Outcome of Multi-drug-resistant Tuberculosis in France - ATS Journals

Outcome of Multi-drug-resistant Tuberculosis in France A Nationwide Case-Control Study MARIE FLAMENT-SAILLOUR, JÉRÔME ROBERT, VINCENT JARLIER, and JACQUES GROSSET National Reference Center for Surveillance of Mycobacterial Infections and Their Drug Resistance; and Service de Bactériologie-Hygiène, Groupe Hospitalier Pitié-Salpêtrière, Paris, France

The factors related to the outcome of 51 cases of multi-drug-resistant tuberculosis (MDR-TB) reported in 1994 to the French National Reference Center were retrospectively analyzed. The patients (median age, 45 yr) were mainly male (75%), foreign-born (63%), and had pulmonary involvement (95%). Sixteen percent were human immunodeficiency virus (HIV)-coinfected. The number of drugs to which the Mycobacterium tuberculosis isolates were susceptible was four. Only 82% of the patients have been hospitalized at any time (median duration, 33 d). Five patients (9%) received no antituberculosis drugs, although three had drug susceptibility results, indicating that two or more active drugs were available; 46 (91%) received drugs, including 37 who received two or more active drugs. Among the nine cases who received only one active drug, three had drug susceptibility results, indicating that two or more active drugs were available. By December 1996, 10 patients were lost before treatment completion, 24 had treatment failure, and 17 had a favorable outcome. The median survival time was 31 mo. Factors related to a poorer outcome were HIV-coinfection (hazard ratio [HR] 5 41), treatment with less than two active drugs (HR 5 9.9), and MDR status knowledge at the time of diagnosis (HR 5 3.3). The country of birth was not associated with a poorer outcome. The management and outcome of MDR-TB in France has to be improved. A solution would be to develop a specialized unit or team for the treatment of MDR-TB, as recommended by the World Health Organization (WHO). Flament-Saillour M, Robert J, Jarlier V, Grosset J. Outcome of multi-drug-resistant AM J RESPIR CRIT CARE MED 1999;160:587–593. tuberculosis in France: a nationwide case-control study.

Multi-drug-resistant tuberculosis (MDR-TB), defined as resistance of Mycobacterium tuberculosis to at least isoniazid (INH) and rifampin (RIF), is a worldwide problem. Patients infected with MDR strains are not only difficult to cure but also more likely to remain sources of infection for a longer period of time than those with drug-susceptible organisms (1). Because of the increased prevalence of MDR-TB in many places of the world (2–4), and outbreaks of MDR-TB among human immunodeficiency virus (HIV)-infected patients, especially in the United States (5–7), a national surveillance system of MDR-TB was developed in France. The aim of the surveillance, monitored by the National Reference Center (NRC) for Surveillance of Mycobacterial Diseases and Their Drug Resistance, was to assess the prevalence of MDR-TB, to differentiate primary from secondary MDR, and to analyze the factors related to MDR-TB in order to suggest public health interventions. In France, the prevalence of MDR-TB in 1994 was 0.7% of bacteriologically confirmed cases of tuberculosis, i.e., 58 out of 7,752 cases, not significantly different from the 1992 and 1993 prevalence rates (0.6 and 0.5%, respectively) (8). Further in( Received in original form January 8, 1999 and in revised form March 29, 1999 ) Supported in part by La Fondation pour la Recherche Médicale. Correspondence and requests for reprints should be addressed to Jacques Grosset, M.D., Bactériologie-Hygiène, Groupe Hospitalier Pitié-Salpêtrière, 47, boulevard de l’Hôpital, 75651 Paris Cedex 13, France. Am J Respir Crit Care Med Vol 160. pp 587–593, 1999 Internet address: www.atsjournals.org

vestigations revealed that 14 of the 58 cases of MDR reported in 1994 had already been reported to the surveillance system during the two preceeding years (8), suggesting that these patients remained active sources of infection for at least 2 yr. To analyze the factors related to the persistance of MDR sources of infection, the management, and the outcome of all MDR-TB cases reported in France during the year 1994 were retrospectively investigated during the year 1997. Baseline characteristics, treatment management, and outcome of cases of MDR-TB were compared with those of patients with drugsusceptible tuberculosis in two case-control studies.

METHODS Study Population MDR-TB surveillance in France relies upon a network of 360 laboratories covering the entire French territory and including a majority of hospital laboratories, military hospital laboratories, Pasteur Institutes, and the major private laboratories performing mycobacteria culture. MDR-TB was defined as bacteriologically confirmed tuberculosis with a M. tuberculosis strain resistant to at least INH and RIF. All patients from metropolitan France (n 5 50) and the French West Indies departments (n 5 1) with MDR-TB reported to the NRC in 1994 were defined as case-patients. Seven patients with MDR-TB from another single overseas department were excluded because of limited investigational possibilities. For each MDR-TB case, the first patient with non-MDR-TB diagnosed after the case-patient in the same setting was chosen through laboratory log books as a control-patient. As country of birth was suspected to be a factor related to MDR-TB and treatment management, a second case-control study was conducted with a second group of control-patients chosen with the same algo-

588

AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE

rithm as the first control-group, but frequency matched on three previously defined regions of birth (i.e., France, Western Europe, and foreign countries outside Western Europe).

