Review: Multidrug-resistant tuberculosis: public health challenges


Dr Richard J. Coker, ECOHOST, Department of Public Health and Policy, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK. Tel.: +44-20-7927 2926; Fax: +44-20-7612 7812; E-mail:

The history of communicable diseases, and in particular tuberculosis, and efforts to control them have long focused policy-makers’ attention, provoked media alarm, and caused public anxiety (Rosenkrantz 1994; Rothman 1995; Coker 2000). Communicable disease control experts frequently tread a fine line. They sometimes have to balance the costs and benefits of provoking anxiety amongst the populace when they highlight public health concerns and counterbalance this with encouraging policy-makers to focus attention on issues they feel may demand urgent attention. Over the past decade or so the challenges and uncertainties associated with the spectre of multidrug-resistant tuberculosis (MDRTB) has concerned those charged with protecting public health at the highest national and international policy-making tables.

In the early 1990s the World Health Organization (WHO) drew public attention to the inadequacies of global tuberculosis control efforts when it described the situation as an ‘emergency’. Also through the 1990s attention started to focus on the challenge posed by strains of tuberculosis resistant to conventional treatment with first-line drugs. In the early 1990s New York City witnessed a substantial and highly publicized epidemic of MDRTB and from 1993 considerable skill, effort and resources were involved in beginning to control it. In 1998, three international organisations, Doctors without Borders (MSF), Medical Emergency Relief International (MERLIN), and the Public Health Research Institute (PHRI) invoked the name of another feared pathogen when they described the spread of multidrug-resistant tuberculosis (MDRTB) in Russia: ‘Ebola with wings’ (Voelker 1998). This captured an image that was also taken up by the World Health Organization (Coghlan 2001), and subsequently built upon in a 1999 report that highlighted both the global nature of the epidemic and the potential for international spread (Open Society Institute 1999). With approximately 2 million people crossing international borders every day and mass movements of people occurring because of economic woes or conflict transnational movements of diseases, including MDRTB may, it is argued, threaten us all (Brundtland 2000).

MDRTB, that is tuberculosis resulting from organisms that are resistant to at least isoniazid and rifampicin, develops through selection pressures. The bacterial population in cavitary pulmonary disease may be between 107 and 109 organisms, and spontaneous mutations leading to drug resistance occur with a frequency of one in 106 to 108 replications depending on the drug (Vareldzis et al. 1994). In any cavitary population there are likely to be, therefore, a few organisms resistant to single anti-tuberculosis drugs. Thus, exposure to one drug only lends a small number of organisms a selective advantage. However, subsequent exposure to a second drug after a period of single drug use will, again, selectively advantage those organisms resistant to the first drug and now also the second drug – hence the development of MDRTB. Acquired drug resistance, i.e. resistance developing in a patient who has received or is receiving treatment, suggests contemporary programme weaknesses. Primary drug resistance occurs when a patient who has previously not received treatment develops disease with an organism that is already resistant, an indication of past programmatic frailties.

Recently the traditional terms ‘acquired drug resistance’ and ‘primary drug resistance’ have been superseded by the terms ‘drug resistance among new cases’ as a proxy of primary resistance, and ‘drug resistance among previously treated cases’ as a proxy of acquired resistance (WHO 2000). The reason for this is that patients may not disclose prior treatment for tuberculosis (which would have led to the term ‘primary drug resistance’ being erroneously used). Alternatively, patients may fail treatment because their strain was resistant from the start and not because it acquired resistance as a consequence of treatment (which would have led to the term ‘acquired drug resistance’ being applied inappropriately). The purpose of this paper is to describe some of the public health challenges, many of which are formidable, facing those involved in countering the threat posed by MDRTB.

Global epidemiology

Much of our understanding of the global epidemiology of MDRTB is based upon data collated from surveys and national surveillance programmes that have been analysed by WHO and IUATLD in collaboration (WHO 2000). Whilst the prevalence of MDRTB in most Western European countries is low, in the order of 1% of all cases of tuberculosis, in some countries, notably in Central and Eastern Europe, high levels have been described. The global picture of MDRTB remains somewhat opaque, however, because of the paucity of data on MDRTB from much of the world including much of Europe and Asia. In Russia, for example, data from only two of 89 regions (or oblasts) informed understanding in this global survey (Figure 1). There remain sizeable gaps in our understanding of the distribution of cases globally, detailed systematic determinations of temporal trends remain patchy, and our comprehension of the magnitude of the global burden of disease because of MDRTB is uncertain. These gaps notwithstanding, estimates of the global burden have been made. One estimate based on data from 64 countries was that the annual incidence of MDRTB was 273 000 cases – a fraction of the estimates of the 8 million annual cases of tuberculosis worldwide (Dye et al. 1999). Projections of future incidence using mathematical modelling suggest that current annual incidence rates may climb and that concerted efforts to control MDRTB will be required and make take years if not decades if rates are to decline (Dye et al. 2002).

Figure 1.

Prevalence of MDRTB among new tuberculosis cases in countries and regions surveyed between 1994 and 1999 (Source: WHO 2000).


Challenges to prevention of MDRTB include problems faced in preventing the development of MDRTB in patients with fully drug-sensitive disease and prevention of transmission of established MDRTB disease to others who then go on to develop MDRTB. Fundamentally, MDRTB is a man-made problem and results principally from clinical mismanagement and programme frailties. Factors in the genesis of MDRTB include inadequate initial treatment regimens, adding single drugs to failing regimens, poor treatment adherence and lack of or intermittent supply of standardized short-course chemotherapy. The prevalence of MDRTB increases 10-fold after unsuccessful treatment of new tuberculosis cases (Pablos-Mendez et al. 1998).

