treatment failures. The proportion of sputum samples from retreatment cases that were found culture-positive for M. tuberculosis was significantly lower than the corresponding value for the new cases. This difference is partly explained by (i) the low isolation rate from treatment failures; (ii) the fact that samples from retreatment cases were relatively more likely to grow mycobacteria other than M. tuberculosis; and (iii) the fact that sputum samples from retreatment cases are relatively more likely to be collected from patients who have already begun treatment for their current episode of tuberculosis.
Since the results recorded by Malawi's Central Reference Laboratory were associated with a 36.7% false-positivity rate and a 0.9% false-negativity rate, the prevalences of MDR tuberculosis that we recorded in Malawi -although low -could overestimate the true values. Given the laboratory's limited capacity and the observation that resistance patterns probably do not vary between smear-positive and smear-negative cases of tuberculosis, 20 we did not investigate the drug resistance of any M. tuberculosis isolates from smear-negative tuberculosis patients.
We found HIV prevalence among new smear-positive cases of pulmonary tuberculosis to be 48.2%. The HIV prevalences reported among all tuberculosis cases by Malawi's National Tuberculosis Programme in 2010 and 2011 were higher, at 63% and 60%, respectively. 21 The programme's observations indicate that HIV prevalence among smearnegative cases of pulmonary tuberculosis exceeded 65% in 2010-2011. By focusing on smear-positive cases, we probably limited the extent to which we could explore associations between HIV and MDR tuberculosis. Although we found no association between HIV and MDR tuberculosis, it is possible that such an association exists in the overall population of people with tuberculosis. The existence of such a link remains a matter of controversy 3,5,22,23 but concomitant HIV infection certainly poses some unique challenges in the management of tuberculosis. 10 Although we collected samples from different areas of Malawi at different times of the year, a retrospective analysis of new tuberculosis case notifications between 1999 and 2007 suggested that there was little variation in the number of new cases occurring in each quarter of the year (James Mpunga, Malawi National Tuberculosis Control Programme, personal communication, 2010).
The low prevalence of MDR tuberculosis that we recorded may be attributable to the success of Malawi's tuberculosis control programme. The frequencies of success in the treatment of tuberculosis in Malawi -88% for new cases and 85% for retreatment cases -are among the highest recorded in sub-Saharan Africa. 4 We recorded higher prevalences of streptomycin resistance than of rifampicin or isoniazid resistance, perhaps because streptomycin was included in the recommended first-line treatment for tuberculosis in Malawi until 1992.  We detected no XDR tuberculosis but did observe some resistance to second-line drugs. Since all of our isolates tested for resistance to ethionamide were found positive, the currently recommended 24-month regimen for the treatment of MDR tuberculosis in Malawi needs to be revised. Resistance to the second-line injectables was detected but not resistance to ofloxacin. At the time of the survey, Malawi's Central Reference Laboratory relied entirely upon the South African Supranational Reference Laboratory for the identification of Malawian cases of XDR tuberculosis. 8 Treatment failure -frequently a forewarning for the development of drug-resistant tuberculosis 24 -was associated with a 29.0% risk of MDR tuberculosis in our survey. The initiation of a standard retreatment regimen while awaiting the results of drug susceptibility testing may amplify resistance in cases with pre-existing MDR tuberculosis. 25,26 Although use of an empirical MDR treatment regimen has been suggested as a replacement for the standard retreatment regimen for all treatment failures, 24,27,28 such a change in Malawi would expose most treatment failures -i.e. those who do not have MDR tuberculosis -to a more toxic and less effective therapy. During our survey, all patients with MDR tuberculosis who were diagnosed by phenotypic testing at the Central Reference Laboratory -including the 11 cases classified as drug sensitive when their sputum samples were investigated in South Africa -were managed with the nationally recommended secondline regimen. This was because (i) the phenotypic results were seen as more predictive of clinical response; (ii) the South African results became available several months after the patients had started second-line therapy; and (iii) it was felt that any changes to treatment made after the South African results became available would be confusing to patients.