Definitions The date of tuberculosis diagnosis was defined as the date of the first positive clinical specimen in the year 1994. A drug was considered active if M. tuberculosis was susceptible in vitro to this drug, as measured by the proportion method (9). For case-patients, the treatment regimen was considered adapted when it had included at least two active drugs, and favorable outcome was considered as having been achieved when patients were clinically improved and remained smearnegative for at least 3 mo by December 31, 1996. For control-patients, favorable outcome was considered when they received 6 mo of standard regimen, and were clinically and radiologically improved and sputum-culture–negative. Treatment failure was considered when case- or control-patients died before treatment completion with positive cultures or remained culture-positive by December 1996. The duration of case-holding was defined as the period of time between the date of diagnosis and the date of the last known contact with the physician in charge of the patient. A patient was defined as “lost” when he had not been retrieved before treatment completion. A patient being homeless, receiving social welfare, or unemployed was considered as having a low socioeconomic status. To measure adherence to antituberculosis medications, we used patient self-report of treatment discontinuation for more than 1 mo and failure to keep clinic appointments during the treatment course.

Data Collection Data including HIV status, prior treatment history, radiologic findings, microbiologic findings with drug susceptibility results (with the exception of pyrazinamide because of controversial interpretation [10]), drug regimens, case-holding characteristics, outcome, and cause of death (if appropriate) were retrospectively collected through physicians in charge of the patients and chart review. Patients or family members were contacted when necessary to assess vital status by December 1996.

Data Management and Statistics Data were collected on standardized forms, computerized and analyzed using Epi-Info 6.1 (Centers for Disease Control and Prevention, Atlanta, GA), and SAS (SAS Institute, Cary, NC) softwares. Categorical variables were compared using either the chi square or Fisher’s exact test. Odds ratios (OR) and their 95% confidence intervals (CI95)

RESULTS Baseline Characteristics

The main characteristics of the 51 patients with MDR-TB included in the study are given in Table 1. Patients were primarily men (74.5%) with a median age of 45 yr (range, 20 to 78 yr). Thirty-two (62.7%) were foreign born, 21 (41.2%) had a low socioeconomic status, 19 (37.3%) had no stable housing, and eight (15.7%) were HIV-coinfected. Tuberculosis was only pulmonary in 40 (78%) case-patients, and both pulmonary and extrapulmonary in nine (17%). Two foreign-born HIV-negative case-patients with a previous history of treatment had only extrapulmonary tuberculosis (Pott’s diseases combined in one of them with cervical lymph node involvement). Of the 49 case-patients with pulmonary involvement, 41 (83.7%) were sputum-smear–positive for acid-fast bacilli (AFB). Out of the 51 case-patients, 40 (78.4%) had relapse or chronic tuberculosis, as defined by the World Health Organization (WHO) (12). Among the 11 patients with no history of previous treatment, seven were foreign-born and HIV-negative, and four were French-born including three HIV-coinfected and one HIV-negative with a MDR-TB household contact. Antituberculosis drug susceptibility testing records of the 51 MDR strains are given in Figure 1. Seventeen (33%) isolates were reported as resistant only to INH1RIF, 18 (35%) as resistant also to streptomycin (SM), four (8%) to ethambutol (EMB), and 12 (24%) to both SM and EMB. Forty-six

BASELINE CHARACTERISTICS OF PATIENTS WITH MULTI-DRUG-RESISTANT TUBERCULOSIS (MDR-TB) AND FIRST CONTROL-PATIENTS WITH DRUG-SUSCEPTIBLE TUBERCULOSIS

Categorical variables Site of tuberculosis Pulmonary only Extrapulmonary only Pulmonary and extrapulmonary If pulmonary involvement: Positive AFB smear Lesions in more than one lobe Relapse/chronic tuberculosis Male Foreign-born Low SES Stable housing Marital status: single HIV-coinfection Continuous variables, median Age, yr (median) CD41 cells/mm3 (HIV-coinfected)

MDR-TB (n 5 51) (%)

Controls (n 5 51) (%)

Odds Ratio (CI95)

p Value

40 (78.4) 2 (3.9) 9 (17.7)

42 (82.4) 2 (3.9) 7 (13.7)

Reference 1.1 (0.1–15.1) 1.4 (0.4–4.7)