A central tenet of tuberculosis control is that treatment is prevention (Crofton 1962). MDRTB is best prevented through robust programmes of tuberculosis control that ensure appropriate treatment with first-line drugs with drug supplies that are assured and of high quality (Crofton 1987). Evidence that short-course chemotherapy with first-line drugs as part of a good control strategy effectively prevents the emergence of drug resistance is suggested by data from the Global Project on Anti-tuberculosis Drug Resistance Surveillance (Raviglione 2000; WHO 2000). This survey compared levels of MDRTB in countries with ‘better’ tuberculosis control achieved by the WHO Directly Observed Treatment, Short-course strategy (DOTS) with results from countries that had ‘poorer’ control programmes. In countries with ‘better’ programmes, the acquired MDRTB index (or, more accurately, the MDRTB index in previously treated cases, i.e. the number MDRTB cases in previously treated cases divided by total number of tuberculosis cases) was significantly lower than in those countries with ‘poorer’ programmes (0.6%vs. 1.8% respectively).

For drug resistance among new cases, the picture from this study was slightly more confusing (Raviglione 2000; WHO 2000).The prevalence of MDRTB among new cases was similar in the two groups of countries. But when MDRTB among new cases was correlated with countries’ rates of treatment success there was a significant association. That is, countries with high levels of treatment success tended to have low MDRTB rates among new cases. Given that MDRTB among new cases is an indicator of historically weak tuberculosis control programmes, the most likely explanation is that robust tuberculosis programmes prevent, in the short term, acquired drug resistance, but that preventing drug resistance among new cases demands sustained efforts.

Moreover, whilst the programmatic approaches to MDRTB usually focus upon outcomes that are individualized and short term (such as cure rates and survival) from a population perspective the critical issue is transmission and which approaches (standardized or individualized second-line regimens, for example) most favourably impact upon transmission dynamics. In some countries, such as the US and Peru, sustained commitment to addressing MDRTB now seems to be resulting in public health benefits.

Prevention of spread of MDRTB, particularly in institutions such as prisons, homeless shelters, and hospitals has been well documented and is a particular concern when many highly vulnerable individuals, for example those infected with human immunodeficiency virus (HIV), congregate (Coronado et al. 1993). Prevention of nosocomial spread can be achieved through a combination of interventions including ensuring adequate ventilation, ultraviolet germicidal irradiation, and masks, respirators and filtration devices. The relative contribution of these interventions or combinations of interventions remains somewhat uncertain however. Moreover, in settings such as the former Soviet Union, where MDRTB rates are high (Farmer et al. 1998; Drobniewski et al. 2002a), care is institutionalized in hospitals and prisons (Coker et al. in press), and as an explosive HIV epidemic unfolds, the challenges faced in combating nosocomial spread are both daunting and demand urgent attention (UNAIDS 2002).


Why is MDRTB a feared scourge? Clinical responses to treatment of MDRTB with standard first-line drug regimens have been poor (Table 1) (Coninx et al. 1999; Ivanovo Oblast 1999; Espinal et al. 2000; Garcia-Garcia et al. 2000; Drobniewski et al. 2002b). Rates of cure with standard short-course treatment range from 5% to 60% (Ivanovo Oblast 1999; Espinal et al. 2000; Lan et al. 2001). In a study from a prison in Baku, Azerbaijan, for example, treatment was successful in only 27% of patients with MDRTB and treatment failure in all patients was associated factors such as the breadth of the spectrum of drug resistance of isolates, positive sputum microscopy at the end of the initial treatment period, cavitary disease, and poor treatment compliance (Coninx et al. 1999). A multi-site study showed, as did the Baku study, that ‘an approximately linear increase in the likelihood of treatment failures was observed as the number of drugs to which the strains were resistant increased’ (Espinal et al. 2000). This later study reported a relative risk of treatment failure and death of 15.4 and 3.73, respectively, in patients with MDRTB compared with patients with drug-sensitive strains. Overall, research suggests that treatment with first-line drugs for MDRB offers little benefit to individuals who are suffering from disease from these strains. Indeed, a retrospective review of cases in Vietnam concluded that this approach produces results ‘similar to historic outcomes when no chemotherapy for tuberculosis was given’ (Lan et al. 2001).

Table 1.  Studies of treatment and outcomes in patients with MDRTB
SettingMethodNo. of casesHIV statusTreatment regimenTreatment failure rateDeath rateMedian survival periodFollow-up periodNotesReferences
  1. * First-line treatment according to schedule in reference (WHO 1997a).

