INTRODUCTION
Tuberculosis (TB) poses a significant global disease burden particularly in Africa, Asia, Eastern Europe and Latin America [Reference Murray and Lopez1–Reference Dye3]. In 2002, the estimated incidence in England, Wales and Northern Ireland was 12·9 cases/100 000 per year, while the corresponding rate of reported TB in the East of England was 6·6 cases/100 000 per year [Reference Crofts4]. New cases of the disease predominantly occur amongst ethnic groups from high-incidence countries, the homeless and those with material deprivation [Reference Crofts4–Reference Delgado6]. The proportion of elderly patients with TB in the United Kingdom is also said to have risen [Reference Teale, Goldman and Pearson7], although recent findings indicate the highest prevalence amongst young adult males [Reference Crofts4].
Patients who have sputum smears positive for acid-fast bacilli are more infectious than sputum smear-negative cases [Reference Reider8–Reference Hayward11]. Moreover, delayed treatment of infectious index cases is associated with increased latent TB infection within a community [Reference Alseda and Godoy12, Reference Ponticiello13]. Consequently, prompt identification and successful treatment of patients and contacts, particularly those with sputum smear-positive pulmonary TB, are fundamental to controlling the spread of the disease.
In England and Wales, Enhanced Tuberculosis Surveillance started in 1999 with the purpose of continuously providing information on the epidemiology of TB including evaluation of disease trends in various population subgroups. In 2002, TB treatment outcome surveillance commenced as part of enhanced surveillance to provide data for determining the effectiveness of TB control [Reference Antoine and French14, Reference Rose15]. Effective control requires that patients complete their planned treatment and remain traceable by the health system. Identification of patients at risk of inadequate treatment could permit the tailoring of interventions to control TB.
Little published information is available on patient outcomes after diagnosis with TB in England. One previous study [Reference Doherty, Spence and Davies16] indicated an increase in mortality from TB in the elderly, who are also less likely to complete their anti-TB treatment [Reference Teale, Goldman and Pearson7]. However, these studies either excluded children or were prone to misclassification of deaths. Furthermore, previous studies have seldom examined the characteristics associated with important patient outcomes. We aimed to describe patient outcomes one year after diagnosis of TB and commencement of treatment, and to explore the factors associated with treatment completion, loss to follow-up and mortality.
METHODS
Study population
The study was set in the East of England, which has a population of 5 400 000 (Office for National Statistics, 2001). Incident TB cases were those reported from 2000 to 2003. In a population-based prospective cohort design, cases of TB reported during the years 2001 and 2002 were followed up for a year after the date of reporting.
Data collection
Cases were reported by clinicians through Enhanced Tuberculosis Surveillance to the Regional Coordinator in the East of England (C.B.). A confirmed case of TB was one that had been verified by culture of Mycobacterium tuberculosis complex. Probable cases were diagnosed clinically and commenced on treatment in the absence of a positive culture. The total number of reported cases excluded those that had been de-notified, which refers to patients initially reported through enhanced surveillance but subsequently determined as not having TB or those who were only receiving chemoprophylaxis against the disease [Reference Rose15].
Outcome data was collected on standardized reporting forms by physicians and TB specialist nurses, and returned to the regional TB surveillance coordinator for collation. Forms were completed 10–12 months after the start of therapy. If outcome data had not been received for a case after fourteen months from the start of treatment, a reminder form was sent out.
Anonymized information on reported cases included demographic, clinical and bacteriological characteristics. Outcomes were classified as treatment completion, loss to follow-up and vital status one year after initiation of anti-TB therapy. Treatment completion was defined as successful completion of a prescribed course of anti-TB treatment within 12 months of commencement. Loss to follow-up included outcomes that were unknown to the clinician and those that were not reported to the surveillance coordinator. Reporting clinicians determined whether TB caused death (considered to be the main cause of death), contributed to death (considered to be contributory to death but was not the main cause) or was incidental to death (not related to death) or whether the association between TB and death was unknown. However, our analyses considered all-cause mortality.