— 1.0 0.78

41 (83.7) 35 (71.4) 40 (78.4) 38 (74.5) 32 (62.7) 21 (41.2) 32 (62.7) 20 (39.2) 8 (15.7)

30 (61.2) 16 (32.7) 9 (17.6) 26 (51.0) 17 (33.3) 18 (35.3) 35 (68.6) 22 (43.1) 9 (17.6)

3.3 (1.2–9.6) 4.8 (1.9–12.1) 17.0 (5.8–51.9) 2.8 (1.1–7.1) 3.4 (1.4–8.3) 1.3 (0.5–3.1) 0.8 (0.3–1.9) 0.9 (0.4–2.0) 0.9 (0.3–2.8)

0.02 , 0.001 , 0.001 0.02 0.005 0.54 0.53 0.69 0.79

45 17 (1–300)

1999

were calculated using standard methods (11). The Mann-Whitney test was used for comparing continuous variables. Logistic regression models were used for multivariate analysis of factors associated with MDR-TB at the time of diagnosis. Survival analysis and identification of factors associated with poor survival in the MDR-TB group were done by the Kaplan-Meier method and Cox proportional hazard model, with death or treatment failure by December 1996 as the outcome. Lost patients were censored at the date of the last known vital status. Factors significantly associated with MDR-TB in univariate analysis and factors clinically reasonable and not colinear were included in the models. The Logrank test was used to determine the level of statistical significance when comparing survival curves; p values are two-tailed and p < 0.05 was considered statistically significant.

TABLE 1

Variables

VOL 160

40 175 (55–673)

Definition of abbreviations: AFB 5 acid-fast bacilli; SES 5 socioeconomic status.

— —

0.30 0.009

589

Flament-Saillour, Robert, Jarlier, et al.: Outcome of MDR-TB in France

Figure 1. Drug resistance of Mycobacterium tuberculosis strains in the 51 MDR-TB cases.

(90%) isolates were susceptible to at least one aminoglycoside, i.e., SM, kanamycin-amikacin, or capreomycin. The remaining five isolates were resistant to SM, but susceptibility tests to kanamycin or capreomycin had not been performed. A susceptibility test to ofloxacin or sparfloxacin had been performed for 44 (86%) of the 51 isolates: 36 (82%) were susceptible and eight (18%) were resistant. All of the latter were from patients previously treated with a fluoroquinolone for their MDR-TB. Of the 36 isolates susceptible to fluoroquinolones, 35 were also susceptible to at least one aminoglycoside. Susceptibility testing to ethionamide-prothionamide was performed for 27 (53%) of the 51 isolates: 16 (59%) were susceptible and 11 (41%) were resistant. In addition to rifampin and isoniazid, a mean of 5.3 drugs had been tested for each isolate (median, 5; range, 2 to 8 drugs). The mean number of drugs to which the isolates were reported as susceptible was four (median, 4; range, 1 to 8 drugs). If aminoglycosides were considered as one single drug, only five (10%) isolates should have been considered as susceptible to less than two drugs and 13 (25%) to less than three drugs. Surprisingly, the susceptibility of all the 13 latter isolates has not been tested to two or more second-line drugs. First Case-Control Study

Compared with the 51 control-patients, the 51 MDR-TB casepatients were similar in age, socioeconomic status, housing, HIV status, and location of tuberculosis (Table 1). However, case-patients were significantly more likely than control-patients to be men (OR 5 2.8; CI95, 1.1 to 7.1), to be foreign-born (OR 5 3.4; CI95, 1.4 to 8.3), to have lesions extended to more than one lobe (OR 5 4.8; CI95, 1.9 to 12.1), to be sputum-smear–positive (OR 5 3.3; CI95, 1.2 to 9.6), and to have relapse or chronic tuberculosis (OR 5 17.0; CI95, 5.8 to 51.9). HIV-infected casepatients had lower CD41 cell count than did HIV-infected control-patients (17 versus 175; p 5 0.009). In multivariate analysis, three variables remained independently associated with MDR-TB: relapse or chronic tuberculosis (OR 5 24.2;