  2. NA, not available; N, no; Y, yes.

First-line drug regimens
Africa, Hong Kong, SingaporeProspective clinical trials11Presumed negativeFour or five first-line drug regimens for 6 months5/11 (45%)NANANATwelve controlled trials conducted by British Medical Research Council. Only 11 of 8212  patients had strains R resistant; of these nine MDRTB. Of the six who responded three relapsed after cessation of therapyMitchison and Nunn 1986
Baku, AzerbaijanProspective cohort30NAStandardized first-line drugs regimen under DOTS*22/30 (73%)NANA9 months after last recruitPrison setting 467 recruits, 131 had complete microbiology results. Failure here defined as failure to achieve three consecutive negative sputums at end of initial treatment phaseConinx et al. 1999
Southern MexicoProspective cohort25Presumed negativeStandardized first-line drugs regimen under DOTS*14/25 (56%)7/25 (28%)NAMedian 24 months for all TB cases284 cases recruited with TB to study. None of deaths were HIV+ (5/221 cases recruited known to be HIV+, none had MDRTB). Failure was defined as positive smear or culture at fifth month of treatmentGarcia-Garcia et al. 2000
Ivanovo, RussiaCase–control26NStandardized first-line drugs regimen under DOTS*18/19 (95%)6/22 (27%)NA12 monthsPrimary MDRTB. Failure means not achieving cure: negative sputum smear at treatment completion and on at least one previous occasionIvanovo Oblast 1999
Ho Chi Minh City, VietnamRetrospective case series42NAStandardized first-line drugs regimen under DOTS*30/42 (71%)1/42 (2%) For duration of treatmentFigure for failures include 1 death. ‘The proportion of patients cured of MDRTB … was similar to historic outcomes when no chemotherapy for TB was given’Lan et al. 2001
Dominican Republic, Hong Kong, Italy, Ivanovo Oblast in Russia, Republic of Korea, PeruRetrospective case series353 (184 new, 169 retreatment cases)Limited data. Five known. HIV positiveStandardized first-line drugs regimen under DOTS* followed in some cases by individualized regimens following DST resultsNew cases 39/184 (21%). Retreatment cases 57/169 (34%)New cases 16/184 (9%). Retreatment cases 18/169 (11%)NANAFailures were patients who had smear- positive status at 5 monthsEspinal et al. 2000
Individualized treatment regimens with second-line drugs
Denver, USARetrospective case series171NIndividualized with second-line drugs47/134 (35%)63/171 (37%)13 years from initial diagnosis of TBMean 51 months from referral to study centrePatients had had TB for a median of 6 years before referral to Denver. Failure defined as persistently positive cultures despite chemotherapy. Further 14% relapsed during follow-up. Mortality attributable to TB 22%Goble et al. 1993
New York, USARetrospective case series17390 (52%) HIV+Individualized with second-line drugsPrincipal outcome measured was death100/141 (71%) 10/35 (29%) in HIV− 68/76 (89%) in HIV+10 years in HIV− 14 months in appropriately treated HIV+; 1.85 months in ‘untreated’ HIV+Up to 10 yearsStudy included HIV- negative and HIV- positive patients, 173 in total. Survival status unknown in 32 cases. Death attributed to TB in seven of 10 (70%) HIV−; 23/68 (34%) in HIV+Park et al. 1996
New York, USARetrospective case series25NIndividualized with second-line drugs1 (4%)1/25 (4%)NAMedian 45 monthsOne clinical failure, died at 5 days; 17 of 17 microbiological response one patient relapsed after 17 months of treatmentTelzak et al. 1995, 1997
TurkeyRetrospective case series158NAIn new patients, individualized after failure of first-line treatment*. Re-treatment cases received individualized treatment13/158 (8%)7/158 (4%)NANAFailure defined as persistence of positive smears or cultures despite treatment for at least 18 or 24 months Unsuccessful outcome included treatment failure, incomplete treatment, death, relapse occurred in 37/158 (23%)Tahaoglu et al. 2001
Masan, KoreaRetrospective case series107NAIndividualized with second-line drugs11/63 (17%)0Median 17 monthsTreatment regimens preferably included four drugs previously unused and to which strains were susceptible; 62 had sufficient follow- up data. Failure means positive culturesPark et al. 1998
Lima, PeruRetrospective case series751/65 (2%) HIV+Individualized with second-line drugs for at least 18 months; 58 different regimens1/75 (1%)17/75 (23%)NA – see commentsMedian 40 monthsNine patients died within 4 months or treatment initiation, five after 4 months, and three after treatment completion. Failure defined as positive culture after 6 months treatmentMitnick et al. 2003
Seoul, KoreaRetrospective case series1011NAIndividualized with second-line drugs for at least 18 months; 73 different regimens82/1011 (39%)3/1011 (0.3%)Mean 12 years from initial treatment (at 3, 11 and 18 months or MDRTB-specific treatment)Up to approximately 12 years487/1011 (48%) cured (defined as two or more negative sputum cultures towards the end of treatment). Failure is two or more positive sputum cultures toward end of treatment or a case that defaulted after 12 months of treatment with positive cultures. Of note, 394 (39%) defaulted (interrupting treatment for two consecutive months other than failed cases)Kim et al. 2001
New York, USARetrospective case series34YIndividualized with second-line drugs17/34 (50%)17/34 (50%)315 daysNAThirty-four of 38 cases HIV+. Rapid death in patients receiving ‘inappropriate’ treatment for two consecutive weeks. Treatment failures defined microbiologically or clinicallyTurett et al. 1995
Standardized treatment regimens with second-line drugs
PeruProspective cohort298NAStandardized 18-month regimen: kanamycin (3 months), ciprofloxacin, ethionamide, pyrazinamide, ethambutol96/298 (32%)32/298 (11%)NA18 monthsNo data provided on relapse ratesSuarez et al. 2002

With second-line drugs, treatment success in MDRTB varies from 48% to >80% of patients cured (or probably cured) (Park et al. 1998; Kim et al. 2001; Mitnick et al. 2003). Death rates vary from 0% to 37% in studies of HIV-seronegative individuals, and up to 89% in HIV-seropositive populations (Goble et al. 1993; Park et al. 1996, 1998). Higher death rates, not surprisingly, were recorded in those studies with longer follow-up periods, but this also suggests that the effect of MDRTB disease in individuals may not be seen for years – a scenario that is very different from drug-sensitive tuberculosis treated with standard first-line drug regimens. Follow-up duration and survival analyses in many of these studies are relatively short (or unclear), however, and the long-term implications of treatment for individuals and the public health impact of these interventions remain somewhat uncertain. Moreover, many of these studies were conducted using a variety of methodologies and have outcomes (of cure, success, failure) that are defined in different ways making comparisons between studies difficult.

