Tuberculosis (TB) is among the top 10 causes of death in the global ranking (1
). Although Peru accounts for only 3% of the population of the Americas, it has 9.5% of the region's TB cases. According to the World Health Organization (WHO), 21,916 new cases of pulmonary TB were reported in Peru between January and November 2013 (2
). Urban areas are more affected, with 59% of all Peruvian TB cases, 82% of multidrug-resistant (MDR) TB cases, and 93% of extensively drug-resistant (XDR) TB cases occurring in Lima. Within the capital area, TB cases are heterogeneously distributed. The most affected districts are located in the northeast and together represent 86% of the reported cases in the capital (3
). San Juan de Lurigancho is the most populated district in this area, with 1,004,339 inhabitants (5
), and reports a pulmonary TB incidence rate of 193 cases per 100,000 inhabitants, a smear-positive TB incidence rate of 126 cases per 100,000 inhabitants (6
), and an overall MDR prevalence of 7% (7
), exceeding the national averages for the three indicators of 103 cases per 100,000 inhabitants, 62 cases per 100,000 inhabitants, and 5.3%, respectively (8
). The HIV prevalence among TB patients in this setting is similar to the national prevalence, which in 2008 was 2.6% (9
Peru has been considered a good example of the beneficial effects of implementing directly observed therapy short course (DOTS) in a country's health system (10
). Nevertheless, national surveys from 1996 and 2006 have shown increases in MDR rates from 2.4% to 5.3% among new cases and from 15.7% to 23.6% among previously treated cases (11
). This paradox of improved TB management and worsening resistance prevalence has been explained by increases in the notified cases, i.e., in case detection, although treatment outcomes remain poor (12
). Basically, the observed increases in MDR-TB rates may result from two factors, i.e., transmission of MDR-TB and acquired resistance due to ineffective TB treatment selecting for spontaneous mutations in specific genes associated with drug resistance (13
Molecular strain typing (genotyping) has significantly contributed worldwide to the understanding of TB epidemiology and transmission dynamics (15
), by confirming outbreaks (17
) and identifying the clonal spread of successful strains, including MDR strains (18
). Furthermore, molecular typing has shown that the Mycobacterium tuberculosis
complex (MTBc) has a diverse population structure, being composed of seven lineages of human importance, subdivided into families (20
) that differ not only in their geographical occurrence but also in their drug resistance profiles (19
). Epidemiological data on TB in Peru have so far consisted largely of reports of the number of cases detected, demographic data for the cases, and drug susceptibility profiles of M. tuberculosis
isolates. Despite the increased number of recent studies on genetic diversity (23–25
), data on the molecular epidemiology of TB in this country are limited; the SpolDB4.0 spoligotyping database includes only 96 Peruvian strains among the 39,609 published entries (26
To gain more insights regarding the extent of ongoing transmission and its contribution to the high prevalence of MDR-TB, we aimed to describe the circulating MTBc genotypes among new pulmonary TB cases. We used a combination of 15-locus myocobacterial interspersed repetitive-unit (MIRU-15)-variable-number tandem repeat (VNTR) analysis and spoligotyping to define the TB population structure and to perform a cluster analysis to measure the level of recent transmission and its possible association with drug resistance.
Despite the decreases in the national TB incidence since the implementation of DOTS, Peru faces increases in the cases of MDR-TB in the poorest districts of the capital. This study presents an in-depth analysis of the population structure of MTBc strains from new pulmonary TB cases in San Juan de Lurigancho, a setting with high TB and MDR-TB burdens in the northeastern part of Lima, Peru. National surveys from 1996 and 2006 showed an increase in the MDR rate from 2.4% to 5.3% among new cases of pulmonary TB, most of which were detected in Lima (11
). The 6.8% MDR-TB rate we observed in our population is in line with these data.
Based on the combination of spoligotyping and MIRU-15 analysis, 75% of the investigated strains were assigned to lineages and families, with lineage 4 being the most prevalent, followed by lineage 2. In contrast to previous reports from Latin America in which LAM was the most common family (11
), our results demonstrated a predominance of the Haarlem family among new cases in the district of San Juan de Lurigancho.