Postcodes were linked to enumeration districts to produce the Townsend Index. This score is a composite measure of material deprivation based on four factors [Reference Townsend, Phillimore and Beattie17] (unemployment, overcrowding, car ownership and home ownership) derived from the 2001 population census for England and Wales, and takes into account the mean and standard deviation relative to England and Wales. A higher score denotes greater deprivation and a score above 0 implies that the deprivation was greater than the mean for England and Wales. Ward level Townsend scores were categorized into quartiles including least (−5·43 to −1·14), low (−1·13 to +1·80), high (+1·82 to +4·73) and highest deprivation (+4·74 to +8·28).
Statistical analysis
We estimated the crude, age, gender and ethnic-specific incidence rates of TB from cases reported to the surveillance unit of the East of England, between 2000 and 2003. Incidence rates were derived by dividing the number of reported cases by the person years (pyr) at risk, obtained by multiplying the estimates of the population of the East of England at the middle of the 4-year observation period (2001) by four. Rates were expressed per 100 000 pyr at risk. Ninety-five per cent Poisson confidence intervals (95% CI) were calculated for incidence rates.
One year after initiation of anti-TB chemotherapy, the number and proportion of cases with available information on different TB treatment outcomes was determined. Independent t tests, Pearson χ2 test and Fisher's exact test were used to compare the following patient characteristics between population subgroups: age, gender, country of birth, Townsend score, disease site, previous illness episode, sputum smear test result and number of years since entry to the United Kingdom for foreign-born cases. Stepwise logistic regression was used to examine the association between these variables and patient outcomes. We examined differences in characteristics of patients with missing data on covariates with those for whom data was available. For all statistical tests, a two-tailed P value <0·05 was considered statistically significant. Analyses were performed using Intercooled stata 8.0 (Stata Corporation, College Station, TX, USA).
RESULTS
Between 2000 and 2003, 1300 cases of TB were reported, 127 having been de-notified. The crude incidence rate of TB was 6·0 cases/100 000 pyr (95% CI 5·7–6·4). The highest incidence rate (14·9, 95% CI 12·9–17·1 cases/100 000 pyr) occurred in the 25–29 years age group, while the lowest was in the 10–14 years age group (0·8, 95% CI 0·4–1·4 cases/100 000 pyr). The age-adjusted incidence rates were 7·1 and 5·3 cases/100 000 pyr for males and females respectively. Black Africans had the highest rates (328·6, 95% CI 286·9–374·6 cases/100 000 pyr) while rates were lowest in the White ethnic group (2·3, 95% CI 2·1–2·5 cases/100 000 pyr). There was a positive correlation between TB incidence and the proportion of the population with a Townsend Index ⩾0 (Spearman's rank correlation coefficient=0·63, P<0·001).
Five hundred and eighty-four confirmed or probable cases of TB were initially reported during 2001 and 2002 (280 and 304 respectively); nine were found not to have TB and therefore excluded from the analysis. The characteristics of patients reported during 2001 and 2002 were not significantly different to those of patients reported in 2000 and 2003, and are shown in Table 1. A total of 343 (59·7%) patients had pulmonary TB and 124 (53·0%) of the pulmonary cases with known sputum smear results were positive for acid alcohol-fast bacilli. Extra-pulmonary disease was mostly localized to extra-thoracic lymph nodes (84, 36·2%), the pleura (33, 14·2%) and intra-thoracic lymph nodes (21, 9·0%). Other extra-pulmonary forms were TB of the bones (15, 6·5%) particularly the spine, gastrointestinal tract and peritoneum (9, 3·9%), the genitourinary system (8, 3·4%) and the central nervous system (7, 3·1%), especially meningitis. Eight (3·5%) patients had miliary or disseminated TB, 32 (13·8%) had mycobacteria localized in other body tissues, and 15 (6·5%) were diagnosed with TB in more than one extra-pulmonary site. The mean age of reported cases was 43·1 years (s.d.=21·2), 55·2% were male.
Table 1. Characteristics of patients completing treatment or lost to follow-up within 12 months of initiation of anti-tuberculosis treatment
* Missing data have been omitted.
† Pearson χ2 test unless otherwise indicated.
‡ χ2 test for trend.
§ Foreign-born cases.
¶ Pulmonary tuberculosis cases.
‖ Fisher's exact test.
Patient outcomes, following initiation of anti-TB therapy, are summarized in the Figure. In total, 439 (76·3%) patients completed treatment, 45 (7·8%) were lost to follow-up; 91 (15·8%) did not complete treatment of which 49 (53·8%) were reported to have died within a year of initiation of treatment.