Reports on experience in treating MDRTB using second-line drugs in middle-income countries have been, until very recently, few in number, and from low-income countries there have been none (Hadiarto et al. 1996; Suo et al. 1996; Park et al. 1998; Yew et al. 2000; Tahaoglu et al. 2001). More recently, however, successful approaches to treatment using second-line drugs in low-income countries, notably Peru, have been reported (Suarez et al. 2002; Mitnick et al. 2003). Different approaches to treatment were taken in these two studies with markedly different outcomes.

The uncertainties that persist in our understanding of the long-term impact of treatment of MDRTB have an important resonance when we contemplate resource allocations (Coker 2002). Moreover, because robustly designed prospectively conducted clinical trials have yet to determine the most effective clinical approaches to management of MDRTB, treatment strategies are still largely dependent upon professional experience and upon drawing inferences from retrospective case series and cohort studies. Several of these reports suggest that poor clinical condition prior to treatment initiation and resistance to a large number of drugs are associated with poor outcomes (Suarez et al. 2002; Mitnick et al. 2003). Furthermore, in those patients in whom sputum sterilization is not achieved, survival may be poor (Goble et al. 1993). Moreover, it appears that more recent treatment strategies that include a quinolone may offer advantages (Yew et al. 2000; Tahaoglu et al. 2001; Mitnick et al. 2003).

One of the challenges clinicians face, therefore, is to decide the most effective drug regimen for any given patient. Broadly, two approaches under a rubric of DOTS-plus have been proposed and these are, in essence, dependent upon local laboratory capacity and resources (Farmer et al. 1999).

One is based on an assessment of the background prevalence of second-line drug resistance and an assumption that chronic patients are likely to have strains of similar pattern (despite the fact that patterns of second-line resistance are largely unknown in almost all regions of the world) (WHO 1997b). Standardized treatment that includes second line drugs are offered, usually for a period of 18 months to 2 years. The National TB Control Programme of Peru adopted this approach using a regimen consisting of kanamycin, ciprofloxacin, ethionamide, pyrazinamide, and ethambutol (Suarez et al. 2002). Advocates of this approach argue that tailored treatment based upon individual determinations of drug sensitivity patterns is beyond the budget of resource-poor countries, laboratory capacity is insufficient to deliver timely, reliable results, dependence on reference laboratories may introduce clinically important delays, and that standardized treatment is the only feasible alternative if patients are to be treated. Using standardized treatment advances the notion that those failing empiric short-course regimens are presumed to have MDRTB. Those individuals whose treatment fails or who relapse are then offered a re-treatment regimen that includes second-line drugs (Crofton et al. 1996; Suarez et al. 2002; Quy et al. 2003). An extension of this approach is to treat those who have failed a first treatment with second-line drugs, offering them the WHO re-treatment regimen of first-line drugs. Those who fail this re-treatment regimen are then offered standardized treatment with second-line drugs. The reason behind this approach is that relatively few failures of treatment may have MDRTB in some settings, such as where rifampicin is not part of the treatment regimen in the continuation phase (22% in a rural Bangladeshi population, for example) compared with most re-treatment failures (87% in the Bangladesh study) (Crofton & Van Deun 2000). Thus one may be able to estimate the probability of MDRTB in an individual and treat accordingly without incurring costs and delays whilst performing and awaiting drug susceptibility test results. Of concern in advancing this approach is that where rifampicin is used throughout the first treatment, failure correlates much more closely with MDRTB (Becerra et al. 2000). In their study of 160 consecutive treatment failure patients Becerra et al. (2000) showed that 150 (95%) had MDRTB. Clinical management should be informed by the prior use first-line treatment regimens.

Few studies have been conducted using standardized treatment approaches for MDRTB, but those that have show outcomes that are worse than for most studies using individualized treatment regimens in expert hands – but better results for individuals than either no treatment or treatment with first-line drugs alone (Suarez et al. 2002). Further research is needed to determine the relative merits of standardized regimens compared with individualized approaches.

The alternative strategy, and one more usually adopted in resource-rich settings, is to use individualized treatment regimens based upon an assessment of the likely sensitivity on the basis of previous regimens used and drug-susceptibility testing derived from patient specimens. This approach, whilst costly, offers a reduction in the potential to amplify resistance further, and may be more effective than a standardized approach (Table 1). Regimens therefore vary for each patient depending upon clinical history and sensitivity patterns of isolates. Delays in the determination of sensitivity patterns (usually about 2 months with traditional methods) mean that the initiation of definitive regimens may be delayed (Mitnick et al. 2003).

