Molecular Characterization and Drug Resistance Patterns of Strains of Mycobacterium tuberculosis Isolated from Patients in an AIDS Counseling Center in Port-au-Prince, Haiti: a 1-Year Study

Abstract

Tuberculosis (TB) is one of the most common opportunistic diseases that appear among human immunodeficiency virus (HIV)-positive patients in Haiti. In this context the probable emergence of multidrug-resistant (MDR) strains of Mycobacterium tuberculosis is of great epidemiological concern. However, as routine culture of M. tuberculosis and drug susceptibility testing are not performed in Haiti, it has not been possible so far to evaluate the rate of drug resistance among M. tuberculosis isolates from circulating TB cases. This report describes the first study on the molecular typing and drug resistance of M. tuberculosis isolates from patients with culture-positive pulmonary tuberculosis monitored at the GHESKIO Centers in Haiti during the year 2000. Clinical, epidemiological, and drug susceptibility testing results were available for 157 patients with confirmed cases of TB, with a total of 8.9% of patients harboring MDR M. tuberculosis. A significant association between the occurrence of resistance and previous TB treatment was observed (P < 0.001), suggesting that a previous history of TB treatment was a risk factor associated with MDR TB in Haiti. The DNAs of individual isolates from 106 samples were available and were typed by spoligotyping and determination of the variable number of tandem DNA repeats. Both typing methods provided interpretable results for 96 isolates, and the clusters observed were further confirmed by ligation-mediated PCR to define potential cases of active transmission. Thirty-three (34%) of the isolates were found to be grouped into 11 clusters with two or more identical patterns. However, an assessment of risk factors (sex, HIV positivity, previous treatment, drug resistance) showed that none was significantly associated with the active transmission of TB. These observations suggest that acquired MDR TB is prevalent in Haiti and may be associated with compliance issues during TB treatment since prior TB therapy is the strongest risk factor associated with MDR TB. Prevention of TB transmission in Haiti should target active case investigation, routine detection of drug resistance, and adequate treatment of patients. The use of directly observed short-course therapy should be enforced throughout the country; and relapses, reactivations, or newly acquired infections should be discriminated by genotyping methods.

With an estimated incidence of about 200 new cases/100,000 inhabitants, tuberculosis (TB) is the most common infectious disease among human immunodeficiency virus (HIV)-positive patients in Haiti (5). In 1998, it was estimated that more than 50% of the 300,000 estimated HIV-positive patients would develop active TB disease within the following 5 years (5). Effective control of disease progression has been hampered by unfavorable socioeconomic conditions. According to a recent report (http://www.paho.org/English/SHA/prflhai.htm), Haiti has the highest population density of all Latin American countries, with 260 inhabitants/km², for a total population of over 8 million people; more than one-third live in the capital, Port-au-Prince. The country has a 70% unemployment rate, and the fertility rate is estimated to be 4.8 children per woman.

The recent emergence of multidrug-resistant (MDR) strains of Mycobacterium tuberculosis is of great epidemiological concern; however, the actual surveillance system does not allow accurate estimation of the proportion of drug-resistant bacilli, as the organisms are not cultured routinely and drug susceptibility testing is not performed due to a lack of resources. Consequently, exact figures permitting assessment of TB incidence, evaluation of risk factors associated with active transmission of the disease, and the prevalence of drug resistance in Haiti are not available.