CI95, 6.9 to 85.1), lesions of more than one lobe (OR 5 7.4; CI95, 2.2 to 25.3), and foreign-born status (OR 5 3.8; CI95, 1.2 to 11.8). HIV coinfection, site of tuberculosis, and sputum smear results were not associated with MDR-TB in multivariate analysis. Patient management. Among the 51 MDR-TB case-patients, 35 were managed by respiratory diseases specialists, seven by infectious diseases specialists, six by internal medicine physicians, and three by other types of physicians, in a total of 42 different clinical sites. Only 42 (82.4%) of the case-patients were hospitalized at any time during the course of the treatment (Table 2). The median duration of initial hospitalization was 18 d (range, 0 to 224 d), and that of total inpatient care was 33 d (range, 0 to 476 d). None of the patients was hospitalized during the whole treatment course and none received directly observed therapy (DOT) when he was an outpatient. Only 46 (91%) out of the 51 case-patients received antimycobacterial drugs. Among them, 19 (41.3%) reported treatment discontinuation for more than 1 mo, and 27 (58.7%) alleged they did not interrupt their medication. The crude median duration of antimycobacterial treatment was 8.3 mo (range, 0 to 34 mo) for the 51 case-patients and 10 mo (range, 0 to 34 mo) for the 35 case-patients who were not lost before treatment completion. If the seven case-patients who died in hospital within the first 2 mo after the diagnosis of tuberculosis were excluded from the analysis, the median duration of treatment would be 13.8 mo (range, 0 to 34.5 mo). Of the 46 case-patients treated with antimycobacterial drugs, 36 (78.3%) received regimens that included at least two active drugs, and 24 (52.2%) received regimens that included at least three active drugs. The median duration of treatment with at least two active drugs was 11 mo (range, 1 to 34 mo) for the patients not lost before treatment completion and would be 12 mo (range, 3 to 34 mo) if the patient who died within the first 2 mo of treatment was excluded. Of the 10 case-patients treated with less than two active drugs, one was lost 21 d after diagnosis, six died before susceptibility test re-

590


VOL 160

1999

TABLE 2 MANAGEMENT AND OUTCOME OF PATIENTS WITH MULTI-DRUG-RESISTANT TUBERCULOSIS (MDR-TB) AND FIRST CONTROL-PATIENTS WITH DRUG-SUSCEPTIBLE TUBERCULOSIS Variables Continuous variables, median Duration of: Initial hospitalization, d Total hospitalization, d Case-holding, mo Treatment, mo Categorical variables Hospitalization for treatment Treatment strategies: Hospital 1 outpatient Hospital 1 institution 1 outpatient Outpatient only Failed to keep clinic appointment Lost to follow-up Treatment stopped . 1 mo Treated with less than two active drugs Treated with less than three active drugs Deceased with active tuberculosis Alive with active tuberculosis Cured (if not lost to follow-up)

MDR-TB (n 5 51) (%)

18 33 11 8.3

Controls (n 5 51) (%)

15 19 11 7.0

42 (82.4)

46 (90.2)

29 (56.9) 13 (25.5) 9 (17.6) 18 (35.3) 16 (31.4) 19 (37.3) 15 (29.4) 27 (52.9) 19 (46.3) 5 (12.2) 17 (41.5)

32 (62.7) 13 (25.5) 6 (11.8) 7 (14.0) 1 (2.0) 9 (18.0) 5 (9.8) 5 (9.8) 5 (9.8) 0 45 (90.0)

sults were available, and three were treated with inactive drugs despite results of drug susceptibility tests, indicating that other drugs were still active. Of the five patients who did not receive any treatment, one died within the first month after diagnosis, one was lost before treatment initiation, and three were not treated after the diagnosis of MDR-TB because of the choice of the physician in charge. The crude median case-holding duration was 11.0 mo (range, 4 d to 34.5 mo). If the seven case-patients who died in hospital within the first 2 mo after the diagnosis of MDR-TB were excluded from the analysis, the median case-holding duration would be 15.0 mo (range, 4 d to 34.5 mo). Eighteen case-patients (35.3%) failed to keep their clinic appointments at least once during the treatment course, and 16 (31.4%) were lost by the physician in charge after a median duration of 8.9 mo (range, 0 to 20 mo). Surprisingly, case-patients with MDR-TB were not more likely than control-patients to have been hospitalized to initiate tuberculosis treatment, to have a longer duration of initial hospitalization, to have been managed in long-term care facilities, and to have a longer duration of case-holding (Table 2). However, the total number of days in hospital was longer for case-patients (33 versus 19 d; p 5 0.06). The case-patients were more likely than the control-patients to fail to keep their clinic appointments (OR 5 3.4; CI95, 1.2 to 10.5), to report treatment discontinuation (OR 5 4.1; CI95, 1.5 to 11.4), to be lost before treatment completion (OR 5 22.9; CI95, 3.2 to 977.2), to be treated with regimens that included less than two (OR 5 3.8; CI95, 1.2 to 13.5) or less than three (OR 5 10.4; CI95, 3.3 to 37.9) active drugs. In multivariate analysis using logistic regression, three characteristics remained associated with MDR-TB treatment and management: regimens that included less than three active drugs (OR 5 101.5; CI95, 16.3 to 632.3), treatment discontinuation (OR 5 4.4; CI95, 1.4 to 13.8), and duration of treatment (in months) (OR 5 1.3; CI95, 1.1 to 1.5). Outcome. Vital status by December 31, 1996 was known for six of the 16 case-patients who were lost by the physician in charge. Three were alive with culture-negative clinical specimens and were considered as having a favorable outcome, one was alive with the last known clinical specimens to be culture-

