Detailed epidemiological analysis as a strategy for evaluating the actual behavior of tuberculosis in an apparently low-incidence region

ABSTRACT

Tuberculosis control in developing regions with apparent low incidence, like the low-income Mexican state of Michoacán, with mean annual incidence rates below 10/100,000 inhabitants, requires knowledge of the actual behavior of the disease. This can be determined using an epidemiological profile at sub-regional level, allowing disclosure of the clinical and social factors that may be hampering efforts to control tuberculosis. In this work, a detailed epidemiological profile was outlined using data of all new monthly cases registered in the National System of Epidemiological Surveillance Database for Michoacán municipalities from 2000 to 2012. Cases were grouped by gender and age, and sociodemographic data were obtained both from the National Institute of Statistics and Geography and from the United Nations Development Programme. Correlations were calculated by Chi-square, Mann-Whitney U, and Kruskal-Wallis H tests. We observed no statistically significant differences between notification rates for the years 2000, 2005 and 2010 (χ² = 0.222, p = 0.895). The percentage of cases is similar between all age groups older than 15, while some regions had low notification rates but high proportions of pediatric cases. Higher proportions of cases of extrapulmonary tuberculosis were observed in municipalities in northern Michoacán. No correlation was found between municipal Human Development Index values and municipal notification rates. Michoacán is undergoing an epidemiological transition with three regions having different epidemiological profiles and particular needs for effective prevention and containment of tuberculosis. Our work shows the importance of the spatial scale of epidemiological profiles for determining specific regional needs of surveillance and containment.

Background

The interaction between pathogenic microorganisms and their hosts is influenced by environmental and social factors, from climatic characteristics to the hygiene practices of the host [1]. The characteristics of the pathogen, the host, and the environment provide the information required for obtaining an epidemiological profile, which consists of data about distribution in space and time of the disease in a given population or community, and the biological, physicochemical, and sociocultural factors involved in its development [1]. Depending on the specific interest, epidemiological profiles can be made at different space and time scales going from the global scale through several decades [2,3] to a single local health facility during a time period spanning a few months or years [4,5]. National level epidemiological profiles are commonly made for several diseases [6,7,8].

Establishing epidemiological profiles allows for planning efficient treatment strategies [9], thus contributing to the containment or eradication of diseases in a specific geographic region [10]. In recent decades, epidemiological profiling has provided a proper framework for guiding the application of molecular biology techniques, which provides the knowledge of isolates’ genotypes of pathogenic microorganisms of the same species. The information from both datasets facilitates the differentiation between isolate outbreaks and sporadic cases, revealing transmission routes, and establishing optimal treatment measures [11,12,13], among other applications.

In the case of tuberculosis, epidemiological profiling has been made at a wide range of time and space scales in order to evaluate the behavior of the disease within defined regions, assess comorbidity, and evaluate the characteristics of health services or other clinical and social factors of public health interest [14,15,16,17,18,19]. Tuberculosis is an infectious disease associated with working age populations – 75% of worldwide patients are aged between 15 and 54 years [20]. The disease is also associated with poverty with 95% of all cases occurring in developing countries [20].

The epidemiological profile of tuberculosis varies as its transmission is reduced [21,22]. Thus, in countries where the incidence of tuberculosis has been lowered, transmission of the disease occurs sporadically, with new cases usually being elderly patients in which the infection has remained latent for many years. On the contrary, in countries with high incidence rates, the transmission of the disease is a frequent event, new cases occurring mainly among young adults. While, infant cases, although being less frequent, tend in general to increase in zones where recent transmission remains uncontrolled [21,22]. Because of these tendencies, the tuberculosis case distribution among age groups can be considered to be an indicator of the dynamics of the disease in a particular region.

The slow and insidious development of extrapulmonary tuberculosis, as well as the reactivation process, causes the proportion of this kind of tuberculosis to increase as recent transmission events become less frequent, in particular in elderly patients. Thus, cases of extrapulmonary infection originate from reactivation of latent cases from transmission events that occurred at least 5 years ago [23].

Finally, association of tuberculosis with other diseases that weaken the immune system – such as infection with the Human Immunodeficiency Virus (HIV), Diabetes Mellitus (DM), malnutrition, and alcoholism – favor the development of active tuberculosis and its evolution toward extrapulmonary infections [21]. The HIV/AIDS pandemic was the most significant factor triggering the values of the incidence rates of tuberculosis throughout the world during the final decade of the twentieth century [21]. In general, regions with the highest proportions of cases of coinfection of HIV and tuberculosis tend to have high incidence rates of tuberculosis [24].