The study described here aimed to build capacity for the development of a national mycobacteria reference laboratory at the GHESKIO Centers, which are already the national reference centers for HIV, sexually transmitted diseases, and diarrheal diseases. Under this program personnel were trained at the Institut Pasteur of Guadeloupe and systematic screening of patients presenting cough, weight loss, and fever began in January 2000 (4). All people presenting with cough are routinely evaluated for TB with a history and physical examination, three sputum smears for acid-fast bacilli (AFB), and a chest radiograph. Sputum cultures for M. tuberculosis are performed for all patients with a positive AFB smear and all HIV-positive patients with wasting syndrome and a negative AFB smear. All patients found to have a positive culture for TB were eligible for the present study. After obtaining informed written consent in Creole, volunteers were enrolled in the study. A complete history and physical examination were taken for all subjects, and they all answered the questions on an epidemiological questionnaire addressing various questions including whether they had previously undergone treatment for TB. All isolates obtained in cultures were systematically tested for their susceptibilities to four first-line antituberculous drugs. In parallel, the strains were typed by three PCR-based genotyping methods in association, i.e., spoligotyping, followed by typing by determination of the variable number of tandem DNA repeats (VNTRs) and ligation-mediated PCR (LM-PCR). Indeed, methods that use molecular markers constitute a complementary tool for epidemiological investigations; they not only confirm transmission between patients with suspected epidemiological connections but also provide information about transmission between patients not suspected to have epidemiological links. Among the methods that use molecular markers, IS6110-based restriction fragment length polymorphism analysis has been widely used for molecular typing of M. tuberculosis (19). However, a combination of spoligotyping and LM-PCR has been proposed as an alternative to the IS6110-based restriction fragment length polymorphism technique (3). An optimal association of PCR methods is of great interest for epidemiological studies, as it saves time, is economical, and requires only minute amounts of DNA (12, 16). The aim of this study was to evaluate the proportion of drug-resistant M. tuberculosis isolates in Port-au-Prince, Haiti, and to detect possible epidemiological links not highlighted by routine surveys.

MATERIALS AND METHODS

Patients and bacterial isolates.

Cultures of sputum for detection of M. tuberculosis are performed for all patients with a positive AFB smear and for all HIV-positive patients with wasting syndrome regardless of their AFB smear results. The population studied comprised patients in Port-au-Prince from whom at least one sample was positive for M. tuberculosis by sputum culture during the year 2000. Cultures were done at the GHESKIO laboratory on Löwenstein-Jensen medium after decontamination with sodium lauryl sulfate. Drug susceptibility testing of all isolates was performed by the 1% proportion method and included testing for susceptibility to isoniazid (1 μg/ml), rifampin (1 μg/ml), and streptomycin (2 μg/ml) on Middlebrook 7H11 medium and testing for susceptibility to ethambutol (10 μg/ml) on Löwenstein-Jensen medium.

DNA preparation and molecular typing.

The DNAs were prepared at the GHESKIO laboratory by the cetyltrimethylammonium bromide (Merck, Darmstadt, Germany) method as described previously (21) and shipped dry at room temperature to the Institut Pasteur laboratory at Guadeloupe according to the International Air Transport Association guidelines. After the DNAs were received they were resuspended and stored in TE buffer (10 mM Tris, 1 mM EDTA [pH 8]) at 4°C.

Spoligotyping.

The spoligotyping method was performed with primers DRa and DRb as reported previously (10). Detection of hybridizing DNA was done with chemiluminescent ECL detection liquid (enhanced chemiluminescence detection kit; Amersham, Little Chalfont, England), followed by exposure to X-ray film (Hyperfilm ECL; Amersham), in accordance with the instructions of the manufacturer.

VNTR typing method.

The VNTR typing method was performed as described previously, with slight modification (7). PCRs were run in a Perkin-Elmer GeneAmp PCR system 9600 (Perkin-Elmer, Norwalk, Conn.). An aliquot of 20 μl from the reaction tubes was run on a 2% agarose gel, and a 100-bp ladder was run every six lanes. The images were digitized by using the Videocopy system (Bioprobe, Montreuil, France), and determination of the molecular weights of the PCR fragments was performed with Taxotron software (Taxolab, Institut Pasteur, Paris, France). The number of copies for each exact tandem DNA repeat (ETR) was deduced by a previously published scheme (7), and the data were documented as five-digit numbers representing allele profiles ETR-A to ETR-E.

LM-PCR.