A comparison between standardized and individualized treatment strategies may be inferred from results of research taking these two approaches in Peru (Suarez et al. 2002; Mitnick et al. 2003). In the study taking an individualized approach few failures were observed and, over a median follow-up period of 40 months, 23% patients died. This model relied upon considerable independent financial support for drugs and drug susceptibility testing was conducted by a laboratory in the US. The standardized approach resulted, by comparison, in significantly higher failure rates (32%) but fewer deaths (11%) over a shorter assessment period.

Overall then, treatment using an individualized approach that is competently implemented drawing effectively upon the skills of clinicians with experience in managing this demanding and complex disease appears to be the most effective approach. But ensuring standards are maintained, that sufficient human and other resources are available and can be drawn upon in a sustained manner are real challenges. If such an approach cannot be assured, and many international experts fear this is true, then many argue that standardized approaches should be implemented. The question, still unanswered, is then, which standard regimens and under which circumstances?

Squandering second-line drugs?

Few drugs with effective anti-mycobacterial properties have been developed in recent decades. Consequently the therapeutic armamentarium to combat MDRTB is largely made up of drugs developed many decades ago – many of which had become almost obsolete because of their relatively poor activity and potent toxicity profiles. Restricting the widespread use of the few newer antimicrobials that do have anti-mycobacterial activity has also been a substantial challenge, and one than has not been met. The widespread use of quinolones in non-tuberculous disease, for example, means that their effectiveness as anti-tuberculosis agents may already be being diminished (Grimaldo et al. 2001; Tahaoglu et al. 2001; Reichman & Tanne 2002).


An important, perhaps the most important, determinant of outcome of MDRTB is the presence of HIV co-infection (Park et al. 1996). Clinical outcomes may be very poor in this setting. This fact was highlighted in outbreaks in the early 1990s in the United States which were associated with extremely high mortality rates (72–89%) with rapid progression from disease to death within weeks (median interval from diagnosis to death, 4–16 weeks) (CDC 1991a,b, 1993; Edlin et al. 1992; Fischl et al. 1992; Pearson et al. 1992). In the mid-1990s Turett et al. (1995) showed that if anti-tuberculosis drugs to which the organism was unlikely to be sensitive were used in treating MDRTB associated with advanced HIV disease, then within 2 months mortality rates approaching 100% were likely to result. If, however, a regimen was used that organisms were likely to be sensitive to then survival at 1 year was in the order of 60% (Turett et al. 1995). More recently, Drobniewski et al. (2002b) showed that immunocompromised individuals in the United Kingdom with MDRTB are nine times more likely to die than patients who are not immunocompromised. Other studies support the notion that early institution of appropriate treatment may extend survival even if individuals are HIV positive (Turett et al. 1995; Park et al. 1996, 1998; Drobniewski et al. 2002b). Experience in management at specialized centres appears, in some studies, to confer prognostic benefits in managing patients co-infected with HIV and MDRTB.

Another important issue is our limited knowledge of the impact of treatment on MDRTB above and beyond the natural history of tuberculosis outside the bounds of experienced centres (or with their support), in centres where results from treatment may not be so accessible through peer-reviewed publications (for example, Coker 2002). Indeed, when contemplating current efforts in treating MDRTB it is worth reflecting upon comments made by Grzybowski (1983) two decades ago: ‘The largely forgotten fact that one-third of patients with advanced disease with positive sputum smears recover on their own, should be kept in mind when claims are made that an inappropriate drug regimen cured a patient with bacterial resistance’. Indeed, there is a remarkable symmetry between some survival curves describing the natural history of tuberculosis in the pre-antibiotic era and a contemporary survival curve associated with MDRTB in a recent analysis from the UK, a country with a tradition of treating MDRTB with individualized treatment regimens (Figure 2) (Stephens 1941; Tattersall 1947; Buhl & Nyboe 1967; National Tuberculosis Institute 1974; Drobniewski et al. 2002b). Most contemporary studies (with some notable exceptions (Park et al. 1996; Kim et al. 2001) describe survival over short periods where benefits are usually marked between, for example, immunocompromised and immunocompetent patients or those treated with ‘appropriate’ treatment compared with those not (Turett et al. 1995; Park et al. 1996, 1998; Drobniewski et al. 2002b). Survival at 5 years may be in the region of only 50% in the absence of HIV co-infection in some settings, a figure not too dissimilar from that noted in pre-chemotherapy studies (Goble et al. 1993; Drobniewski et al. 2002b).

Figure 2.

Survival trends in untreated individuals and those treated for MDRTB (sources: Drobniewski et al. 2002b; National Tuberculosis Institute 1974).


Choice of appropriate treatment is an important determinant of more favourable outcomes. If patients receive early treatment that the organism proves be sensitive to then treatment is more likely to be successful (Turett et al. 1995; Tahaoglu et al. 2001; Drobniewski et al. 2002b). Hence a number of challenges face those managing MDRTB in the diagnostic arena.

The first is whether rapid diagnostics using amplification techniques to determine rifampicin resistance (as a marker for MDRTB) early after presentation are an effective (and cost-effective) clinical tool such that treatment for individuals might be more appropriately tailored using second-line drugs and outcomes potentially improved.

A further important challenge is to answer the question whether such an approach, in public health terms, will impact upon efforts to control the public health threat that is MDRTB by reducing the potential for transmission by shortening the time to initiation of appropriate treatment and duration of infectiousness.