The American continent is one of the regions that has achieved the Stop TB Partnership’s goal of reducing incidence, prevalence, and mortality of this disease to one-half of the same values recorded for the year 1990 [24]. Within the Americas, Mexico registered in 2014 a tuberculosis incidence rate of 21 cases for every 100,000 inhabitants [25], a higher value than the incidence rate reported in 1990 of 17 cases for every 100,000 inhabitants [26]. The more recent report of the Mexican Ministry of Health (Secretaría de Salud) including data prior to the year 2010 states that in over 60% of the tuberculosis cases there is another associated disease, mainly DM (20%) and malnutrition (13.15%). The largest number of cases of tuberculosis in Mexico occurred in the population between ages 25 and 44 years, and 9% of the cases of pulmonary tuberculosis in the country were diagnosed in municipalities with high indigenous populations (i.e. those who belong to pre-Columbian ethnic groups) [26]. This report shows that the incidence rate of tuberculosis varies among the country’s regional states, ranging from 3.9/100,000 inhabitants in Tlaxcala to 54.1/100,000 inhabitants in Baja California, which are states located in central and northwestern Mexico, respectively. The state of Michoacán is located in central western Mexico, and it is the fifth state with a lower incidence rate of tuberculosis with 7.7 cases for each 100,000 inhabitants [26]. The value of incidence of tuberculosis for Michoacán reported by the Mexican National System of Epidemiological Surveillance (Sistema Nacional de Vigilancia Epidemiológica, SINAVE) places the state at the level of countries with low-incidence rates of the disease (<10/100,000 inhabitants) [27], such as Switzerland (7.2/100,000 inhabitants), Norway (8.2/100,000 inhabitants), Austria (8.4/100,000 inhabitants), and New Zealand (7.3/100,000 inhabitants) [25]. These figures would suggest that tuberculosis is a controlled disease in the state of Michoacán, more so if it is compared to other Mexican states such as Guerrero (32/100,000 inhabitants), Tamaulipas (35.9/100,000 inhabitants), or Baja California (54.1/100,000 inhabitants) [26]. However, there are no available detailed epidemiological profiles of tuberculosis for any of the Mexican states, including Michoacán. This hampers the evaluation of the disease’s behavior in the state in order to know if it corresponds to other world regions with comparable incidence rates. Regions of low incidence of tuberculosis have once more become relevant in the fight against the disease, because within these zones it should be possible to achieve significant advances toward its eradication [27]. The design of strategies for the elimination of tuberculosis is based on currently available epidemiological data, and has been focused in previously identified vulnerable groups such as immunocompromised patients and immigrants [28].

The objective of this work was to make a detailed epidemiological profile of tuberculosis in the Mexican state of Michoacán. The epidemiological profiling obtained was intended to be useful for the design of programs for district surveillance, control, and eradication that are based on the epidemiological behavior of tuberculosis in each one of the state’s regions. The approach followed here for building an epidemiological profile can also be useful for containing the spread of tuberculosis in other regions of the world. In particular, the detailed regionalization of the disease’s behavior becomes more relevant in developing countries in which national or regional incidence rates are low. It has also been proposed that global strategies must be complemented by country-level implementation guidelines that begin with local epidemiology and rely on tools developed for specific epidemiological profiles [29].

Methods

Epidemiological data used for the present work were obtained from the database of the SINAVE [30] with previous authorization of the Public Health State Laboratory of Michoacán. The database contains confirmed cases, ruling out all suspected cases. Annual reports of the tuberculosis surveillance program were generated for each year for the time period between 2000 and 2012, according to the following selection criteria: new cases, monthly frequency, and analysis by locality of patients’ residence. Municipal cases of tuberculosis were obtained by month, gender, age group, location of the infection in the host (organs infected), and associated diseases. Additionally, we obtained data of total cases recorded in the state for each age group by location of the infection in the host. Annual reports were obtained for each pair of variables.

Data about total population, population by sex, age group, and indigenous language speakers for each municipality of Michoacán for the years 2000, 2005, and 2010 were obtained from census information for the corresponding years gathered by the National Institute for Statistics and Geography (INEGI) [31]. The Human Development Index (HDI) for the state of Michoacán was obtained from the Human Development Report 2007 for Michoacán [32].

The descriptive analysis consisted in calculation of frequencies, means, medians, and rates of the variables of interest. Although the incidence rate has been calculated for Mexico, there is no estimation of the detection gaps for the regions herein analyzed. Therefore, our analysis relied only on notification rates. The notification rate of tuberculosis (N_TB) is defined as the total cases recorded by the SINAVE for every 100,000 inhabitants. Values of N_TB were calculated for each municipality, sanitary district, age group, and census year by the following equation:

$$N_{TB}=\frac{(\mathrm{R}\mathrm{e}\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{d}\mathrm{e}\mathrm{d}\mathrm{c}\mathrm{a}\mathrm{s}\mathrm{e}\mathrm{s})}{(\mathrm{P}\mathrm{o}\mathrm{p}\mathrm{u}\mathrm{l}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{s}\mathrm{i}\mathrm{z}\mathrm{e})}\times 100,000$$

Notification rate values were grouped according to the categories for incidence rates used by the WHO in its Global Tuberculosis Report for 2013 [33], having 10/100,000 inhabitants as its upper limit for the lower of its ranges. In the present work that value is considered as the maximum expected rate for a low-incidence zone. For the analysis at the municipal level, total cases per municipality during the studied period were considered, each municipality being accounted as a case. Since the data were found to be not normally distributed (data not shown), the Spearman r correlation for nonparametric data was used for calculating correlations between quantitative variables such as municipal notification rates, municipal human development index, percentage of indigenous population, ratio of cases by gender, proportion of pediatric cases, percentage of cases of extrapulmonary tuberculosis, percentage of cases associated to Diabetes Mellitus and percentage of cases associated to HIV/AIDS.

