Introduction

The new Post-2015 Global Tuberculosis Strategy identifies action on social determinants of tuberculosis as a main component of the bold policies and supportive systems required for TB control. In line with this Strategy, there is a general agreement that research is needed to assess and measure how social determinants affect TB risk [1].

At the population-level, improved socioeconomic conditions have been accompanied by a decline in the tuberculosis burden in industrialized countries over the past century. However, the association between socioeconomic status (SES) and TB is mediated by different proximal risk factors operating at different points in the natural history of TB [2,3]. The magnitude and direction of the association may therefore vary by setting, study population and the clinical features of the disease [3].

Scant data are available on the association between SES and multidrug-resistant tuberculosis (MDR-TB). Drug-resistant TB can result from two mechanisms: i) the selection for resistant bacteria in a patient undergoing treatment—defined as acquired (or secondary) resistance, or ii) infection with a drug resistant strain, defined as transmitted (or primary) resistance. Although previous studies have identified risk factors for MDR-TB [4,5], few studies have differentiated between risk factors for acquired and transmitted resistance or elucidated the role of SES in these two mechanisms.

Distinguishing between risk factors for acquired and those for transmitted MDR-TB may provide useful evidence to plan, implement, and evaluate targeted, preventive interventions in the context of TB control programs. Here, we analyze the relationship between individual and household-level SES and behavioral factors and acquired and transmitted MDR-TB in an urban South American setting.

Methods

Results

Between October 2009 and August 2012, we identified 3379 cases of culture-confirmed TB of which 411(12%) had at least one multi-drug resistant Mtb isolate at some point during the study. 29 subjects met our definition for acquired resistance and 61 for transmitted resistance. The remaining MDR cases could not be classified as having acquired or transmitted resistance based on these criteria.

Table 1 describes the characteristics of MDR patients by acquired, transmitted and unclassified status. Socioeconomic status was not differentially distributed between MDR-TB subjects included and excluded from our analysis (p = 0.7), nor were the other considered variables(Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Socio-economic risk factors for acquired and transmitted resistance.

https://doi.org/10.1371/journal.pone.0146642.t001

In the univariate analyses, subjects with transmitted resistance were younger, of higher socioeconomic status, more educated, more likely to have a history of incarceration or to be heavy drinkers than those with acquired resistance, although only age met the p = .05 criteria for statistical significance. After adjustment for age, sex and socioeconomic variables in the final model(Table 1), participants of “medium” and “high” SES were 1.31(95%CI:0.37–4.64) and 3.15(95%CI: 0.81–12.18) times as likely to have transmitted compared to acquired resistance (p for trend = 0.08). When we conducted a sensitivity analysis that considered housing alone among the SES variables, we found that participants with transmitted resistance were 3.18(95%CI: 1.10–9.58) more likely to have better housing than those with acquired resistance.

Discussion

Our results suggest that in urban Lima higher socioeconomic status is associated with a nearly 3-fold increased risk of transmitted compared to acquired resistance and, conversely, that lower socioeconomic status is associated with an increased risk of acquired resistance. Quality of housing mediates most of the impact of SES. Despite this suggestive evidence, the results do not allow us to determine whether this effect is due to an increased risk of transmitted resistance in subjects of higher SES or an increased risk of acquired resistance in subjects of lower SES (although the distribution of SES status among the cases suggests the former) and we cannot quantify how much socioeconomic vulnerability contributes to MDR-TB burden through either mechanism.

One possible interpretation of our results is that better houses—where walls, roofs and windows are built with more solid materials—are less well ventilated and therefore increase the risk of within-household transmission. It is also possible that those of higher SES status work less and spend more time at home, thereby increasing their exposure within a household. Another possible pathway is that those with better houses may spend more time socializing indoors with increased risk of household crowding. On the other hand, wealthier people might have better access to healthcare, earlier access to TB diagnosis and supports permitting better treatment compliance, this reducing the risk of acquiring drug resistance. Case detection rates may also vary by SES strata and this might bias the results; although Peru has a strong public National TB Program that captures most cases, we cannot exclude this possibility in our study setting.