Fig. Treatment outcomes of tuberculosis (TB) cases reported through the TB treatment outcome surveillance in the East of England between 2001 and 2002 (* excluded from the analysis).
Treatment completion
Patients who completed treatment were more likely to be younger (χ2 for trend=20·8, P<0·001) and more deprived (χ2 for trend=5·2, P=0·023). Patients with extra-pulmonary disease were also more likely to complete treatment compared to those with pulmonary TB (Table 1).
Loss to follow-up
Foreign-born patients were 2·35 (95% CI 1·03–5·32, P=0·041) times more likely to be lost to follow-up compared to those born in the United Kingdom as shown in Table 2. Moreover, amongst the foreign-born cases, those who entered the United Kingdom more than 3 years prior to notification were less likely to be lost to follow-up than cases that were reported within 3 years of entry in the United Kingdom (Table 1). There was no significant association between deprivation and loss to follow-up.
Table 2. Association between patient characteristics and outcomes within twelve months of initiation of anti-tuberculosis treatment
* Logistic regression model of best fit.
† Wald test P value.
‡ P value from likelihood ratio statistics.
§ The odds ratio for an increase in age of 1 year.
¶ There was collinearity between country of birth and number of years since entry in the United Kingdom; the latter was omitted during regression analysis. Foreign birth was the only independent factor associated with loss to follow-up on stepwise multivariate analysis.
All-cause mortality
A majority of deaths (57·1%) were reported to be TB-associated. Mortality was highest amongst older patients (χ2 for trend=50·7, P<0·001), patients with pulmonary disease (χ2=12·2, P<0·001), patients born in the United Kingdom (χ2=8·3, P=0·004) and those who reported a previous TB episode (χ2=5·5, P=0·019). Notably, few (8, 3·8%) patients with extra-pulmonary TB were reported to have died.
After adjustment for variables that remained significant during stepwise logistic regression, only increasing age [odds ratio (OR) 1·05, 95% CI 1·04–1·07] and pulmonary disease site (OR 2·73, 95% CI 1·21–6·15) were independently associated with mortality within 12 months (Table 2).
DISCUSSION
The clinically reported incidence of TB in this setting was low but differed considerably by ethnicity, as reported in other studies [Reference Crofts4–Reference Teale, Goldman and Pearson7]. When incidence rates amongst various ethnic groups are compared, they appear to reflect those from the patients' places of birth. Consequently, the high incidence amongst Black Africans was similar to the rate in most of Sub-Saharan Africa [2]. The high background risk of infection could offer a partial explanation, especially because a majority of foreign-born cases reported recent entry in the United Kingdom [Reference Reider8]. However, whether infection predominantly occurs in the country of birth would require molecular studies on latent infection among foreign-born individuals.
Completion of prescribed anti-TB treatment is often used as an outcome measure. The majority of patients in this cohort completed their treatment. It is also evident that failure to report treatment completion was mainly because of longer durations of drug therapy and the occurrence of death. Nevertheless, treatment completion was slightly lower than the national average of 78% [Reference Antoine and French14] and the World Health Organisation (WHO) target of achieving 85% treatment success, except for cases aged <25 years [Reference Zumla and Mullan18–Reference Sharma and Liu20]. Our results indicate that patients, who were younger, more deprived and those who had extra-pulmonary TB were more likely to complete treatment. The association between age and treatment completion was comparable to findings reported by the Health Protection Agency [Reference Antoine and French14]. Older patients frequently have multiple pathologies and are likely to be on complex drug regimens making adherence to anti-TB treatment difficult. More importantly, they could have succumbed to the disease before completing planned treatment.
The comparatively greater proportion of more deprived patients completing treatment is not consistent with previous findings. Current smoking [Reference Chang, Leung and Tam21], patients running out of anti-TB agents during self-treatment [Reference Kaona22] and unstable housing [Reference Tulsky23] have been linked to lower treatment completion rates. The seemingly higher proportion of treatment completion amongst more deprived patients in this study can be explained by the comparative excess in deaths of less deprived patients who were also older, before treatment completion. It is also conceivable that with universal access to anti-TB treatment, completion of long-term treatment should not differ significantly by deprivation except when patients have other risk factors like homelessness and alcoholism.