Municipalities of Michoacán have diverse demographic characteristics, ranging from low-populated municipalities below 3,000 inhabitants living in rural settlements, to densely populated municipalities over one million inhabitants frequently distributed in more than one urban settlement. Such demographic heterogeneity yields the observation of atypical cases and hampers the generation of clearly defined patterns. For that reason, we decided to make regional analysis grouping cases within the eight sanitary districts established in Michoacán by the SS for operating public health and epidemiological surveillance programs (Figure 1(a), Additional file S2). Sanitary districts were treated both as a single case (all cases within the district together, without municipal distinction) or as a group of cases (being each municipality within the district one case). Dependent variables compared between sanitary districts were mean notification rate, percentage of cases per age group, percentage of pulmonary and extrapulmonary cases, percentage of cases by gender, and percentage of cases associated with other diseases. The distribution of cases within districts showed a non-normal distribution, while data across time (i.e. annual notification rate) were normally distributed. To know if there were differences between districts as cases, a χ² test was made. When clustered municipal values were considered a Kruskal-Wallis H test (variables with three or more possible values) was made. If we observed a variable to be differentially distributed, paired comparisons between districts were made by means of a Mann-Whitney U test (variables with two possible values) in order to know exactly which district had a significantly distinct distribution. The directionality of such a difference was assigned based on the box graphs representing the distribution in each region. Statistical processing was made using the software PASW Statistics 18.0.0 (IBM).

Figure 1.

Behavior of tuberculosis by sanitary district: (a) Sanitary districts in the state of Michoacán. (b) Mean municipal notification rates. (c) Percentage of cases by age group. (d) Percentage of pediatric cases in each district. (e) Percentage of extrapulmonary cases in each district. (f) Percentage of cases associated to HIV or DM in each sanitary district.

Maps were generated in the software QGIS 1.8.0-Lisboa [34], using as inputs the vectorial layers of the topographic maps for the state of Michoacán and its municipalities at a 1:250,000 scale obtained from the web page of INEGI [35].

Results

During the time period 2000–2012 in the state of Michoacán 4,307 new cases of tuberculosis were recorded with a mean of 339 new cases per year. The notification rate for the state was calculated for the census years 2000 (9.61/100,000 inhabitants), 2005 (8.70/100,000 inhabitants), and 2010 (8.00/100,000 inhabitants), finding no significant differences between these values (χ² = 0.222; p = 0.895; degrees of freedom, df = 2).

Because the available projections for the population for the region gave highly imprecise results regarding the data obtained during the last census (Additional file S1), we decided not to calculate the notification for inter-census periods. When data were analyzed at the municipal level, the total number of cases was reduced to 4,202. This discrepancy between the total number of cases generated by adding the total municipal cases (4,202) and the total cases reported for the state (4,307) may have been due to incomplete information in case records. Annual differences were larger during the first five years of the evaluated period, while the records of the last seven years did not present that issue. When municipalities were analyzed as groups within the eight sanitary districts, the mean notification rate was significantly higher in districts 7 and 8 (Figure 1(b); χ² = 44.244; p < 0.001; df = 7; residual district 7 = 7.8; residual district 8 = 16.8). The percentage of cases by age group (Figure 1(c)) shows significant differences for the 15–29 age group (χ² = 14.462; p = 0.044; df = 7; residual district 2 = 2.5; residual district 4 = 0.5; residual district 5 = 2.5; residual district 7 = 6.5; residual district 8 = 7.5). For the group aged 60 and over, districts 3, 5, and 6 showed significant difference (χ² = 22.692; p = 0.002; df = 7; residual district 3 = 20.6; residual district 6 = 0.6; residual district 6 = 5.6). For the pediatric age group (0 to 14 years old), there was no significant difference between districts (χ² = 7.653, p = 0.346, df = 7), whose values are shown in Figure 1(d). It is noteworthy that the percentage of pediatric tuberculosis is inversely related to the mean notification of the disease for the same district.

The percentages of extrapulmonary cases (Figure 1(e)) are significantly different between sanitary districts (χ² = 31.993; p < 0.001; df = 7) being higher in districts 1 (residual = 11.4), 2 (residual = 8.4), 4 (residual = 3.4), 5 (residual = 2.4), and 6 (residual = 5.4). Thus, the districts with lower mean notification values have a higher percentage of extrapulmonary cases.

Regarding the association between tuberculosis and DM (Figure 1(f)), there were no significant differences between sanitary districts (χ² = 8.860; p = 0.263; df = 7). However, significant differences were observed when comparing the percentage of tuberculosis cases associated to HIV/AIDS (χ² = 20.800; p = 0.004; df = 7). In the latter comparison, district 2 (residual = 4.0), 5 (residual = 1.0), and 8 (residual = 7.0) had a significantly higher percentage of tuberculosis cases associated to HIV/AIDS.