Odds Ratio (CI95)

p Value

— — — —

0.80 0.06 0.68 0.44

0.5 (0.1–1.9) Reference 1.1 (0.4–3.1) 1.7 (0.5–6.4) 3.4 (1.2–10.5) 22.9 (3.2–977.2) 3.3 (1.2–9.4) 3.8 (1.2–13.5) 10.4 (3.3–37.9) 7.8 (2.3–29.5) — 0.05 (0.01–0.2)

0.25 — 0.83 0.39 0.02 , 0.001 0.02 0.01 , 0.001 , 0.001 0.02 , 0.001

positive, one was alive with an unknown bacteriologic status, and one died while still having culture-positive clinical specimens. The remaining 10 case-patients lost before treatment completion were all foreign-born, and six of them had returned to their country of birth, including five who were still culture-positive at the time of their departure. Finally, the outcome at least 2 yr after initiation of antituberculosis medication for MDR-TB was obtained for 41 casepatients. Favorable clinical and bacteriologic response was recorded in 17 (41%), and treatment failure was recorded in 24 (59%). Of the latter, 10 (42%) were foreign-born and 14 (58%) were French-born, and only five (21%) were still alive. Among the 51 control-patients, 45 (88%) had favorable outcome, one (2%) was lost to follow-up, and five (10%) died before treatment completion, including two HIV-coinfected patients, one 86-yr-old patient who died from tuberculosis, one 75-yr-old patient who died from neoplasia, and the last one from a traffic fatality. The median duration of survival of MDR-TB cases was 30.7 mo (CI95, 25.2 to 36.1). There was an excess mortality among patients with MDR-TB compared with patients with drug-susceptible tuberculosis (Logrank test 5 15.45, p , 0.001). MDRTB case-patients coinfected with HIV had a shorter duration of survival (median, 2.1 mo; CI95, 0.9 to 3.3) than did drug-susceptible HIV-coinfected control-patients (median . 36 mo). Second Case-Control Study

Only 47 control-patients (second controls) with country of birth similar to that of the corresponding case-patients were eligible for the second case-control study. Compared with the 47 MDRTB case-patients, the 47 second controls had similar baseline characteristics with regard to age, sex, localizations of tuberculosis, socioeconomic status, and housing. However, casepatients were more likely than the second controls to have relapse/chronic tuberculosis (OR 5 21.1; CI95, 6.6 to 71.3), extension of pulmonary tuberculosis to more than one lobe (OR 5 4.1; CI95, 1.5 to 11.1), and positive smear (OR 5 3.8; CI95, 1.3 to 12.3). After multivariate analysis, two baseline characteristics remained associated with MDR-TB: relapse/chronic

591

Flament-Saillour, Robert, Jarlier, et al.: Outcome of MDR-TB in France TABLE 3 CHARACTERISTICS ASSOCIATED WITH TREATMENT FAILURE AMONG THE 41 MDR-TB CASES WITH KNOWN OUTCOME Variables Categorical variables Foreign-born HIV-coinfection Acquired resistance MDR status knowledge (Yes) Hospitalization for treatment Failed to keep clinic appointment Treatment discontinued . 1 mo Treatment with less than two active drugs Treatment with less than three active drugs M. tuberculosis susceptible to less than two drugs M. tuberculosis susceptible to less than three drugs Continuous variables Median duration of: Total hospitalization, d Case-holding, mo Treatment with two active drugs, months

Failure (n 5 24) (%)

Success (n 5 17) (%)

Odds Ratio (CI95)

p Value

10 (41.7) 7 (29.2) 19 (79.2) 15 (62.5) 17 (70.8) 9 (37.5) 15 (62.5) 13 (54.2) 18 (75.0) 5 (20.8) 11 (45.8)

12 (70.6) 0 (0) 8 (47.1) 4 (23.5) 16 (94.1) 5 (29.4) 6 (35.3) — 3 (17.6) 0 (0) 1 (5.9)

0.3 (0.06–1.3) — 4.3 (0.9–21.3) 5.4 (1.1–28.9) 0.2 (0.01–1.5) 1.4 (0.3–7.0) 3.1 (0.7–13.7) — 14.0 (2.5–95.1) — 13.5 (1.5–618)