Rapid determination of drug resistance profiles of isolates causing disease would be a useful step in enabling clinicians to tailor treatment appropriately earlier in the disease course and potentially reduce morbidity, mortality, and duration of infectiousness and thereby reduce public health threat. Greater access to drug susceptibility testing, particularly in the developing world, might also facilitate epidemiological surveillance of drug resistance. Conventional methods for determining susceptibility such as the proportion method and the absolute concentration method are based on the measurement of growth in culture media containing antibiotics (Canetti et al. 1969; Kent et al. 1985). Whilst relatively inexpensive and undemanding of sophisticated equipment, results usually take weeks and this is problematic, particularly where inappropriate choice of treatment regimen may result in death within weeks of initiation and before results arrive at the bedside. Whilst the introduction of the BACTEC® radiometric system, and its modification for drug susceptibility testing (BACTEC® TB-460), has been a significant advance (Roberts et al. 1983), this expensive technology remains beyond the reach of most tuberculosis control programmes and is hindered by the need to dispose of large volumes of radioactive materials (Heifets & Cangelosi 1999). Moreover, whilst turnaround time is significantly better than conventional methods, it may still be in the order of 3 weeks (Roberts et al. 1983). Considerable research efforts have gone into the development of novel rapid drug susceptibility tests, yet many of these remain, at present, restricted to developed countries or reference laboratories because of their expense and sophistication (Palomino 2000). Moreover, the application of these approaches to support individualized treatment through determination of second-line drug susceptibility profiles remains largely unexplored implying that their application in support of individualized treatment of MDRTB remains uncertain. Novel approaches broadly fall into two groups, genotypic methods and phenotypic methods.

Genotypic techniques (including automated DNA sequencing (Pai et al. 1997), polymerase chain reaction (PCR)-single strand conformation polymorphism (Pretorius et al. 1996), PCR-heteroduplex formation (Wiliams et al. 1998) have resulted from recent insights into the molecular basis of drug resistance and the application of new molecular biology tools. All demand DNA extraction, gene amplification, and detection of mutation (for example in the rpoB gene for rifampicin resistance) and therefore are relatively expensive and sophisticated, demanding resources and skills that are usually unavailable in most regions where rates of MDRTB are high. A further limitation of these approaches at present is that not all molecular mechanisms of drug resistance are known. Their potential advantage is that there is no need for growth of the organism and drug susceptibility results can be determined in days rather than weeks. Moreover, research suggests that they can be highly reliable (Palomino 2000).

Perhaps the most encouraging development is the adoption of phage technology to rapidly detect drug-resistance profiles (McNerney et al. 2000; Albert et al. 2002). Several tools have been developed and the high correlation with standard approaches suggests that they may be useful in facilitating the rapid initiation of appropriate management of patients with MDRTB (Takiff & Heiferts 2002).

Novel rapid phenotypic methods to determine drug susceptibility harness technologies that can detect metabolic activity or early visualization of microcolonies. The most advanced and well validated of these is the Mycobacteria Growth Indicator Tube (MGIT) which detects oxygen consumption in the presence or absence of drug (Reisner et al. 1995). As with genotypic approaches, detection of drug resistance can be accomplished in days rather than weeks but is again constrained by cost. Alternative approaches that might lend themselves to laboratories in developing countries because of their relatively low cost are microcolony detection but further evaluation of these approaches are needed (Schaberg et al. 1995; Mejia et al. 1999).

Despite substantial advances in rapid diagnostics and susceptibility testing in recent years significant challenges persist. Perhaps the most critical of these is cost. Most novel rapid diagnostic approaches developed require expensive equipment and highly trained personnel. Others that might lend themselves to resource-poor settings will require careful evaluation and field testing to ensure acceptable levels of sensitivity, specificity and reproducibility. Finally, and of greatest relevance if individualized second-line regimens are to become widely used, rapid determination of susceptibility patterns to second-line drugs remains a substantial challenge.

Whilst high laboratory sensitivity and specificity measures for isoniazid and rifampicin resistance support surveillance, only in recent years has the sensitivity for ethambutol improved to above 90% through the Global Network of Supranational Reference Laboratories (WHO 2000). Moreover, for pyrazinamide, drug susceptibility results remain less reliable (Zhang et al. 2002). For second-line drugs, where an understanding of susceptibility profiles could inform clinical management either through an understanding of individual or population-based profiles, standardization and quality assurance schemes are in their infancy and, moreover, not yet included in international programmes.


Many experts have argued that treatment of MDRTB in low-income countries might be an inefficient use of limited resources (Crofton et al. 1996; Dye et al. 2002). Treatment of MDRTB is certainly considerably more expensive than treating drug-sensitive disease. Indeed, Dye et al. (2002) has suggested that treating the global burden MDRTB may cost as much as treating all the world's remaining drug-sensitive cases. And there has been much debate over whether relatively resource-poor countries, where much of the burden of MDRTB falls, can afford to manage such cases (Farmer et al. 1998; Coker 2002; Mitnick et al. 2003).

Until recently, the response to the epidemic of MDRTB in New York was held up as an example of the potential costs that might be incurred to adequately address MDRTB in some settings. Reports of the $1 billion expended to support tuberculosis control efforts in the early 1990s in that city have meant that many have considered the management of MDRTB beyond the resources of lower income nations whilst at the same time the success of this programme offered a glimpse of the possible (Frieden et al. 1995). The fact that this $1 billion was expended on an array of issues including a profound restructuring of New York's control programme is often overlooked (Frieden et al. 1995). In London, more recently, estimates of the cost of treating a patient with MDRTB are in the region of £60 000, 10 times the cost of treating drug-sensitive disease (White & Moore-Gillon 2000). In other parts of the world, for example in India, local production and local procurement mechanisms for second-line drugs can mean that treatment costs which include second-line drugs may be considerably lower.