The wide values distribution observed in different sanitary districts for some variables is evidence of high heterogeneity between municipalities within districts (Figure 2(a–f)).

Figure 2.

Distribution of cases of tuberculosis grouped by epidemiological variables in each sanitary district: (a) Mean municipal notification rate. (b) Distribution of tuberculosis cases by age group. (c) Distribution of percentage of tuberculosis cases among the pediatric population. (d) Distribution of the percentage of cases of extrapulmonary tuberculosis. (e) Distribution of the percentage of cases of tuberculosis associated to HIV/AIDS. (f) Distribution of the percentage of cases of tuberculosis associated to Diabetes Mellitus.

The distribution of mean municipal notification rates within each district was assessed. In most cases distributions were non-normal (Figure 2(a)). Pairwise comparison of the mean notification rates (Figure 2(a); Additional file S3) showed that districts 2, 3, 6, and 7 had, on average, higher notification rates; although district 8 showed a very large interquartile range and included municipalities with mean notification rates above those in the municipalities forming the rest of the districts.

The only significant difference found in the analysis of tuberculosis cases by age group was between districts in the populations between 15 and 29 years old, and 60 and above (Figure 2(b), Table 1). Pairwise comparisons determined that the municipalities in district 3, 4, 5, 6, and 8 registered lower percentages of cases in the population between 15 and 29 years old. On the other hand, districts 1, 2, and 7 presented higher medians for this variable (Additional file S3).

Table 1.

Comparison of the distribution of cases of tuberculosis among sanitary districts.

	Sanitary Districts†¶ χ2 (p)
XN	21.72 (0.003)
PED	8.444 (0.296)
15–29	14.933 (0.037)
30–44	4.449 (0.727)
45–59	7.299 (0.405)
60+	23.805 (0.001)
ETB	31.614 (<0.001)
DM	9.36 (0.228)
HIV	27.48 (<0.001)

Notes: ^† Kruskal-Wallis test. Bold characters indicate significant differences between at least two districts. ^¶Eight sanitary districts were included in the analysis; degrees of freedom = 7 in all the tests. Abbreviations and symbols: X_N: Mean municipal notification rate; PED: Proportion of pediatric cases, i.e. under 15 years old; 15–29: Proportion of cases in the population from 15 to 29 years old; 30–44: Proportion of cases in the population from 30 to 44 years old; Proportion of cases in the population from 45 to 59 years old; 60+: Proportion of cases among the population from 60 years old and above; ETB: Percentage of cases of extrapulmonary tuberculosis; DM: Percentage of cases associated to Diabetes Mellitus; HIV: Percentage of cases associated to HIV/AIDS.

Observation of the pediatric cases in municipalities within each district revealed several districts in which the interquartile range included percentages of pediatric tuberculosis of above 10% (Figure 2(c)), and municipalities with atypical data recording pediatric tuberculosis values of nearly 50% of the total cases. These atypical cases have to be considered with caution since they may be due to the low number of cases. The complete data for municipalities having percentages of pediatric tuberculosis above 5% are shown in Additional file S4. Of the cases in the over 59 years old population, the 1, 3, 5, 6, and 8 districts had the highest median value for this variable (Figure 2(c), Additional file S3).

Significant differences were also found in the proportions of extrapulmonary cases among districts (Table 1). The distribution of municipal values and pairwise comparisons (Figure 2(d), Additional file S3) allowed for determining that districts 3, 7, and 8 had low proportions of extrapulmonary cases, while in the remaining districts such proportions had medians of above 20%.

Both DM and HIV/AIDS are factors favoring the development of active tuberculosis. The proportion of tuberculosis cases associated to HIV/AIDS was significantly larger in districts 2 and 8 (Figure 2(e), Additional file S3), but we found no significant differences among districts in tuberculosis cases associated to DM (Table 1, Figure 2(f)). It was however noticeable that the interquartile ranges of five of the eight districts had percentages of comorbidity with DM of above 20% (Figure 2(f)).

Because some municipalities had very high percentages of extrapulmonary tuberculosis, we assessed the possibility of finding some correlation between percentages of pulmonary tuberculosis and the proportion of pediatric cases, concomitance with DM, and concomitance with HIV/AIDS (all factors of risk for the dissemination of the disease), as well as with the mean of municipal notification rate (Table 2). However, the only correlation found was with the concomitance with HIV/AIDS.

Table 2.

Correlation between sociodemographic and epidemiological variables.