0.07 0.02 0.07 0.01 0.11 0.60 0.09 , 0.001 , 0.001 0.07 0.006

23.5 3.2 0

55.0 26.0 14.0

0.02 , 0.001 , 0.001

tuberculosis (OR 5 30.5; CI95, 8.3 to 111.7), and lesions of more than one lobe (OR 5 3.9; CI95, 1.1 to 14.5). Positive smear remained associated with MDR-TB, but the association was not statistically significant (OR 5 3.4; CI95, 0.8 to 14.1). Data on patient management were similar to those of the first case-control study with regard to number of patients hospitalized for treatment, duration of initial hospitalization, number of patients with treatment discontinuation, and failure to keep clinic appointments. There were six patients lost before treatment completion, all foreign-born, among the second controls. Among the 41 second controls who were not lost, one died before treatment completion and the remaining 40 had favorable outcomes. Case-patients were more likely than the second controls to have a longer duration of total hospitalization (38 versus 19 d; p 5 0.03) and, though not significantly, to receive regimens that included less than two active drugs (OR 5 2.6; p 5 0.07). In multivariate analysis, the characteristics that remained associated with management of MDR-TB were treatment regimens that included less than two active drugs (OR 5 5.6; CI95, 1.7 to 18.6) and duration of treatment (OR 5 1.02 for 1 mo; CI95, 1.0 to 1.04).

within the 4 mo after the diagnosis of MDR-TB. Case-patients treated with three or more active drugs had a median survival time longer than those treated with less than three active drugs (. 36 mo versus 22.2 mo, p , 0.001). Kaplan-Meier analysis after stratification on treatment with less than two active-drug regimens yielded similar results (data not shown). After multivariate analysis using the Cox model, HIV-coinfection was strongly associated with a poorer survival with a hazard ratio (HR) of 41.2 (CI95, 9.6 to 176.4). Inadequacy of treatment, i.e., treatment with less than three active drugs, and not the number of drugs to which M. tuberculosis was still susceptible, was highly associated with a poorer survival. Surprisingly, previous knowledge of the MDR status was also highly correlated with a shorter outcome (HR 5 3.3; CI95, 1.2 to 8.7). Failure to keep clinic appointments and country of origin were not significantly associated with a poorer outcome in this model. Similarly, in a second model that included treatment regimens containing less than two active drugs instead of regimens containing less than three active drugs, treatment regimens containing less than two active drugs were also related to a poorer survival (HR 5 9.9; CI95, 3.8 to 25.2).

Risk Factors for Treatment Failure among MDR-TB Cases

DISCUSSION

Factors associated with treatment failure in the 41 MDR-TB case-patients with a known outcome are summarized in Table 3. Foreign-born case-patients had a lower risk of treatment failure than did French-born case-patients, but the association was not statistically significant (OR 5 0.3; CI95, 0.06 to 1.3). Case-patients with treatment failure were more likely to be HIV-coinfected (p 5 0.02), to have secondary MDR-TB (OR 5 4.3; CI95, 0.9 to 21.3), and previously known MDR-TB status (OR 5 5.4, CI95; 1.1 to 28.9), to receive less than two (p , 0.001) or less than three (OR 5 14.0; CI95, 2.5 to 95.1) active drugs. Treatment discontinuation for more than 1 mo and absence of hospitalization for treatment were not significantly associated with treatment failure. There was an excess of mortality among HIV-coinfected case-patients compared with other case-patients (Logrank test 5 28.0, p , 0.001) (Figure 2). The median survival time of the former was 2.1 mo (CI95, 0.9 to 3.3) compared with 32.3 mo (CI95, 26.7 to 37.9) for the latter. Out of eight HIV-coinfected case-patients, seven died

The results of the present case-control study conducted to determine the management and the outcome of the 51 cases of MDR-TB reported in France during the year 1994 provided information that calls for Public Health action. First, the management of the MDR-TB cases has not been dramatically different from that of the cases with drug-susceptible M. tuberculosis. Second, despite considerable efforts made to retrieve all cases, the outcome of only 41 (80%) cases has been sucessfully assessed; and third, only 17 of the latter 41 (41%) cases have had a favorable outcome. In France, the diagnosis and treament of tuberculosis are free of charge and almost everywhere decentralized and integrated into general health services. This explains why the 51 MDR-TB cases reported during the year 1994 have been managed in 42 different clinical sites located throughout France, a majority of them having no special dedication to patients with difficult-to-treat tuberculosis. As a consequence, the great majority of MDR-TB patients were managed in the same way as other patients with tuberculosis with drug-susceptible organ-

592


Figure 2. Kaplan-Meier curves for survival of MDR-TB cases comparing HIV-negative cases and HIV-positive cases (A), and cases who received three or more active drugs with those receiving less than three active drugs (B) (1: censored cases).