We followed a recently described procedure for the LM-PCR (12). Briefly, DNA was digested with SalI, and an asymmetrical double-stranded oligonucleotide was ligated to the cut ends. The linker is an ordinary DNA strand that is stable under ligation conditions but not at the temperature used for the PCR. The linker primer site therefore does not remain connected to the template DNA strand during PCR and therefore cannot serve as a binding site for the linker primer. Restriction fragments containing the IS6110 sequence were amplified by using an IS6110-specific primer, and the linker primer amplified DNA-tagged fragments containing the IS6110-flanking sequence on the 5′ side. The amplified products were separated in a 2% agarose gel. The images were digitized by using the Videocopy system (Bioprobe), and the molecular weights of the PCR fragments were determined with Bionumerics software (version 2.5; Applied Math, Sint-Martens-Latem, Belgium) by using the Dice coefficient.

Combined numerical analysis.

The combined numerical analysis was performed with Bionumerics software. Each file with experimental data from the spoligotyping, VNTR, and LM-PCR analyses was merged as a composite data set in the Bionumerics database, with the similarity coefficient option taken from each experiment. The matrices from the individual experiments were averaged according to the same defined weight, and an individual similarity matrix was calculated in such a way that all characters had an equal influence on similarity. A dendrogram was drawn by using the unweighted pair group method with arithmetic averages (UPGMA) with a tolerance of 1%. Under these conditions a “cluster” of strains was considered clonal if the similarity was ≥98% by all three methods. All clustering events were also visually confirmed in parallel. From the pattern matching, an estimate of clustering was done, and recent transmission was estimated by the formula T(c) − N(c)/T(a), where T(c) is the total number of clustered isolates, N(c) is the number of clusters, and T(a) is the total number of isolates (15).

Approval of the study.

The study protocol was approved by the Comité des Droits Humains de GHESKIO, the local institutional review board, and the institutional review board of the Weill Medical College of Cornell University, New York, N.Y.

RESULTS

Patient characteristics and drug resistance patterns.

The ages of the patients included in this study ranged from 14 and 68 years (mean, 31 years). Drug susceptibility results were available for 157 isolates; 32 (20.4%) patient isolates were resistant to one or more drugs (Table 1). A significant association between the occurrence of resistance and previous TB treatment was observed (P < 0.001). Whereas 15.8% of untreated patients had drug-resistant TB (the M. tuberculosis isolates were resistant to one or more drugs), as many as 55.5% of treated patients harbored drug-resistant bacilli. As summarized in Table 1, rates of resistance to any single drug or a combination of drugs, including the proportion of MDR isolates, were considerably higher for previously treated patients. Fourteen (8.9%) patients were infected with an MDR strain (resistance to at least isoniazid and rifampin), a value comparable to the proportions of MDR strains in the neighboring country of the Dominican Republic (11). Whereas 6 of 139 (4.3%) cases of MDR were found among patients with new cases, as many as 8 of 18 (44.4%) cases of MDR were found among retreated patients (Table 1). The rate of resistance to any drug and the prevalence of HIV infection were highest in those ages 35 to 44 years, i.e., 25.8 and 64.5%, respectively (the values were 14.9 and 14.9%, respectively, for those younger than 25 years of age; 38.1 and 19.0%, respectively, for those ages 25 to 34 years; and 43.8 and 12.5%, respectively, for those older than 44 years of age). However, the occurrence of resistance was not significantly related to either HIV infection (22 of 99 resistant isolates were from HIV-negative patients and 10 of 58 resistant isolates were from HIV-positive patients [P < 0.65]) or age and sex (data not shown).

TABLE 1.

Drug resistance in M. tuberculosis isolates from Haitian patients with new and previously treated TB cases