		HDI 2000	HDI 2005	IND 2000	IND 2005	IND 2010	M/F	PED	ETB	HIV	DM
XN	ρ	–	–	–	–	–	0.094	−0.040	−0.121	0.272	0.148
	p	–	–	–	–	–	0.322	0.674	0.202	0.004	0.116
N2000	ρ	0.231	–	0.109	–	–	–	–	–	–	–
	p	0.014	–	0.250	–	–	–	–	–	–	–
N2005	ρ	–	0.160	–	0.182	–	–	–	–	–	–
	p	–	0.091	–	0.054	–	–	–	–	–	–
N2010	ρ	–	–	–	–	0.116	–	–	–	–	–
	p	–	–	–	–	0.220	–	–	–	–	–
M/F	ρ	0.161	0.156	−0.179	−0.152	−0.74	–	0.060	−0.114	0.063	0.277
	p	0.089	0.100	0.057	0.109	0.406	–	0.526	0.231	0.504	0.003
PED	ρ	−0.028	−0.012	0.027	0.074	0.095	0.060	–	0.155	0.087	−0.150
	p	0.770	0.902	0.776	0.436	0.314	0.526	–	0.102	0.357	0.112
ETB	ρ	0.302	0.290	−0.062	0.054	−0.005	−0.114	0.155	–	0.201	0.158
	p	0.001	0.002	0.513	0.571	0.960	0.231	0.102	–	0.033	0.095
HIV	ρ	0.333	0.408	0.115	0.099	0.144	0.063	0.087	0.201	–	–
	p	<0.001	<0.001	0.225	0.296	0.128	0.504	0.357	0.033	–	–
DM	ρ	0.041	0.025	−0.096	−0.045	−0.031	0.201	−0.150	0.158	–	–
	p	0.668	0.792	0.314	0.633	0.743	0.033	0.112	0.095	–	–

Notes and abbreviations: ^† X_N: Mean municipal notification rate; N_{2000, 2005, 2010}: municipal notification rate for the corresponding year. HDI_{2000, 2005, 2010}: Human development index for the corresponding year; IND_{2000, 2005, 2010}: Percentage of indigenous population for the corresponding year; M/F: Ratio of cases by gender; PED: Proportion of pediatric cases; ETB: Percentage of cases of extrapulmonary tuberculosis; DM: Percentage of cases associated to Diabetes Mellitus; HIV: Percentage of cases associated to HIV/AIDS. ^§Bold characters indicate significant correlations. All data for this analysis can be found in additional file S2.

Visual inspection of the values analyzed at the municipal level shows the noteworthy fact that extrapulmonary tuberculosis is more frequent in the center and north of the state (Figure 3(a)), while regions in the southeast of the state showed high percentages of pediatric tuberculosis, despite that in these latter municipalities low notification rates are reported (Figure 3(b)).

Figure 3.

Relation of pediatric tuberculosis with both the mean municipal notification rate and extrapulmonary tuberculosis: (a) Extrapulmonary tuberculosis contrasted with pediatric cases. (b) Mean municipal notification rate contrasted with cases of pediatric tuberculosis.

Although analyzed, no correlations were observed in the proportion of cases by gender, human development index, or cases between habitants in indigenous ethnic groups (Table 2).

Discussion

The epidemiological profile of tuberculosis generated in the present work shows that in the state of Michoacán the disease has a heterogeneous behavior among the sanitary districts. This can be observed both when the sanitary districts are analyzed as single cases, or as groups of municipal cases. The analysis of districts as groups of municipalities allows pinpointing the municipalities that drive the behavior displayed by an entire district. While some of these districts have characteristics similar to those in low-incidence countries, others have the typical behavior of high incidence populations [36]. For example, districts 7 and 8 – with headquarters in the municipalities of Apatzingán and Lázaro Cárdenas respectively – have a behavior typical of high incidence zones. Both Lázaro Cárdenas and Apatzingán are urban settlements, which may account for higher notification rates compared to other municipalities. Even though living in densely populated urban settlements is a risk factor as has been reported in several countries [37,38], no correlation of tuberculosis municipal notification rate with human development index was found. We expected a positive correlation, since urban communities have a higher value for this index due to overcrowding, high HIV prevalence, and occupational transmission [32,39]. Given the need for reducing the contagion of tuberculosis in these districts, the possible influence of commercial activity and the associated flow of population must be evaluated in detail in further studies.

In contrast, in other regions of the state the generated epidemiological profile indicates contradictions and the mixture of characteristics proper of low and high incidence zones. This is particularly evident in district 4 in the central portion of the state, which presents municipalities with above 5% of pediatric tuberculosis as well as a more homogeneous distribution of cases within the population younger than 60, both being characteristics of high incidence regions. Nevertheless, the incidence value of district 4 is among the lowest, and the high proportion of extrapulmonary cases matches that data. Other districts also have a high percentage of cases of pediatric tuberculosis, which is an epidemiological marker that must not be overlooked. Children develop symptoms after 6 to 12 months from the time of primary exposure [40], because of which, pediatric cases are markers of recent transmission events in high incidence zones. In low-incidence countries, the cases of tuberculosis among children below 15 years of age are up to 5% of the total number of cases, while in high incidence countries that same percentage rises up to 20% [22]. The latter becomes relevant, since adolescents tend to present a substantial increase in tuberculosis incidence [41]. As Figures 2(b) and 3(b) show, there are municipalities for every district in which no cases are reported, some of them adjacent to municipalities with high notification rate values. Because of the lack of additional information, we are not able to identify the cause of these values, although the existence of free tuberculosis zones cannot be discarded. Nevertheless, it is also probable that some of these municipalities are actually underreporting ongoing tuberculosis cases. An intended search is recommended in these municipalities. Taking all these data together we can state that the obtained profile for eastern Michoacán suggests there is underreporting of cases of tuberculosis among the adult population.