isms. As usual, after a short period of hospitalization, if any, to establish the diagnosis and initiate standard short-course chemotherapy, the drug treatment had been ambulatory and selfadministered. The initial period of hospitalization was not significantly different for MDR-TB patients and control-patients. Only the total duration of hospitalization, though very short, had been on average significantly longer in MDR-TB patients (33 d) than in control-patients (19 d). As the case-control study was conducted retrospectively in 42 clinical sites, we had to deal with nonstandardized data. In addition, because the follow-up of patients had not been planned, 10 (20%) of the 51 MDR-TB patients versus one (2%) of the 51 first control-patients, and six (12.8%) of the second controls matched on region of birth, were lost before treatment completion. As expected, a majority of the patients lost to follow-up were foreign-born: all 10 MDR-TB patients and all six second controls, whereas the single first control was French-born. A high percentage of lost to follow-up has also been observed among MDR-TB patients in other reported studies. For example, 13% of 171 HIV-negative MDR-TB patients hospitalized in Denver from 1973 to 1983 (13), and 19% of 107 HIV-negative MDR-TB patients hospitalized in a Korean tuberculosis hospital from 1993 to 1995 (14) were lost to follow-up. In The Netherlands (15), among the 19 MDR-TB

VOL 160

1999

patients reported in 1993 and 1994, including 15 foreign-born, six (32%) were reported as defaulters and four (21%) left the country. In contrast, in New York City, 25 of the 26 HIVnegative patients with MDR-TB hospitalized between March 1991 and September 1994 in seven hospitals remained available for outcome analysis (16). The 41% of favorable responses to treatment in our cohort may be considered as poor compared with an 82% cure rate obtained in Korea (14), to at least 64% in New York (16), and to 100% among 11 HIV-coinfected MDR-TB cases treated at the Beth Israel Medical Center in New York City (17). It is still lower than the 56% of favorable response in the Denver cohort (13), and the 60% cure rate (nine of 15) among the MDR patients who did not leave The Netherlands before completion of treatment (15). It might even be not better than the 45% survival rate observed among the 173 MDR-TB patients, including 52% of HIV seropositive, referred to Bellevue Hospital Center in New York between 1983 and 1993 (18). In most of the reports on treatment outcome of MDR-TB (13, 14, 16, 17, 19), cure rates were related to the low number of drugs to which the isolates were still susceptible. In our study, we also observed that the cure rate was nil among the 13 patients who received less than two active drugs, and 61% among the 28 patients who received at least two active drugs. But, as in another study (18), we did not observe any relationship between the cure rate or the survival time and the number of drugs to which the isolates were susceptible. Actually, our investigation revealed that the mean number of drugs to which the isolates were still susceptible was high. For example, 10 of the 13 isolates from the MDR-TB patients who received less than two active drugs were still susceptible to at least two drugs. Consequently, one of the major causes for the high failure rate was the lack of prescription of available active drugs and not the extension of drug resistance. Among the explanations for such a finding are that, first, susceptibility tests to second-line drugs were not performed for all of the MDR strains because the policy is not to perform additional susceptibility testing without specific indication given to the laboratory by the clinician in charge of the patients, an indication rarely given; second, some laboratories did not know for which second-line drugs susceptibility tests had to be performed; third, in a number of cases, the physician in charge was not aware of the laboratory results; fourth, there was a waning interest and expertise for the management of MDR-TB patients. Although MDR-TB patients are an epidemiologic threat to the community, it is not easy to improve significantly their management because of their relatively limited number in France as in other western European countries. One cannot expect all respiratory departments and microbiology laboratories of all university hospitals to keep or develop an expertise in the management of MDR-TB when the number of affected patients is so small. A reasonable solution would be to develop a national specialized unit or team for the treatment of MDR-TB, as recommended in the WHO guidelines for the management of drug-resistant tuberculosis (20, 21). This solution will be submitted to the French Public Health Authorities. Its implementation will require the close collaboration with the associations of medical specialists involved (22). Acknowledgment : The writers wish to gratefully acknowledge all of their colleagues for sharing their data and D. Costagliola for statistical advice.

References 1. Centers for Disease Control and Prevention. 1990. Nosocomial transmission of multidrug resistant tuberculosis to health care workers and

593

Flament-Saillour, Robert, Jarlier, et al.: Outcome of MDR-TB in France

2.

3. 4.

5.

6.

7.

8.

9.

10.

11.

