Tuberculosis (TB) is still a public health problem the world over, with a great part of its morbidity and mortality concentrated in developing countries.1 Since 1999, Peru is among the 23 countries with the greatest TB load, according to the WHO classification. In 2002, Vietnam and Peru were removed from this list because, according to the reports of these countries, they were able to meet the objectives of the millennium for the detection and treatment of cases: 70% of cases detected and 85% of cases cured.1,2 However, despite these achievements, Peru still maintains the highest rate of incidence of TB in Latin America together with Bolivia and Ecuador.1 Likewise, it is the country with the greatest number of multidrug-resistant TB (MDR-TB) cases and extremely drug-resistant TB (XDR-TB) cases in the continent.1,3 One of the main causes of this problem has been the high rates of treatment non-compliance before the 1990s (50%). Moreover, even though the prevalence of non-compliance in the last five years has been less than 5%,4 this average is not representative of the high rates of several cities and provinces or departamentos.

Non-adherence to anti-tuberculosis treatment is defined as not coming to the clinic for treatment for more than thirty consecutive days.4 It is an important barrier for controlling TB because it is related with treatment failure, the development of drug resistances, dissemination of the infection and increased treatment costs.5 Among the possible reasons for non-compliance are: the prolonged duration of the treatment (6–8 months), the high number of pills (minimum of 11), adverse reactions to anti-tuberculosis medicine and lack of information about the disease and its treatment.

The identification of factors that influence the probability of discontinuing the treatment is the first step in designing interventions that improve treatment adherence and the control of the disease. Therefore, the objective of this study was to identify the risk factors associated with non-compliance of anti-tuberculosis treatment in the four Peruvian provinces with the highest incidence.

We carried out a study of unmatched cases and controls (1:2). The study sample was randomly selected from among patients who initiated treatment in 2004–2005 and who completed treatment up until September 2006 in four Peruvian provinces (Loreto, Cusco, Ica and Lima (Callao)). Together, these provinces accumulated 80% of the total number of patients who had abandoned treatment. In order to select the provinces, the Pareto method was used because, due to logistical and economic difficulties, all the provinces of the country could not be covered (Table 1 and Fig. 1). The data collection was done from January 2007 to January 2008 (basically due to the difficult geographic accessibility in the provinces).

Figure 1.

Peruvian provinces or departamentos included in the study.

(0.2MB).

The sample size was determined based on a pilot study that had evaluated different variables, among these being previously discontinued treatments. This variable was what allowed us to obtain a larger sample size. The prevalence in this pilot study for the controls was 0.02 (patients who did not abandon) and when we considered a power of 80%, an odds ratio of 3.14 and a 95% confidence interval, we obtained a sample size of 265 cases and 530 controls.

The participants were patients treated with Directly Observed Treatment (DOT) with treatment schemes I and II (I: indicated for never-treated patients; and II: indicated for relapses and those who had previously discontinued treatment, then returned). Scheme I included: Isoniazid (H), Rifampicin (R), Pyrazinamide (Z) and Ethambutol (E). The duration was 6 months, until 82 doses were completed, divided into two stages: Phase 1: 50 doses (daily from Monday to Saturday with HRZE); Phase 2: 32 doses (2 times a week, with RH). Scheme II included: Rifampicin (R), Isoniazid (H), Pyrazinamide (Z), Ethambutol (E) and Streptomycin (S). The duration was 8 months (approximately 32 weeks), until 125 doses were completed, divided into two stages: Phase 1: 75 doses (2 months, daily, from Monday to Saturday with RHZES, followed by one month daily from Monday to Saturday with RHZE); Phase 2: 40 doses (intermittent, 2 times per week, with RHE). We excluded unconfirmed transfer patients because their actual condition was unknown (potential non-compliers or possible controls) and those patients with irregular treatment.

We defined cases (non-compliers) as those patients who stopped taking the treatment for a period of 30 consecutive days and controls as those patients who completed the treatment satisfactorily and were discharged from the National Health-Care Strategy for the Prevention and Control of Tuberculosis (in Spanish, ESNPyCTBC) as cured.