According to the WHO, unreported cases include: unreported cases diagnosed and treated by the public health systems; unrecorded cases diagnosed and treated by the private health sector; undiagnosed cases of patients recurring to either public or private health services; cases of patients having access to some kind of health services, but that do not recur to them; and cases of patients without access to health services [42]. The magnitude of the underreporting of cases can be evaluated by comparing the number of incident cases recorded with the number of mortality records attributable to tuberculosis [43], or by means of the active search of cases through health surveillance efforts [42]. These strategies have allowed for the identification of the main causes of case underreporting to be the lack of microbiological confirmation of negative sputum smear samples and the poor diagnosis of patients with mild symptoms [44]. In addition, a recent study formulated a mathematical age-structured model of tuberculosis transmission. This model uses some parameters described in this work, such as tuberculosis incidence in children or adult populations to apply optimal control strategies [45]. It has also been observed that cases of elderly patients and of patients with higher education levels have a higher risk of being overlooked by health surveillance systems [46]. The above-mentioned observations and factors regarding underreporting of cases of tuberculosis should be taken into account for future optimization of patient recording in eastern Michoacán. We recommend that the districts containing municipalities with a high percentage of pediatric tuberculosis to undertake an intended search of patients, increase microbiological confirmation for negative sputum samples, increase the percentage of positive sputum samples that are cultured to isolate and identify the pathogen or adopt approved detection techniques based in molecular biology tools. All of these actions would increase the detection rate of both human transmitted and zoonotic tuberculosis.

The highest number of extrapulmonary cases was recorded in districts 1, 2, 4, 5, and 6, covering the central and northern portions of the state, a region displaying heterogeneity in notification rates and in case distribution by age group, and in which a substantial number of municipalities have low notification rate values. The latter characteristic can explain the high percentage of extrapulmonary cases, because as the number of incident cases decreases, extrapulmonary tuberculosis becomes more prevalent [46,47]. This form of the disease takes place when the bacillus invades any organ of the host through hematogenous dispersion, and due to oxygenation in most tissues being below the levels optimal for development of the pathogen, extrapulmonary infection becomes concealed [37]. The extrapulmonary infection occurs in only between 10% and 20% of the immunocompetent patients developing an active infection, while among immunocompromised patients that same percentage is increased [48], rising to 60% [49]. However, despite that at the state level the proportion of extrapulmonary cases has increased since the year 2000, the notification rate did not significantly decrease, suggesting other factors must be contributing to the increase in cases of extrapulmonary tuberculosis.

Districts 2, 4, 5, and 6 include cattle-raising regions with above 0.5% prevalence of bovine tuberculosis caused by M. bovis [50], while in municipalities of district 1 backyard milk production takes place, a productive activity known for its low level of technification [51]. Presence of zoonotic tuberculosis can in part explain the high percentage of extrapulmonary tuberculosis cases, given that M. bovis is highly prone to dissemination to other organs [52,53]. In some Latin American countries, M. bovis has been recorded to be responsible for between 1% and 2.5% of total cases of human tuberculosis; however, the absence of permanent programs for intentional search for zoonotic cases lead us to suppose that these data could be underestimating the problem [54], since in Mexico, the median percentage of human tuberculosis caused by M. bovis is 7.6% [55]. In the neighboring states to Michoacán, such as in Querétaro and Jalisco, the isolation of M. bovis has been reported in 13.8% and 28% of the studied patients, respectively [56,57]. Both of these states are adjoining with regions of the state of Michoacán in which cases of extrapulmonary tuberculosis are more prevalent. Otherwise, the presence of M. bovis in patients with pulmonary tuberculosis has already been recorded in Michoacán [58], where one of 46 analyzed isolates was identified as such (2.1%). Unfortunately, that study failed to record the geographic origin of the samples, for which it is impossible to know how well represented were the municipalities in the central and northern portions of the state. Because of this, and in order to establish the relevance of M. bovis in the extrapulmonary cases recorded in central and northern Michoacán, we recommend that culture, identification, and molecular genetic typing of the causal agent become regular protocols.

Another aspect to be considered is the percentage of cases of pediatric tuberculosis in the state. Dissemination of the bacillus outside the lungs is more common in pediatric than in adult patients. However, only 16.17% of the cases of extrapulmonary tuberculosis in Michoacán occur in the pediatric population, which suggests that pediatric cases are not responsible for the rise in the number of this form of tuberculosis.