Observationsa	Total	No. (%) of isolates from patients with:
		New TB cases	Previously treated TB cases
Total cases	157 (100)	139 (88.5)	18 (11.5)
Male patients	84 (53.5)	73 (52.5)	11 (61.1)
Female patients	73 (46.5)	66 (47.5)	7 (38.9)
HIV-positive patients	58 (36.9)	55 (39.6)	3 (16.7)
HIV-negative patients	99 (63.1)	84 (60.4)	15 (83.3)
Resistance to any drug	32 (20.4)	22 (15.8)	10 (55.5)
HIV-positive cases with resistant isolates	10 (6.4)	9 (6.5)	1 (5.5)
HIV-negative cases with resistant Isolates	22 (14.0)	13 (9.3)	9 (50.0)
Resistance to one drug only	11 (7.0)	9 (6.5)	2 (11.1)
INH	6 (3.8)	5 (3.6)	1 (5.6)
RIF	4 (2.5)	3 (2.1)	1 (5.6)
SM	1 (0.6)	1 (0.7)	0
EMB	0	0	0
Resistance to more than one drug (excluding MDR TB)	4 (2.5)	4 (2.9)	0
INH + SM	2 (1.3)	2 (1.4)	0
INH + EMB	2 (1.3)	2 (1.4)	0
Resistance to more than one drug (MDR TB)	14 (8.9)	6 (4.3)	8 (44.4)
INH + RIF	4 (2.5)	2 (1.4)	2 (11.1)
INH + RIF + SM	4 (2.5)	1 (0.7)	3 (16.7)
INH + RIF + EMB	3 (1.9)	2 (1.4)	1 (5.6)
INH + RIF + SM + EMB	3 (1.9)	1 (0.7)	2 (11.1)

^aINH, isoniazid; RIF, rifampin; SM, streptomycin; EMB, ethambutol.

DNA polymorphisms of M. tuberculosis strains.

The DNAs of individual isolates were available from 106 samples and were typed by spoligotyping and the VNTR method. Interpretable results were obtained for 96 isolates by both typing methods, and the clusters observed were further confirmed by LM-PCR to define potential cases of active transmission. The clustering of 96 isolates for which DNA was available on the basis of the various typing methods is summarized in Table 2. Detailed results for spoligotyping and VNTR analysis are shown in Fig. 1, and those of combined numerical analysis following spoligotyping, VNTR analysis, and LM-PCR are illustrated in Fig. 2.

TABLE 2.

Clustering of M. tuberculosis isolates in Port-au-Prince based on spoligotyping, the VNTR method, and LM-PCR

Method	No. of distinct profiles	No. of clusters [N (c)]	No. of clustered isolates [T (c)]	% Clustered isolates	No. of unique types	Largest cluster	Smallest cluster	Estimate of recent transmission rate	HGDIa
Spoligotyping (n = 96)	43	14	67	70	29	12	2	55	0.957
VNTR (n = 96)	28	12	80	83	16	28	2	70	0.882
Spoligotyping + VNTR (n = 96)	61	13	48	50	48	9	2	36	0.980
Spoligotyping + VNTR + LM-PCR (n = 48)	26	11	33	69	15	6	2	46	0.991

^aHGDI, Hunter Gaston discriminatory index (9).

FIG.1.

Combined numerical analysis of data obtained by spoligotyping and the VNTR method for 96 M. tuberculosis clinical isolates from Haiti. The bar represents the similarity index obtained by the UPGMA method. Column I, spoligotype patterns; column II, VNTR method data; column III, isolate number; column IV, spoligotype pattern according to spoligotype database designations; column V, VNTR results in a numerical format. The different shades of grey from lighter to darker in column II provide a visual representation of the variations in the number of ETR alleles from 1 to 5. The letters A to M indicate M. tuberculosis clusters identified by spoligotyping and VNTR analysis.

FIG. 2.

Combined numerical analysis based on three genotyping methods, i.e., spoligotyping, the VNTR method, and LM-PCR. Only 48 of 96 isolates found to be previously clustered by spoligotyping and VNTR (Fig. 1) were retained for this analysis. A total of 11 clusters containing isolates from two to six patients can be seen. The bar represents the similarity index obtained by the UPGMA method. Column I, spoligotype pattern; column II, LM-PCR fingerprints; column III, isolate number; column IV, spoligotype number according to the worldwide spoligotype database; column V, VNTR results in a numerical format. Clusters that were not further subdivided upon LM-PCR were designated with the same letter, as in Fig. 1; on the other hand, if a cluster was further subdivided by LM-PCR, the original letter designation was followed by a number, e.g., clusters A1, A2, C1, D1, and F1.