The cases were selected randomly, without any substitutions, and there were no patient deaths among the selected cases. The controls were selected from the same health-care center where the cases had been identified, using the patient registry of the ESNPyCTBC, obtaining data about the initial patient condition, clinical evolution, punctuality when taking treatment and Adverse Drug Reactions (ADR). Prior to the enrollment of the patients, informed consent was given by all the participants. Later, they answered a field survey that was designed according to the Lalonde Laframboise model2,4 that classifies risk factors into 4 groups: health-care service factors, human biology factors, lifestyle factors and environmental factors (including physical, social and psychological setting). The survey was applied at the participants’ homes and/or at the health-care centers, and the majority of the questions were closed. The interviewers were physicians and nurses who worked at the health-care establishments and had been trained.

Poverty was divided into three categories: extremely poor, poor and non-poor (according to the Peruvian National Institute for Statistics – INEI).6 For the evaluation of any possible alcoholism, the CAGE questionnaire proposed by the WHO was used, which classifies alcohol use into: no alcohol consumption, low risk for alcoholism (social drinker) and high risk of alcoholism (prejudicial consumption), the said test was modified in order to be used in retrospect. In order to evaluate the satisfaction of the subjects with the information received, we used the ISQ-1 test,7,8 which classifies this perception as either excellent, good, acceptable, unsatisfied and very unsatisfied. Last of all, to measure family support, the Olson and Wilson family satisfaction scale was used to classify the levels of satisfaction as high, average or low.9

For the statistical analysis, the data collected were subjected to a quality control, followed by a bivariate analysis comparing the distribution of variables between cases and controls. The resulting significant variables were input in a multivariate logistic regression analysis. The analyses were done with the SPSS 16.0 software package.

This study was completed and financed by the Epidemiology Agency of the Peruvian Health Ministry and was approved by the Ethics Committee of the Hospital Dos de Mayo in Lima.

We analyzed 870 patients, 265 of which were cases and 605 controls. Out of the total number of patients, 60% were men, the predominant age group was between 21 and 30 (40%), 50.9% were single and 61% had completed secondary school.

As for their poverty level, 55% were not considered poor, 27% poor and 17% extremely poor. 70% of the patients were recruited from the outpatient consultations, and the remainder was identified by means of an active search (identification of patients with respiratory symptoms in the community). 15.2% had a previous history of non-compliance (35.8% [95] of the cases and 6.8% [41] of the controls) and 57.7% felt discomfort during the treatment (66.9% of cases and 54.2% of the controls). Likewise, 24.9% consumed alcohol, 4.3% addictive drugs (marihuana, cocaine, crack) and 3.6% recreational drugs (terokal, ecstasy).

At the bivariate level, we observed that the factors associated with risk of discontinuing treatment were: being male, feeling malaise during treatment, expressing some sort of fear about the treatment and not having a good family environment, considering the quantity of medication excessive, having a bad relationship with health-care workers, doubting the capability of the health-care personnel, presenting a history of previous non-compliance, not being satisfied with the information received, as well the consumption of alcohol, recreational drugs and addictive drugs.

The protection factors that lowered non-adherence were: higher education, having a spouse/partner, trusting the treatment, having an adequate family relationship, not considering the quantity of medication excessive, having a good relationship with the health-care personnel, having received a home visit after three days of absence and considering the information received between good and excellent (Table 2).

The multivariate analysis confirmed the influence of the variables of being male, presenting some type of malaise during treatment, having a history of previously discontinuing treatment and use of recreational drugs. The protective factors were: presenting an excellent level of satisfaction with the information received and considering the patient care hours at the health-care centers adequate.

On the other hand, it has been observed that the risk for non-compliance among patients with a history of previous treatment discontinuation is 7.95 (4.76–13.27). This risk increases to 11.24 (4–31.62) if, in addition to presenting a history of previous non-adherence, the patient is also poor.

The control of TB largely depends on its being treated successfully. In this study of four Peruvian provinces, we have identified that non-compliance with anti-tuberculosis treatment is still relatively frequent and associated with non-modifiable factors (such as masculine sex, from a biological standpoint, and a history of previous non-compliance) and other factors whose control would improve adherence (malaise during treatment, consumption of recreational drugs, poverty, lack of satisfaction with the information received and the patient-care schedule). We should mention that sex could also be considered a modifiable factor as it may be influenced by either cultural or other factors.