With the exception of a study on MDR-TB household contacts, in which LAM was most frequent (41
), this predominance of the Haarlem genotype has been observed in all Peruvian studies to date (23–25
). Despite heterogeneity in the design and size of those studies, the combined data suggest a trend of increasing prevalence of the Haarlem genotype in Lima in the past decade, with an increase from 23.8% to 35.3% (23
). Among a selection of MDR-TB and XDR-TB isolates collected in Peru in 2007 to 2009, the prevalence reached 43.6% (24
). The widespread occurrence of the genotype in Peru could be due to the stability of the genotype in this population and/or a high rate of recent transmission, as documented by the higher probability of Haarlem strains to be in a cluster in our study. Additionally, our results are in accordance with the finding of LAM as the second most prevalent family in Peru, followed by the Beijing family (23–25
), with a seeming increase in Beijing strains in Lima from around 5% in early 2000 (23
) to around 15% a decade later (42
; this study).
Unlike the situation observed in other South American countries, in which the Beijing family was reported to represent less than 1%, its proportion in Peru was found to be relatively high, i.e., 5.5% (44/794 isolates) for samples obtained between 1999 and 2005 (23
), 5.9% (11/185 isolates) in 1999 (43
), 9.3% for samples obtained between 2004 and 2006 (25
), and 14.1% for samples obtained during 7 months in 2009 (42
). This observation was recently explained by Iwamoto et al. (31
) as the result of Chinese and Japanese immigrants settling in Peru in the 19th century. The 16.4% prevalence rate observed in this study confirms the trend of increasing prevalence of this family, an expansion due to transmission after a relatively recent introduction. Although Beijing strains overall clustered less in our study population, the largest cluster included 46 Beijing isolates.
In this study, we observed a significant proportion (23.5%) of strains that did not match well-defined lineages but were responsible for large clusters (see Fig. S2 in the supplemental material). Among them, 37% were grouped in the poorly defined T group, and 63% had spoligotyping and/or MIRU patterns that did not match an international type. Future analyses including the additional 9 loci for 24-locus MIRU analysis and/or whole-genome sequencing could help to discriminate better between these strains and to elucidate the real clusters.
Clustering is considered a surrogate marker for strains involved in recent transmission. The efficacy of TB control programs can be assessed by the degree of recent TB transmission in the population (44–46
). The validity of molecular epidemiological studies is determined by the resolution of the applied typing method, among other factors. Spoligotyping alone is an unreliable tool for formal phylogenetic analyses, due to the occurrence of independent mutational events in different evolutionary lineages, which may result in identical convergent spoligotyping patterns (47
). However, spoligotyping combined with MIRU-15 analysis allows for high-resolution genotyping to study recent transmission (32
). We excluded 24 samples lacking data for 2 or more MIRU loci. An occasional lack of PCR amplification of some loci has been reported (32
) and might be explained by chromosomal deletions, nucleotide polymorphisms in the sequences complementary to the PCR primers (48
), or insufficient DNA quality. MIRU-15 analysis may lack sufficient power of resolution for the classification of Beijing strains (32
). However, the comparison between MIRU-15 and MIRU-24 analyses showed similar clusters. Therefore, it is likely that the majority of the remaining strains typed by MIRU-15 analysis would mostly remain clustered if MIRU-VNTR typing was extended to 24 loci (32
Duration and geographic/population coverage are other important determinants for the validity of molecular epidemiological studies. Due to logistic constraints and human and/or microbiological factors, it is usually not possible to obtain cultures and DNA fingerprints for all eligible cases. In this study, we focused only on new TB cases, which represented approximately 89% of all pulmonary TB cases, and among these we enrolled only smear-positive cases, which represented approximately 65.3% of the new cases (6
). Borgdorff et al. confirmed that clustering rates may be underestimated as a result of random sampling in time or in space (49
). In settings in which risk factors for clustering may be interpreted as risk factors for recent transmission, these risk factors can be robustly calculated even when sampling is incomplete, i.e., they are also associated with larger cluster sizes. Therefore, odds ratios are generally insensitive to random sampling unless larger clusters are excluded and sampling fractions are small (49
). This study analyzed a sampling fraction of almost 50% of the eligible cases, i.e., registered new smear-positive cases of pulmonary TB, and did not exclude any clusters. As a result, the odds ratios of our study are likely robust estimates.
TB is known to cluster in hyperendemic “hot spots,” which often are characterized by social determinants such as crowding, poverty, and HIV infections (50–52
). Compared with other infectious diseases, however, for which 20% of the population may generate 80% of transmission, TB transmission appears relatively more homogeneous (53
). Previous studies in areas with high TB incidence rates have shown a wide range of transmission rates (clustering rates), ranging from 37 to 72% in settings with TB incidence rates of over 200 cases per 100,000 inhabitants (55–57
). In our study population, we observed a relatively high clustering rate of 69.7%, mainly among Haarlem strains.