HIV-infected patients in a urban hospital-Florida. M.M.W.R. C.D.C. Surveill. Summ. 39:718–722. Frieden, T. R., T. Sterling, A. Pablos-Mendez, J. O. Kilburn, G. M. Cauthen, and S. W. Dooley. 1993. The emergence of drug-resistant tuberculosis in New York City. N. Engl. J. Med. 328:521–526. Miskinis, K., and B. Purvaneckiène. 1995. Drug-resistant tuberculosis in Lithuania. Tuber. Lung Dis. 76(Suppl. 2):100. Pablos-Méndez, A., M. C. Raviglione, A. Laszlo, N. Binkin, H. L. Rieder, F. Bustreo, D. L. Cohn, C. S. B. Lambregts-van Weezenbeek, S. Kim, P. Chaulet, and P. Nunn. 1998. Global surveillance for antituberculosis-drug resistance 1994–1997. N. Engl. J. Med. 338:1641–1649. Beck-Sagué, C., S. W. Dooley, M. D. Hutton, J. Otten, A. Breeden, J. T. Crawford, A. E. Pitchenik, C. Woodley, G. Cauthen, and W. R. Jarvis. 1992. Hospital outbreak of multidrug-resistant Mycobacterium tuberculosis infections. Factors in transmission to staff and HIV-infected patients. J.A.M.A. 268:1280–1286. Fischl, M. A., G. L. Daikos, R. B. Uttamchandani, R. B. Poblete, J. N. Moreno, R. R. Reyes, A. M. Boota, L. M. Thompson, T. J. Cleary, M. J. Saldana, and S. Lai. 1992. Clinical presentation and outcome of patients with HIV infection and tuberculosis caused by multiple-drugresistant bacilli. Ann. Intern. Med. 117:184–190. Pearson, M. L., J. A. Jereb, T. R. Frieden, J. T. Crawford, B. J. Davis, S. W. Dooley, and W. R. Jarvis. 1992. Nosocomial transmission of multidrug-resistant Mycobacterium tuberculosis: a risk to patients and health care workers. Ann. Intern. Med. 117:191–196. Grosset, J., D. Trystram, A.-C. de Benoist, V. Schwoebel, V. Vincent, M. C. Gutierrez, C. Truffot-Pernot, and J. Robert. 1998. La tuberculose à bacilles multirésistants en France en 1995. Bull. Epid. Hebdo. 13:53–54. Canetti, G., N. Rist, and J. Grosset. 1963. Mesure de la sensibilité du bacille tuberculeux aux drogues antibacillaires par la méthode des proportions. Rev. Tuberc. Pneumol. 27:217–272. Canetti, G., W. Fox, A. Khomenko, H. T. Mahler, N. K. Menon, D. A. Mitchison, N. Rist, and N. A. Smelev. 1969. Advances in techniques of testing mycobacterial drug sensitivity, and the use of sensitivity tests in tuberculosis control programmes. Bull. World Health Organ. 41: 21–43. Mehta, C. R., N. R. Patel, and R. J. Gray. 1985. Computing an exact con-

12. 13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

fidence interval for the common odds ratio in several 2 3 2 contingency tables. J. Am. Stat. Assoc. 80:969–973. World Health Organization. 1997. Treatment of tuberculosis: guidelines for national programmes. WHO/TB/97.220. Geneva. Goble, M., M. D. Iseman, L. A. Madsen, D. Waite, L. Ackerson, and C. R. Horsburgh. 1993. Treatment of 171 patients with pulmonary tuberculosis resistant to isoniazid and rifampin. N. Engl. J. Med. 328:527– 532. Park, S. K., C. T. Kim, and S. T. Song. 1998. Outcome of chemotherapy in 107 patients with pulmonary tuberculosis resistant to isoniazid and rifampin. Int. J. Tuberc. Lung Dis. 2:877–884. Lambregts-van Weezenbeek, C. S. B., H. M. Jansen, N. J. D. Nagelkerke, B. van Klingeren, and J. Veen. 1998. Nationwide surveillance of drug-resistant tuberculosis in The Netherlands: rates, risk factors and treatment outcome. Int. J. Tuberc. Lung Dis. 2:288–295. Telzak, E. E., K. Sepkowitz, P. Alpert, S. Mannheimer, F. Medard, W. El-Sadr, S. Blum, A. Gagliardi, N. Salomon, and G. Turett. 1995. Multidrug-resistant tuberculosis in patient without HIV infection. N. Engl. J. Med. 333:907–911. Salomon, N., D. C. Perlman, P. Friedmann, S. Buchstein, B. N. Kreiswirth, and D. Mildvan. 1995. Predictors and outcome of multidrug-resistant tuberculosis. Clin. Infect. Dis. 21:1245–1252. Park, M. M., N. W. Schluger, H. Cohen, and W. N. Rom. 1996. Outcome of MDR-TB patients, 1983–1993. Prolonged survival with appropriate therapy. Am. J. Respir. Crit. Care Med. 153:317–324. Turett, G. S., E. E. Telzak, L. V. Torian, D. Alland, I. Weisfuse, and B. A. Fazal. 1995. Improved outcomes for patients with multidrugresistant tuberculosis. Clin. Infect. Dis. 21:1238–1244. Crofton, J., P. Chaulet, and D. Maher. 1997. Guidelines for the management of drug-resistant tuberculosis. World Health Organization, Geneva. Iseman, M. D. 1998. MDR-TB and the developing world—a problem no longer to be ignored: the WHO announces ‘DOTS Plus’ strategy. Int. J. Tuberc. Lung Dis. 2:867. Lambregts-van Weezenbeek, C. S. B. 1998. Drug-resistant tuberculosis in the Netherlands: trifle or threat? University of Amsterdam, Amsterdam.