Recently, a cost-effectiveness analysis of a treatment programme in Peru of patients with MDRTB offered hope by suggesting that cases might be effectively treated under a standardized treatment approach for as little as approximately $2400 per case, producing a cost per disability adjusted life year (DALY) gained of $211 – a figure not too dissimilar from what the World Bank has suggested should be viewed as an attractive investment in low-income countries (Suarez et al. 2002). This important study also showed that such an approach might be feasible in some low-income settings. However, as noted above, with failure rates of one-third, important questions remain regarding the effectiveness of a standardized treatment approach which includes second-line drugs. Moreover, it may not be a simple task to replicate elsewhere the organizational capacity, and political, patient, and professional commitment demonstrated in Peru. How this capacity and commitment might be developed elsewhere is an important lesson to be drawn from Peru.

One advance in the promotion of treatment strategies to combat MDRTB has been the effectiveness of the Green Light Committee. This multi-institutional partnership has been charged with negotiating lower prices for second-line drugs to combat MDRTB and facilitating access to these drugs at negotiated prices for programmes that are deemed to adhere to internationally accepted control practices. Substantial price reductions have been achieved for a raft of second-line drugs and over the past couple of years estimated costs for an 18-month course of treatment have fallen from approximately from $5000 to $1800 (Gupta et al. 2001). Whilst the Green Light Committee has been a useful agency to negotiate price reductions and ensure drug quality for programmes adhering to the DOTS strategy, local production of second-line drugs in some countries ensures that these negotiated prices can be undercut but quality may not be assured. Moreover, it might be argued that the success of advocacy for MDRTB treatment programmes has forced some governments, such as those in the former Soviet Union, to respond by initiating MDRTB treatment programmes prematurely, not after developing robust DOTS programmes as advanced by WHO, but by building upon already frail programmes in a piece-meal manner. In other regions the private sector is playing a role, a role that is sometimes ill-coordinated and ineffective.

Clearly the issue of drug costs is only a small part of the costs incurred in MDRTB treatment programmes. Hospitalization practices which may be very costly, for example, vary considerably in different regions of the world and are informed by clinical traditions, social demands, and health system approaches to caring for the ill as much as the epidemiology of MDRTB. In New York and London, for example, hospitalization costs for MDTRTB are considerable (Frieden et al. 1995; White & Moore-Gillon 2000). In Russia, while costs are much lower, the tradition of hospitalization for prolonged periods is embedded more firmly and may offer benefits as well as costs in controlling MDRTB (Polivakho & Sharaburova 2000).

Moreover, whilst the costs of treating MDRTB relate to the absolute numbers of cases, where rates are high (such as in the Baltic states), there may be greater political commitment because absolute numbers indicate that a programme of control might be feasible. In settings where rates are moderate but absolute numbers likely to be very substantial, such as in India, then the magnitude of the problem (and its geographical spread) has the potential to create political inertia. This problem seems likely to be compounded when both rates and absolute numbers are likely to be high, such as in Russia.

Infectiousness and virulence

Critical to informing policy making and the allocation of resources to address MDRTB is an understanding of the transmission dynamics of MDRTB. If epidemics of MDRTB are self-limiting even if treatment strategies are not adopted, policy makers may prefer to allocate scarce resources to individual and public health interventions that provide greater returns on their investment (Dye et al. 2002). The ethical and cultural frameworks within which such questions are asked are important (Coker 2002).

One of the most significant challenges facing those committed to controlling MDRTB is understanding more about the relative infectiousness of resistant strains of Mycobacterium tuberculosis compared with drug-sensitive strains and using his understanding to tailor control policies more effectively. Somewhat conflicting epidemiological evidence suggesting that MDRTB strains are as or more infectious comes from a variety of sources. Cases of MDRTB, in some settings, appear to cluster less than drug-sensitive strains (e.g. Garcia-Garcia et al. 2000; Nitta et al. 2002). Furthermore, molecular fingerprinting to identify clustering has, in some studies, suggested a reduced propensity to cluster for drug-resistant strains (van Soolingen et al. 1999). In this study, isoniazid resistant strains were less likely to cluster suggesting that some strains are less likely to produce secondary cases. Other epidemiological studies including longitudinal studies have, however, suggested higher cluster rates with MDRTB (Snider et al. 1985; Alland et al. 1994; Frieden et al. 1996; Bifani et al. 2001; Teixeira et al. 2001).

The other basic question, again one that remains unresolved, is whether drug-resistant strains are as virulent (i.e. less able to cause disease) as drug-susceptible strains. In animal models in the 1950s and 1960s research suggested that strains resistant to isoniazid and streptomycin were less infectious and less virulent than susceptible strains (Wolinksky et al. 1956; Mitchison et al. 1960; Cohn & Davis 1970). In vitro research supports this notion by showing that some isoniazid-resistant strains grow less well, and some rifampicin-resistant strains have also been shown to grow less well (De Beenhouwer et al. 1995). Contradicting the notion that resistant strains might be less virulent is the epidemiological evidence of clustering and other research showing that resistance, at least to some strains with mutations conferring rifampicin and isoniazid resistance, may not attenuate pathogenic behaviour (Wolinksky et al. 1956; Billington et al. 1999).