Little published evidence is available on loss to follow-up, particularly in England. The proportion of TB patients lost to follow-up was slightly higher than that reported (6%) by Conaty et al. [Reference Conaty24] at a central London hospital. In our study, only foreign country of birth was associated with loss to follow-up, especially for patients who reported entry in the United Kingdom within the preceding 4 years. It is possible that this was because some foreign-born patients returned to their countries of birth or permanent residence. Studies elsewhere [Reference Albuquerque25–Reference Helbling27] have found an association between loss to follow-up or default and alcoholism, homelessness and foreign birth with uncertain legal status rather than the patient's country of birth.
About half of the foreign-born patients had pulmonary TB. Those who could not be traced were therefore potentially infectious and also at risk of antimicrobial resistance. Additionally, pulmonary cases had more unfavourable mortality experiences. This underscores the importance of enhancing follow-up of foreign-born cases on anti-TB chemotherapy, with the aim of ascertaining favourable outcomes. For example, attention should be paid to lifestyle and environmental factors that may adversely affect follow-up [Reference Brainard26, Reference Coker28, Reference Raviglione and Uplekar29].
Directly observed treatment (DOT) has been recommended for patients with risk factors for poor adherence or default from treatment. In the United Kingdom, guidelines from the Joint Tuberculosis Committee recommend DOT for patients who are less likely to adhere, for example the homeless, alcohol or drug abusers, patients with a history of poor adherence; or those with multiple drug resistance [30]. Perhaps recent migrants should be added to this list of risk factors. Whereas evidence from systematic reviews indicate that the benefits and cost effectiveness of universal DOT relative to alternative adherence-promoting strategies, in a developed country setting, remain uncertain [Reference Volmink and Garner31, Reference Garner and Volmink32], patients at risk of unfavourable outcomes are likely to benefit. Indeed, tailored approaches for such risk populations are an important component of the WHO's new Stop TB strategy to TB control [Reference Garner and Volmink32].
We also found that pulmonary disease site and increasing age were independently associated with all-cause mortality. This study corroborates previous findings of the higher case fatality of pulmonary TB [Reference Reider8] compared to extra-pulmonary disease, largely because the latter mostly occurred as benign, less severe lymph node TB. Teale et al. [Reference Teale, Goldman and Pearson7] reported that elderly patients in the United Kingdom were at greatest risk of death from TB. Despite there having been no adjustment for disease site in the previous study, a preponderance of deaths amongst the elderly can be expected.
Findings elsewhere [Reference Centis and Migliori33] have pointed to a higher age-related death rate amongst indigenous residents compared to immigrants, although the later were more likely to default on their anti-TB treatment. We did not find differential death rates by country of origin. Outcomes after diagnosis of TB therefore appear to vary by distribution of socio-demographic and clinical characteristics in different geographical settings and could be influenced by programmatic treatment strategies [Reference Helbling27, Reference Samman34–Reference Diez Ruiz-Navarro37].
Limitations
Surveillance data on sputum smear test and specimen culture were incomplete, potentially leading to misclassification of TB cases. Missing data though, had no effect on the association with patient outcomes. The relative contribution to death, treatment completion or loss to follow-up, of comorbid states and health behaviours such as smoking, alcohol and drug abuse, were beyond the scope of this study. Notably, HIV-seropositive TB patients seem to have different disease characteristics and outcomes [Reference Centis and Migliori33, Reference Aliyu, Salihu and Ratard38] regardless of adherence to drug treatment [Reference Rocha39].
CONCLUSIONS
The population-based approach to TB surveillance provides a very useful basis for examining determinants of patient outcomes after diagnosis, despite the drawback of incomplete microbiological data. The majority of patients completed their recommended treatment and were followed up. Elderly patients were both less likely to complete treatment and at greater risk of death, suggesting the need for closer monitoring of TB treatment. Similarly, more attention should be drawn towards patients with pulmonary TB to guarantee completion of treatment, minimize spread and possibly avert premature death. Foreign-born TB patients, especially those reporting recent entry into the United Kingdom, might benefit from closer follow-up to ensure favourable outcomes.
ACKNOWLEDGEMENTS
We thank Mike Streather of the Eastern Region Public Health Observatory for providing information on indices of material deprivation and population data. The efforts of reporting clinicians and TB nurses are also gratefully acknowledged.
DECLARATION OF INTEREST
None.