Spoligotyping generated 43 distinct patterns among the 96 isolates that were typed (Table 2; Fig. 1); a total of 67 (70%) isolates were grouped into 14 clusters, whereas 29 (30%) remained unclustered (Table 2). The spoligotype designation was attributed by comparison of the patterns that we obtained with those included in an international spoligotype database that provides information on the shared-type distributions of M. tuberculosis spoligotypes worldwide (17). At the time of this comparison, that database had compiled data on 11,160 isolates from 90 countries, comprising 817 shared types (shared by two or more patients) and 1,546 orphan isolates. New shared types were created when the pattern for an isolate with an orphan pattern from this study was similar to that for another orphan isolate in the database (e.g., type 455). In this way, we were able to add to the 67 clustered isolates from this study another 14 isolates from Haiti with orphan patterns that were already present in the databases and attributed new shared-type designations (i.e., types 1, 5, 52, 65, 77, 150, 151, 175, 205, 242, 245, 294, 455, and 801). Thus, a total of 81 isolates belonged to 28 spoligotyping-defined clusters and 15 isolates presented orphan patterns. Most of the shared types found in Haiti were previously reported from the Americas (52%; particularly North America [37%]) and Europe (24%). Some shared types found in Haiti were previously reported only from Europe (types 161 and 294) or the United States (types 295 and 455).

Typing by VNTR analysis alone was less discriminatory than spoligotyping, as it generated 28 distinct patterns instead of the 43 profiles observed by spoligotyping (Table 2; 80 or 83% of the isolates grouped in 12 clusters and 16 or 17% of the isolates were unclustered). However, combined analysis by spoligotyping and the VNTR method was highly discriminatory and generated 61 distinct patterns (48 isolates were grouped into 13 clusters containing 2 to 9 isolates, and 48 isolates were unclustered [Tables 2, 3, and 4).

TABLE 3.

Distribution of clustered isolates by the various typing methods^a

Spoligotype	Spoligotyping alone			Subclustering by VNTR method		Subclustering by LM-PCR
	No. of isolates	Frequency (%)b	Phylogenetic cladesc	Digit No.	Cluster (no. of isolates)	No. of bands in subclustered isolates	Subclusterd (no. of isolates)
50	12	4.71	Haarlem	32333	A (9)	6	A1 (2)
						3	A2 (2)
				32323	C (3)	3	C1 (2)
2	11	0.95	Haarlem	32333	L (6)	6	L (6)
				32323	M (2)	5	M (2)
				22333	K (2)	2 unrelated patterns	NAe
				21333	1	NA	NA
91	6	0.69	Carib (X3)	32333	D (5)	2	D1 (4)
				32323	1	NA	NA
42	6	3.11	LAM	21433	F (4)	4	F (3)
				21233	1	NA	NA
				21234	1	NA	NA
93	6	0.42	LAM	21433	G (4)	5	G (4)
				21533	H (2)	5	H (2)
17	5	0.94	LAM	22433	J (3)	3 unrelated patterns	NA
				12433	1	NA	NA
				22423	1	NA	NA
70	4	0.31	Carib (X3)	32313	1	NA	NA
				32323	1	NA	NA
				32333	1	NA	NA
				32433	1	NA	NA
53	3	7.43	Tf	22432	1	NA	NA
				22431	1	NA	NA
				22232	1	NA	NA
20	3	1.32	LAM	22433	1 (3)	6	1 (3)
714	3	0.02	LAM	32333	B (3)	2	B (3)
295	2	0.05	Haarlem	32434	E (2)	2 unrelated patterns	NA
193	2	0.07	LAM	22433	1	NA	NA
				22423	1	NA	NA
118	2	0.30	Haarlem	22333	1	NA	NA
				22434	1	NA	NA
34	2	0.72	LAM	32433	1	NA	NA
				42442	1	NA	NA
Orphan	29	NA	NA	NA	NA	NA	NA

^aEpidemiological observations and the links between patients are provided for each cluster in Table 4.