Issues regarding whether differences in clustering result from altered infectiousness or virulence, combinations of the two, or differences in these characteristics in subpopulations of resistant strains therefore remain to be clarified. Whether some strains, for example Beijing which is frequently associated with drug resistance and clusters and may be an escape mutant from BCG vaccination, gain a selective advantage is a further issue that requires research (Lopez et al. 2003).


The challenges posed by MDRTB are substantial. They start with considerable uncertainty regarding the magnitude of the problem, and the potential impact of another epidemic, that of HIV in parts of the world reporting high prevalence rates of MDRTB. Most notable, and perhaps most worrying, is the situation in some states of the former Soviet Union (FSU). Along with huge political, economic, social and cultural changes, many countries of the FSU have health care and social welfare infrastructures that have crumbled or are fractured almost beyond repair and are ill equipped to respond to these new threats. With regional borders shifting with the enlargement of the European Union, other international borders remaining porous, along with conflicts and other causes of socio-political turmoil precipitating mass migrations, it seems sensible that nation states of the affluent west remain focused on supporting efforts to address MDRTB globally– if not for humanitarian reasons then out of self-interest.

Where this support should be focused remains uncertain, however. The urgency of the task, some would argue, means that programmes should be initiated immediately drawing upon the evidence existing to support practical approaches. This means that although the evidence base upon which programmes may sit may not be as robust as some purists might wish, the humanitarian disaster that threatens to unfold demands action now rather than waiting until the evidence base is more robust. Others argue that if the adopting approaches that have not been subjected to robust critical analysis from a public health perspective then programmes may result in inefficient, indeed counterproductive, outcomes. At the heart of this challenge is considerable uncertainty about the transmission dynamics of MDRTB strains and the impact of the DOTS and DOTS-plus strategies upon these dynamics. Controversy has centred on the appropriateness and effectiveness of the DOTS strategy in regions with differing prevalence rates of MDRTB. What seems clear is that robust DOTS programmes will prevent the emergence of drug-resistant strains if rates of pre-existing MDRTB are low. The controversial issue is what should be done where MDRTB is already endemic (Farmer & Kim 1998). Some have argued that MDRTB strains may be less ‘fit’ and that, with a robust DOTS programme strains of MDRTB would eventually fall because of higher fatality rates and reduced transmission without focused attention on MDRTB (Dye et al. 2002). From the perspective of an individual with MDRTB, however, programmes that adhere to a DOTS strategy using first-line drugs alone seems to offer little beyond the natural history of tuberculosis in the pre-chemotherapy era (Coninx et al. 1999; Kimerling et al. 1999; Lan et al. 2001). And there may be a risk in settings where single drug resistance is prevalent that use of DOTS may create selection pressures that facilitate the emergence of MDRTB – the so-called ‘amplifier’ effect (Farmer et al. 1998). One challenge is, therefore, to determine the right mix of interventions in a variety of contextual settings to ensure control of drug-susceptible TB and prevent the emergence of MDRTB whilst treating those that have MDRTB effectively in order to ensure their survival and protect the public from transmission of these strains.

Other issues, of course, challenge MDRTB control. Evidence in support of preventive therapy is almost non-existent, a challenge that will assume considerably greater importance as the epidemics of MDRTB and HIV collide in parts of the FSU. New, highly effective drugs for M. tuberculosis remain elusive whilst the potential benefits offered by some newer drugs with anti-mycobacterial activity, such as the quinolones, are squandered because more profitable markets exist in treating diseases more prevalent in the West and because of unconstrained use in the developing world (Reichman & Tanne 2002).

Other important challenges not addressed in this paper include the ethical and legal dilemmas of managing patients who either decline treatment or are untreatable but threaten public health (Coker 2001). If effective treatment for some individuals with MDRTB remains illusory, do these patients face a disease condition that parallels disease in the pre-chemotherapeutic era, where about 30% of cases remain smear positive and infectious (Grzybowski 1983)? And what policies should be developed to deal with these populations?

The criminal justice system, which appears to fuel the epidemic of MDRTB in some countries, must be reformed to ensure that, in Vivien Stern's words, prisoners are not ‘sentenced to die’ (Stern 1999). Whilst nosocomial transmission of MDRTB is most obviously an unjust consequence of incarceration, the threat of spread within institutions in the civil sector remains a very real possibility, particularly where lengthy hopitalizations for tuberculosis are the norm, as in the FSU. Indeed, more broadly, the challenge of MDRTB raises other equally intransigent challenges associated with poverty, injustice and alienation.

A further challenge for public health practitioners will be to understand systems that respond effectively to the challenge of MDRTB. Robust analytical frameworks that capture the political, legal, economic, technological, social, and organizational factors which make up the context of programmes charged with controlling MDRTB are needed so that lessons can be drawn and applied effectively, in a nuanced, sophisticated, and culturally appropriate fashion (Atun et al. in press).

The public health challenges raised by the spectre of MDRTB are manifold. They range from the epidemiological to the diagnostic, from the economic to the legal, and from the social to the political. In facing these challenges and contemplating remedies we should perhaps reflect on the suggestion that tuberculosis is ‘the perfect expression of our imperfect civilization’ (Dormandy 1998). That MDRTB, a man-made phenomenon, only adds to the authority of these sentiments.


I am grateful to the anonymous reviewers who commented on an earlier draft of this paper and made valuable suggestions.