For proper adherence to anti-tuberculosis treatment, it is important for several factors to interact.10 In this study, the male sex has been identified as a risk factor for treatment discontinuation. This coincides with other studies mainly due to the fact that in countries like Peru men are more actively employed, although the precarious working conditions often entail: lack of formalized work, lack of unemployment benefits, contracts without vacation time and lack of subsidies for illness, all of which make adherence to treatment even more difficult.11–13 No association was identified between the age groups, although other studies mention that patients in extreme age groups (older than 54 and younger than 15) presented a greater tendency towards non-adherence; this was not observed in our study, possibly due to the homogeneity of the sample.14

Another risk factor for non-compliance was feeling malaise during the treatment; this aspect was assessed in another study that recognizes there being a greater percentage of patients with side effects among the non-complier patients.15 Previous lack of adherence was also identified as a risk factor (OR=7.9).16 Likewise, it is known that people who have abandoned another treatment and those who have presented difficulties for caring for themselves are considered at high risk for discontinuing treatment.17 If the variable for previous non-compliance interacts with poverty, this risk increases to 11.24, which is a worrying fact because the large majority of patients in Peru are poor and many have a history of previous non-compliance.

In the drug use analysis, there was a demonstrated association between recreational drugs and risk of non-compliance. In this case it must be kept in mind that TB patients are a low-income population, and therefore the most widely accessible drug in the country is terokal (adhesive). In the literature, we have found no articles associating this type of drug with non-compliance; however, subjects addicted to other drugs in Nicaragua18 were identified as having a risk that was five times higher. This was also observed in cocaine addicts in New York,19 and in heroin addicts in Barcelona.20 Furthermore, alcohol consumption has also been related with risk of non-compliance,18 although this relationship was not confirmed in the multivariate analysis.

Among the protective factors that were identified, it was observed that the patients who presented an excellent level of satisfaction with the information received had less of a probability of discontinuing treatment, as observed in other studies.19,21 Another protective factor was considering the patient-care hours adequate, (in Peru the primary-care centers offer a minimum of 12h of patient care per day). Likewise, in another study it was mentioned that the flexibility of the timetable and the comfortable setting of the health-care center were related with better treatment adherence.22 Contrarily, risk for non-compliance is three times higher among those who find it difficult to go to the health-care centers.18

In the literature, we found associations between non-adherence and lower levels of education,18 lack of confidence in the treatment23,24 and absence of a stable family environment.14,15,24 Nevertheless, none of these variables related with the patient environment was confirmed in the multivariate analysis. In the same manner, the variables depending on the health-care services related with non-compliance, such as considering the number of medications to be excessive,25 having a negative opinion of the care provided,26 or a poor relationship with the health-care professionals27,28 did not present any association.

The results of this study should be interpreted while considering some limitations: retrospective data collection is a complex task affected by memory biases; the interviewers were conscious of the conditions of the cases and could have influenced the collection of the information; our study did not obtain samples from all the provinces of the country, although it did include four departamentos that represent 80% of cases with discontinued treatment, and in this manner our results could be generalized to the majority of the affected population in Peru. Last of all, despite the delay in data collection and the analysis of these results, the technical regulations for the administration of treatment have not undergone any important changes; therefore, the results are still current and valid.

In conclusion, this study identified that the profile of Peruvian patients who do not comply with anti-tuberculosis treatment are subjects who are usually male, report some type of malaise during treatment, have a history of previous non-compliance, use recreational drugs, do not demonstrate an excellent level of satisfaction with the information received and also feel that they do not have enough time to go to the ESNPyCTBC (National Health-care Strategy for the Prevention and Control of Tuberculosis). On the other hand, it has been observed that the risk for discontinuing treatment in patients with a previous history of non-adherence increases substantially with the presence of poverty. The contribution of our study should be of particular interest to health-care professionals working with TB patients in the ESNPyCTBC as it demonstrates that special care should be taken in patients who present some of the mentioned factors.

Dirección General de Epidemiología, Peruvian Ministry of Health.

The authors declare no conflicts of interest.