Our findings extend to transmitted MDR-TB the results of several studies from Africa which reported an unexpected association between measures of relative wealth and TB. Two consecutive cohorts in Malawi found that subjects living in higher quality houses were at higher risk of TB [10] while other measures of SES such as asset ownership were associated with diminished risk [10]. Similar findings have been reported from both high and low-incidence TB settings.

A large literature is available on how to measure socioeconomic status in relation to health outcomes [12]. Different socioeconomic measures capture different pathways of the association between socioeconomic status and tuberculosis. The rationale for using principal component analysis to derive a composite measure of relative socioeconomic status was to apply a recognized and increasingly used statistical technique and to obtain results easily comparable to other settings’ findings. We have previously applied the same analytical technique to explore the role of proximal risk factors in the relationship between TB and poverty in other settings [13].

The clustering definition we applied was supported by recent findings that demonstrated fully identical 24 loci VNTR typing result to be the most appropriate cluster definition in TB control practice [8].

Our work has some limitations. The conservative case definitions we adopted excluded most of the MDR-TB cases recruited in the cohort and we were therefore not able to assess the relative contribution of the two mechanisms to the MDR-TB burden in the study setting. Consequently, our small sample size did not allow us to identify any other than the most extreme associations and null findings should not be interpreted to suggest that other risk factors had no effect. Despite the limitations for statistical inference with small sample sizes, the p-value of .08 is weakly significant. Our strict definitions to differentiate between acquired and transmitted MDR-TB cases that combined both epidemiological and molecular features adds to the body of knowledge on the topic available in the literature and is one of the strengths of our study. A second limitation is that our findings do not allow us to derive conclusions generalizable to the entire Lima population; still, they allow us to suggest—on the basis of rigorously conducted analysis—that acquired and transmitted MDR-TB are associated with different risk factors that need to be taken into account when planning and implementing control programmes and interventions [14]. A third limitation, addressed above, is that we were unable to distinguish between the scenario in which higher SES facilitate transmission of MDR TB and that in which lower SES enabled acquisition of MDR TB.

To our knowledge this is the first study to assess the association between SES and MDR-TB distinguishing between acquired and transmitted resistance, especially using a prospective design and rigorous definitions that create more certainty about the classification of MDR TB into these categories. In particular, scant evidence is available on whether specific social determinants are proxies for exposure to patients with drug-resistant TB [15], or instead are associated with suboptimal care that selects for drug resistance. In the past, a history of previous TB treatment in drug resistant TB patients was used as a proxy for acquired resistance. However, because the ‘previously treated’ patient category can comprise patients primarily infected or re-infected with resistant strains, this approach may not reliably identify subjects with acquired drug resistance. Molecular epidemiologic approaches have sometime been used to differentiate between primary and secondary resistance; here, investigators have assumed that clustered MDR cases are those with primary resistance and those that are not clustered are due to acquired resistance. Like other molecular epidemiologic studies, however, the validity of this approach depends on the completeness of sampling and errors often results from convenience or small samples [16].

As stated in the recently released Global Tuberculosis Report—selected risk factors, including the emergence of DR and MDR tuberculosis poses a significant threat to global tuberculosis control strategies [17,18,19,20]. Acquired and transmitted resistance arise from two different mechanisms that are associated with different risk factors. We analyzed data from a large-scale prospective study and provide evidence suggesting that better living conditions might be associated with increased risk of MDR-TB transmission in urban settings. Health education programs should target specific subgroups of the population, promote good ventilation and prompt healthcare seeking and diagnosis so as to interrupt TB transmission. Further research is needed to confirm our findings. Our next step will be to geo-reference TB cases to explore the impact of ecological-level factors on MDR-TB transmission.

Supporting Information