Finally, we should state that we were not able to test the correlation between BCG vaccination and extrapulmonary tuberculosis. BCG vaccine administration protects children against miliary and meningeal tuberculosis, but their protective effects in adults are variable [59]. However, detecting a low immunization percentage between cases may suggest one of the causes for the rise of extrapulmonary tuberculosis. Unfortunately, for the first 5 years of the studied period, there were incomplete records in this matter. The sum of vaccinated and unvaccinated cases for a year was less than the total cases reported for the same year. A clue about the immunization status of the population in Michoacán can be found in the immunization coverage estimates that Mexico has reported for the BCG vaccine, which ranges from 91% to 99% in the period of study [60].

Epidemiological transition in terms of age groups

As amply reported for other countries, the distribution of tuberculosis by age groups is related to the incidence of the disease. In countries with low incidence the patients tend to be older than those in high incidence countries [61,62,63,64,65,66,67]. The age of patients may be used as a marker of tuberculosis dynamics that underlie notification or incidence rates, since the differentiation in age means or predominant age groups between high- and low-incidence countries are conserved even in the case of migration. An example of this is the fact that the epidemiological profile of tuberculosis in low-incidence countries has a highly important migratory component, given that immigrants tend to come from developing countries having high rates of incidence of the disease [47,61–67]. Recent studies in developing countries tend to discern between the results from native patients from those of immigrant patients coming from zones of high incidence of tuberculosis. While native patients have a mean age above 50 years, migrant patients tend to have a mean age of about 30 years [61–67], therefore the immigrant population with tuberculosis reflects the behavior of the disease in countries with high incidence rates in which most cases occur among young adults [47,61–67]. This behavior can also be observed at a subnational level: In the Mexican states with the highest incidence rates, such as Nuevo León and Guerrero, the age of the majority of the patients oscillates between 20 and 40 years [68,69]. In the states of Chiapas (25.2/100,000 inhabitants) and San Luis Potosí (15.8/100,000 inhabitants) most cases occurred in patients between 36 and 55 years of age [70,71]. It must however be emphasized that these studies cover a sample of the population and not the total of the state’s records.

Most cases of tuberculosis in the state of Michoacán concentrate in the population older than 59 years. However, districts 2, 4, and 6 show a more similar distribution in the range between 15 and 59 years of age, with districts 2 and 4 having the lower notification rates. However, it can be observed that the distribution of cases in the state does not completely correspond to regions of low or high incidence, but indicates an epidemiological transition in which the elderly population begins to become predominant, and in which reduction of cases among younger population – usually due to recent events of transmission – has not yet been achieved [72]. These epidemiological transitions have been observed in other countries. During the time period between 1986 and 1996 in Brazil, an increase in the rate of incidence of tuberculosis was observed among the population above 30 years of age, while that rate remained above 10/100,000 inhabitants in the age groups from 0 to 20 years [73]. Along a ten-year time period in India and the Republic of Korea, the tuberculosis incidence rates decreased in age groups of 15, 20, and 40 years, but increased in the population above 50 and 60 years of age [74]. In the case of Brazil, the observed change was attributed to the decreased risk of contagion after the implementation of programs for the control of tuberculosis [73]. In the cases of the Republic of Korea and India, the transitions appear to be caused by malnutrition and DM in adults, more so in India, a country in which the incidence of DM has been rising [74]. It is important to establish in the future if these or other causes are involved in the epidemiological transition observed in Michoacán.

Diseases associated to cases of tuberculosis

Coinfection of M. tuberculosis with HIV/AIDS has been the focus of the attention worldwide because the HIV/AIDS pandemic was the cause of the increase of cases of tuberculosis during the late twentieth century [75,76]. Currently, countries with higher prevalence of HIV are also those with higher incidence of tuberculosis [24,76,77].

According to official information for Mexico, the most frequent comorbidities in cases of tuberculosis are DM (20.1% of the cases), followed by malnutrition (13.1%), HIV/AIDS (5.8%), and alcoholism (5.7%) [26]. Our data failed to find a correlation in the state of Michoacán between the cases of tuberculosis associated with DM and the mean municipal notification rate of tuberculosis. Such a lack of correlation can be due to the inexactitude of the records of comorbidity present in patients with tuberculosis [78], given that the association between these two diseases had been previously observed both in Mexico [79,80] and elsewhere in the world [81], and such association had been recognized to occur even before treatment with antibiotics was available [82]. The review made by Jeon and Murray (2008) shows that this association is made further clearer when medical records are supplemented with measurements of the level of glucose in patients. Because of this fact, in order to have more precise data about comorbidity and for improving the control of tuberculosis, some authors have suggested to perform an intended search of tuberculosis among patients of DM, and vice versa [81,83].

Conversely, the cases of tuberculosis associated to HIV/AIDS correlated with the municipal notification rate of tuberculosis, which suggest that this comorbidity could play an important role in the dissemination of tuberculosis in the state of Michoacán, as it occurs in other regions [24], even though it is present in few cases. Districts 2 and 8 presented the highest number of cases of tuberculosis associated with HIV/AIDS. This comorbidity could also contribute to the higher number of extrapulmonary cases observed in district 2, but the correlation between both variables was not evident. District 8 presents one of the highest notification rates of tuberculosis, but our data could not establish the extent in which HIV/AIDS contributes to the generation of new cases of tuberculosis.