^bFrequency in the wordwide spoligotyping database (14, 17), consulted on 30 March 2002.

^cMajor clades of the tubercle bacilli according to the wordwide spoligotyping database (14, 17). Note that the terms Carib and X3 have been used alternatively to designate the same major clade.

^dSubclusters defined by the VNTR method that were not further subdivided upon LM-PCR were designated with the same letter (e.g., clusters L, M, G, H, I, and B). On the other hand, if a cluster was further subdivided by LM-PCR, the original letter designation was followed by a number (e.g., clusters A1, A2, C1, D1, and F1).

^eNA, not applicable.

^fThe T clade is highly ubiquitous and differs from the Haarlem type 50 by the presence of a single spacer (spacer 31).

TABLE 4.

Epidemiological observations and links

Clustera	New case(s)	Drug resistance pattern	Patient(s)	Median age (yr)	Link(s)
A1	Yes	Susceptible	One HIV-positive male (age, 34 yr) and one HIV-negative female (age, 25 yr)	29.5	None evident
A2	Yes	Susceptible	Two HIV-negative females (ages, 33 and 22 yr, respectively)	27.5	None evident
B	Yesb	Mixedc	Two HIV-negative females (ages, 45 and 18 yr, respectively) and one HIV-negative male (age, 28 yr)	30	The two females were a mother and daughter who lived together.
C1	Yes	Susceptible	One HIV-positive female (age, 23 yr) and one HIV-negative female (age, 27 yr)	25	None evident
D1	Yesd	Susceptible	One HIV-positive male (age, 47 yr) and one HIV-positive female (age, 17 yr) (both new cases)c	31	Infection probably occurred through two HIV-negative patients attending the same facility for TB treatment 12 mo earlier
F1	Yes	Susceptible	Two HIV-negative females and one HIV-negative male	29	None evident
G	Yes	Mixede	One HIV-negative female (age, 34 yr) (a new case of isoniazid monoresistance), one HIV-negative female (age, 31 yr) infected with a drug-susceptible isolate, and two HIV-negative males (ages, 21 and 44 yr, respectively) with MDR TB (resistance to isoniazid, rifampin, and streptomycin)	32.5	No link between infected females was evident; one HIV-negative male was probably infected through the other, who had attended the same facility for TB treatment 12 mo earlier
H	Yes	Susceptible	Two HIV-positive males	39.5	Patients attended the same health care facility
I	Yes	Susceptible	Two HIV-positive males and one HIV-positive female	29	All HIV-positive patients attending the same health care facility
L	Yes	Susceptible	One HIV-negative female (age, 20 yr), one HIV-positive female (age, 26 yr), one HIV-positive male (age, 41 yr), and three HIV-negative males (age range, 22 to 47 yr)	32.5	None evident
M	Yes	Mixedf	Two HIV-negative males (ages, 19 and 26, respectively), one of whom presented with a case of resistance to isoniazid streptomycin	22.5	None evident

^aSee Table 3, footnote d, for explanation of cluster designations.

^bOne new case and two previously treated cases.

^cThe isolates from the women were MDR, and the isolate from the man was susceptible.

^dThree new cases and one previously treated case.

^eNot all patients are described here.

^fNo single drug resistance pattern was observed.

Table 3 summarizes the overall distribution of clustered isolates by the various typing methods used alone or in combination, and the results obtained are presented in decreasing order of the cluster size. The major identifiable spoligotyping-defined phylogenetic clades were identified according to recent publications (14, 17) and showed that only three major families covered all clustered isolates: ubiquitous type Haarlem (30 isolates), the Latin America and Mediterranean or LAM family (27 isolates), and the Carib family reported from the Caribbean area (10 isolates). The patterns obtained by spoligotyping and VNTR in combination were also analyzed and are shown in Table 3. Some of the spoligotyping-defined clusters were easily subdivided upon VNTR typing; e.g., cluster 50 (12 isolates) was subdivided into two groups (combinations 50-32333 and 50-32323, containing 9 and 3 isolates, respectively), and cluster 2 was subdivided into three groups (combinations 2-32333, 2-22333, and 2-32323, containing 6, 2, and 2 isolates, respectively).