In agreement with previous national and regional surveillance data (58–60
), we found high rates of resistance to isoniazid and/or rifampin in this setting. This high MDR rate does not seem to be attributable primarily to recent transmission among new cases, as our MDR strains were not systematically clustered together. Indeed, they were grouped together with non-MDR strains, mostly resistant to at least isoniazid. Thus, the additional acquisition of rifampin resistance could have occurred in epidemiologically linked retreated cases that were excluded from this study.
MDR-TB strains were underrepresented among clustered strains. There are three possible explanations for the lack of association between clustering and the presence of MDR-TB strains observed among new cases in our setting. First, recent transmission could have occurred from MDR-TB in previously treated patients who were excluded from our study. The fact that MDR strains clustered with strains that were resistant to isoniazid or rifampin supports this hypothesis of acquired MDR in retreated cases, which potentially remained infectious for extended periods if not diagnosed properly and treated appropriately. In Peru, the time to change from the standard schedule to the MDR-TB schedule varies from 28 days for new TB patients with risk factors to 2 months for patients without risk factors (61
). Second, transmission could have occurred prior to our sampling period. When we compared our results with those of a previous study conducted in the same district in 2009 (42
), however, we found few shared clusters between the two sets of samples; the 2009 study focused on low-MDR-risk patients, who did not represent the entire pool of MDR-TB cases in the area. Moreover, our study covered almost 2 years of sampling, which is considered the minimum time to study recent transmission (62
). Third, resistant strains could have been imported from other districts of Lima. According to reports from the Ministry of Health (see Fig. S3 in the supplemental material), however, San Juan de Lurigancho has one of the highest TB incidence rates in the capital. Therefore, export (rather than import) from San Juan de Lurigancho to other districts might be more frequent, as a result of high interdistrict mobility. On the other hand, according to the latest census (performed in 2007), 54.7% of the inhabitants living in San Juan de Lurigancho were born in a different region and 15.7% had moved to the district in the past 5 years (63
), which might have caused import of (resistant) strains from those regions. Therefore, infection from previously treated cases, which are at higher risk of being MDR-TB, seems the most plausible explanation for the observed high rates of MDR-TB among new cases.
In the specific case of the Beijing family, worldwide concern has been raised about its frequent association with outbreaks of MDR variants (48
); however, this was contested by Glynn et al. (66
), who found regional differences in the Beijing family's association with drug resistance, with Colombia and Cuba lacking this association. Our results agree with this observation and previous Peruvian studies in which the Beijing family was present in relatively high proportions but was not associated with multidrug resistance (23
). The prevailing Beijing isolates in Lima are associated with the modern subfamily coming from China, consistent with the worldwide trend except in the cases of Japan and Korea, where the ancient subfamily predominates. One particular clone, the PCT001 genotype, has been circulating in Lima since at least 1999 and might have gained a selective advantage allowing for enhanced spreading in the past decade. Indeed, our largest Beijing cluster (n
= 46) displayed the typical PCT001 profile, whereas the second cluster (n
= 15) had the PCT002 pattern (31
Our study has some limitations. First, since this was a passive surveillance study, only patients attending health care facilities in the public sector were included. In Peru, however, most cases of TB are treated within this sector (67
). Second, this study was embedded in a prospective cohort study of new cases of sputum smear-positive pulmonary TB; therefore, only new TB cases that were smear positive were enrolled. Third, the sampling period covers almost 2 years, which is less than the 4-year window that yielded maximum clustering in a study conducted in Malawi (62
In conclusion, our study confirms a diverse population structure of M. tuberculosis in the San Juan de Lurigancho district. The predominant family was the Haarlem family and, although the strains of this family showed the greatest probability to cluster, they were less prone to be resistant to isoniazid and rifampin. The high rate of clustering indicates active recent transmission of M. tuberculosis among new cases but was not associated with drug resistance in this study population. Our results suggest that the high MDR-TB rate among new TB cases in Lima is primarily due not to recent transmission among new cases but rather to treatment failure in previous TB infections, which causes the selection of spontaneous mutations in specific genes associated with drug resistance. Future prospective community-based studies, including new and previously treated patients, should aim to study the transmission of drug-resistant strains in the general population, to confirm this hypothesis. Meanwhile, rapid diagnosis and effective treatment of TB remain crucial for interrupting the chains of transmission, with an emphasis on retreated cases for MDR-TB.