The analysis of the official data from the epidemiological surveillance system together with general demographic statistics allowed us to observe regional differences in the behavior of tuberculosis. Delimitation of regions with particular epidemiological characteristics may significantly contribute to the implementation of specific actions for preventing and controlling the disease. We suggest regionalization in three epidemiological zones (Figure 4), each one having particular epidemiological profiles of tuberculosis and recommendations. We assume that notification rates are lower than incidence rates, so high notification rate values are indicative of high incidence rates. Therefore, we can describe the three zones as follows. Epidemiological region 1: A zone of high incidence including sanitary districts 7 and 8, with high frequency of transmission events. We suggest evaluating a larger number of contacts of recorded patients, if possible, by means of annual checkup appointments, as well as to take into consideration making intentional searches of cases of latent infection. Epidemiological region 2: A zone with possible underreporting of cases includes the municipalities of sanitary districts 3 and 4. We suggest the screening for cases of active tuberculosis between patients that attend the districts’ health centers for any other reason. Epidemiological region 3: A zone with possible prevalence of zoonotic tuberculosis including municipalities in sanitary districts 1, 2, 4, 5, and 6. We suggest a program of culture and identification supported by molecular genetic techniques of all extrapulmonary samples, and coordination with bovine tuberculosis surveillance programs for intentional search of cases in communities with high prevalence of zoonotic tuberculosis.

Figure 4.

Regionalization of municipalities based on specific needs for documentation, containment, and prevention of tuberculosis. The limits of the regions were establishing to include complete sanitary districts. Hatched municipalities presented both high percentages of pediatric tuberculosis and of extrapulmonary cases.

Our analysis shows the importance of identifying sociodemographic factors preventing the significant reduction of new cases in high incidence zones. Also relevant was establishing the importance of zoonotic tuberculosis in the northern regions of the state; in the case of zoonosis being infrequent, it must be understood if some characteristics associated with the host or the pathogen are favoring the development of extrapulmonary infection.

Explicit comments must be made about the limitations found in our work in order for these to be considered in future demographic profiles made from the same inputs we used. The main limitation encountered was the use of previously aggregated data of confirmed cases, which limits the kind of statistical analyses that can be made to a chi-squared test. However, percentage values are commonly used elsewhere [22,23,39,40] and so, these values allowed us to make pertinent comparisons. We strongly suggest adopting an Open Data policy, with sensitive data protection, in order to allow more detailed analysis of the behavior of this disease. Another source of possible bias in data is missing records of some variables, such as microbiological identification of the pathogen, vaccination records, and comorbidities of the patient. An additional weakness of our study was the lack of microbiological and molecular data limiting inferences made about the origin and characteristics of mycobacterial populations occurring within each region. This latter constraint also limits the evidence supporting our epidemiological regionalization. However, our work reveals the need to document by region, intended search of cases, a more frequent culture and identification of causal agents, incorporation of molecular epidemiology, and changes in data access policies.

Our results are presented as a background for future work combining clinical epidemiology, microbiology, and molecular genetic typing of the causal agents in order to overcome the above-mentioned limitations and providing a better reference framework for future activities designed for the epidemiological surveillance of tuberculosis.

Conclusions

The epidemiological profile of tuberculosis in Michoacán is complex and cannot be accurately explained in terms of the official state statistics based on simple averages. The state of Michoacán is located in a zone of epidemiological transition that is not homogeneous throughout its territory. The detailed epidemiological profile we made shows that tuberculosis behaves heterogeneously and not as expected in low-incidence zones, and we have established three epidemiological regions with a particular need for attention and surveillance. Such needs are not reflected in values of state incidence. The existence of high incidence zones suggests that transmission of tuberculosis is not very effective in such areas although a deeper epidemiological characterization is needed in order to determine the sociodemographic factors involved in the behavior of the disease at the local level. Such characterization should include survey-based studies that allow assessment of lifestyle, healthcare-seeking behavior, and adherence to treatment of each patient. This information is vital for the design of strategies for the surveillance, prevention, and control of tuberculosis. Detailed epidemiological profiling can be applied in any other region of the world, in particular in areas with apparent low incidence and with contrasting sociodemographic circumstances.

Funding Statement

This work was supported by the Consejo Nacional de Ciencia y Tecnología [34145]; Fondo Mixto Consejo Nacional de Ciencia y Tecnología y Gobierno de Michoacán [MICH-2012-C05-197631]; Fondo Regional para el Desarrollo Científico, Tecnológico de Innovación [193512/2012-02]; Coordinación de la Investigación Científica UMSNH [16.4]; Fondos Mixtos Michoacán 2009 (http://www.conacyt.mx) [127128].

Disclosure statement

The authors declare that they have no competing interests.

Supplementary-Material

Supplemental data for this article can be accessed here.