Finally, LM-PCR was used to further discriminate the 48 isolates clustered by the combination of spoligotyping and the VNTR method and generated 26 distinct patterns: 33 (69%) isolates grouped in 11 clusters and 15 (31%) isolates were unclustered (Tables 2 and 3; Fig. 2). It is worth mentioning that seven of the clusters defined by spoligotyping plus the VNTR method were further subdivided by LM-PCR, as detailed in Table 3. However, six of all the clusters described by the previous methods (clusters L, M, G, H, I, and B) were undifferentiated.

Characteristics of clustered isolates and associated factors.

Characteristics such as the sex and age of the patients, HIV positivity, previous treatment for TB, and resistance of isolates to drugs were compared for the clustered (n = 33) and unclustered (n = 62) isolates (Table 4). However, none of these characteristics differed significantly between the two groups, as determined by univariate analysis of the risk factors that placed a patient in a TB transmission group (results not shown). For example, the mean age of the patients with clustered isolates was 30 years (range, 18 to 50 years), the sex ratio was 1.2, and 33.3% of the patients were coinfected with HIV. These figures were very similar to those obtained for patients harboring unclustered isolates (mean age, 32 years; age range, 14 to 68 years; sex ratio, 1.1; rate of HIV-TB coinfection, 33.8%. Nonetheless, the rate of MDR TB was slightly higher among patients with clustered isolates (4 of 33 clustered isolates compared to 5 of 62 unclustered isolates; the difference was not significant statistically).

Twenty-six of 33 (79%) of the clustered isolates were pansusceptible to all the first-line drugs, whereas the remaining 7 isolates were associated with various degrees of resistance: one case of isoniazid resistance only, one case of streptomycin resistance only, one case of resistance to two drugs (rifampin and streptomycin), three cases of resistance to three drugs (isoniazid, rifampin, and streptomycin), and one case of resistance to all four first-line drugs. These findings underline the high rates of resistance to first-line drugs among clustered isolates in Port-au-Prince: 6 of 33 (18%) isolates for isoniazid and streptomycin and 4 of 33 (12%) isolates for rifampin. On the other hand, the rate of resistance to ethambutol (1 of 33 [3%] isolates) was significantly lower.

The epidemiological observations and genotyping characteristics of the isolates from the patients harboring 33 isolates in 11 clusters are summarized in Table 4 and show that a definite epidemiological link could be established for only a minority of the patients. Furthermore, only two clusters (clusters B and G) concerned MDR cases of TB, and in each case, it concerned an HIV-negative patient who had previously been treated for TB, suggesting that the MDR TB in this study originated due to a lack of compliance during treatment. This observation is also supported by the fact that no MDR M. tuberculosis strain was isolated from any of the untreated patients harboring clustered isolates. Cluster B represents two previously treated patients (a mother and a daughter) who were infected with MDR isolates presenting similar genetic characteristics. There was no evidence of an epidemiological link for the third person in cluster B. Two other cases of MDR TB among previously treated cases were clustered, on the one hand, with an isolate sensitive to all four drugs and, on the other hand, with an isoniazid-monoresistant isolate (cluster G).

In conclusion, as many as eight of nine (89%) of patients infected with an MDR isolate typed by the three methods described above were previously treated. The only MDR M. tuberculosis isolate in a new case concerned a 42-year-old, HIV-positive female. This strain did not cluster either with any of the isolates in this study or with any of the 11,160 isolates representing the worldwide diversity of M. tuberculosis spoligotypes in the database of the Institut Pasteur. These findings underline the fact that even though the prevalence of MDR TB in Port-au-Prince is high, the rate of active transmission of MDR M. tuberculosis strains appears to be low. For the time being, the emergence of MDR TB appears to be nearly exclusively linked to a lack of patient compliance with anti-TB treatment. Thus, the fact that previously treated patients are often retreated with the same drugs without performing drug susceptibility testing certainly contributes to the emergence of MDR TB in Haiti and sooner or later will lead to the circulation and transmission of these MDR clones of tubercle bacilli.

DISCUSSION

This study presents a first picture of the M. tuberculosis organisms circulating in Port-au-Prince, Haiti, and their drug resistance. The percentage of clustered isolates (34%) observed in this study was slightly lower than those observed in other studies performed in The Netherlands, Denmark, and France and in other large studies performed in the United States (range, 35 to 57%) (2, 8, 15, 20). However, this preliminary study included 157 patients, and DNAs from only 96 individual isolates among the 14,400 estimated new cases of TB per year in Haiti could be fingerprinted. Consequently, it is impossible to establish the transmission chain correctly. Therefore, the preliminary estimate of the recent rate of transmission of TB in Haiti, according to this study (23%), does not reflect the expected incidence of the disease. Indeed, statistical methods have shown that appropriate estimation of the recent transmission rate by molecular biology methods should be evaluated differently by age group and should be performed at least over a period of 2 years (22). Similarly, the 1998 prediction that 50% of the 300,000 estimated HIV-positive patients go on to develop active TB disease within the following 5 years in Haiti (5) could not be evaluated because of the short duration of the study and the relatively small number of patients included. These aspects may be addressed in the near future, as the present study is ongoing in Haiti.

Comparison of spoligotypes from Haiti with those in the worldwide database with the spoligotypes of isolates from 90 countries provided some important clues on the phylogenetic origins of the circulating clades (Table 3). It is noteworthy that all the clustered isolates belonged to three major clades of tubercle bacilli (Haarlem, LAM, and Carib) and probably originated from the Americas (North America, Central and South America, and the Caribbean) and Europe (14, 17). Despite the high degree of strain diversity in the present study, more than half of the M. tuberculosis spoligotypes present in Port-au-Prince were traced to the spoligotypes already reported from neighboring Caribbean islands. Recent studies have underlined the potential interregional transmission of M. tuberculosis between Haiti and its neighboring countries (6, 13). Lastly, among the unclustered isolates, a single genotype Beijing isolate of Asian descent was found in a 25-year-old, HIV-negative male; however, even though the majority of Beijing isolates are often associated with drug resistance, including MDR TB (1), this isolate was pansusceptible.

As reported for other parts of the world (11), we recorded a higher proportion of drug resistance among isolates from previously treated patients (55.5%) compared to the proportion among isolates from patients with new cases (15.8%). Thus, the only risk factor associated with MDR TB in the present setting was retreatment. We have not found any evidence for the active transmission of MDR TB among patients with new cases, at least for the time being. Even though patients in Haiti receive the standard anti-TB treatment, much uncertainty exists as to which drugs should be used and the duration of use, as a lack of compliance during treatment remains the main hurdle to efficient chemotherapy of TB. The dynamics of TB in response to 10 years of intensive control efforts in Peru through intensive short-course chemotherapy showed that compliance with treatment is one of the major factors affecting TB control measures (18).

Another problem is the fact that culture of the tubercle bacilli and drug susceptibility testing are not routinely performed in Haiti. This often leads to the retreatment of patients with the same drug combinations used in the initial phase of treatment, even though the strains may have acquired resistance to one or more drugs during that treatment. Thus, efforts to prevent the transmission of drug-resistant bacilli in Haiti should also target adequate case investigation, culture of the bacilli followed by drug susceptibility determination, and adequate treatment.

In conclusion, even though the incidence of TB in Haiti is high and the rate of MDR TB, which was essentially associated with treatment failures in this study, was high, there appears to be a low rate of active transmission of MDR TB. On the other hand, phylogenetic analysis shows the presence of three major clades, with the potential circulation of M. tuberculosis isolates between Haiti and the Americas. This preliminary study will be continued for an additional 2 years, with culture, drug susceptibility testing, and DNA fingerprinting performed on all strains isolated at the GHESKIO Centers and obtained from the 20 institutions throughout Haiti that the GHESKIO Centers serve. The national data obtained will help to establish priorities for TB control programs in this part of the developing world.