Tuberculosis treatment outcomes in Europe: a systematic review

Identification and selection of articles

Published reports dealing with TB treatment outcomes in Europe were reviewed. Studies were identified through a computerised search of the following databases: PubMed (1993–2003), EMBASE (1993–2003), TOXLINE (1993–2003), Serfile (1998–2003), and CAB Abstracts (1993–2003). MEDLINE was searched through PubMed. The term tuberculosis was searched with all of the following as a combination of free text and thesaurus terms in different variations: treatment outcome, treatment failure, successful therapy, unsuccessful therapy, mortality due to, surveillance, surveys, and therapy follow-up. WHO documents were systematically searched on their website. The International Journal of Tuberculosis and Lung Disease was selected as the key journal for hand-searching.

Any original study from a European country was included in this review, whereas comments, editorials and reviews were excluded. The articles were included if they estimated both successful and unsuccessful treatment outcomes in the total population of TB cases who started treatment, including patients lost to follow-up, those transferring out and/or those for whom outcomes were not reported.

Articles were excluded if they dealt with extrapulmonary TB alone, chronic TB or MDR-TB, a mycobacterium other than tuberculosis, high-risk groups such as HIV-positive or neoplastic patients, or previously treated cases alone. Case studies reporting results regarding diagnostic/laboratory methods, drug efficacy tested in vitro or through clinical trials, TB surgical procedures, side-effects of drugs, TB vaccination, and indications for management or treatment of TB cases were also excluded.

Outcome measures and definitions

The outcome measures of TB treatment were assessed as percentages of successful and unsuccessful outcomes, and of deaths during treatment, among all patients who started therapy.

Cases reported in the articles presented each WHO outcome definition for surveillance concerning TB patients before and after therapy 1, 4, 5. Successful outcomes in the review included cured patients and those who completed treatment. Two definitions were used for unsuccessful outcomes, one included failures and defaulters 1, 4, 5 and the other patients lost to follow-up or transferred-out.

European countries were selected according to the WHO regional classification, which includes Israel and Turkey.

Statistical analysis

Where appropriate, the pooled percentages for each TB treatment outcome were used in a random effects meta-analysis. The se and 95% confidence interval (CI) of outcomes were calculated from data reported in each article. Heterogeneity between studies was assessed using the Chi-squared test, and the variation due to heterogeneity across the studies estimated by calculating I². Meta-regression was performed in order to analyse the association between treatment outcome and: mean age; percentages of males, immigrants and MDR-TB among patients; year of TB occurrence; and study design. It was not used to analyse the association with percentages of patients treated with WHO standard chemotherapy because only nine studies reported this information. For variables associated with treatment outcome, pooled stratified estimates were calculated and sensitivity analysis of potential bias performed.

Successful tuberculosis treatment outcomes

Successful TB treatment outcomes were below the 85% threshold suggested by the WHO. The estimate of 74.4% found is coherent with those from WHO surveillance 3. The lower percentage of successes observed in previously treated than in new TB cases, although estimated only for three studies, was also consistent with previous knowledge, since previously treated patients are more likely to have had or acquired multidrug resistance to Mycobacterium tuberculosis 35.

The estimate of cured patients could not be used even though it is the best indicator of the impact of a TB control programme. It measures transmission interruption, as well as cured patients, but requires microbiological tests to be carried out before and after therapy in each pulmonary TB case. From the low percentage of cured patients, it was hypothesised that this practise is not common in European countries, and, therefore, that this percentage underestimates the conversion rate and impact of TB control programmes.

Although lower than in Western countries, the pooled estimate of treatment success in the former USSR was not significantly different. This was not as expected because of the high prevalence of MDR reported in the former USSR by the WHO 36. The present authors are unable to explain these results, but suggest that most studies from Western countries in the present review, although population-based, do not give valid estimates of MDR-TB prevalence since the drug sensitivity data referred to TB patients selected on the basis of clinical criteria. It is possible that those who did not respond to standardised therapy have been more frequently studied for sensitivity than others; therefore, the prevalence of MDR-TB could be overestimated by the studies from Western Europe. This could also explain, in part, the low observed percentages of successful therapy.

The heterogeneity between the studies could not be explained by subgrouping by national data in Switzerland and Italy. Differences in MDR-TB levels explained the heterogeneity observed among Italian studies.

Finally, the results obtained in Poland could be due to the epidemiological characteristics of infection in that country; however, the methods used for TB surveillance could also have influenced the observed results.

Determinants of outcome

MDR-TB levels were inversely associated with successful treatment. It was found that a proportion of >9% MDR-TB resulted in an important significant reduction in successful outcomes. Although this result needs to be confirmed by other observations, it represents the first quantitative estimate of the MDR-TB threshold in the population that influences TB treatment outcomes. The selection of patients studied for MDR prevents estimation of the true MDR-TB prevalence in the population, as discussed above, but is unlikely to distort the results of the regression.

People aged <45 yrs have already been reported as at higher risk of MDR-TB in many European countries 37–40, the rationale for this being that rifampicin was introduced as part of therapy in 1969, more recently than other drugs 41. The lower percentage of successful outcomes found in people aged <44 yrs could be due to a higher prevalence of MDR in younger patients.

The lack of association between successful outcome and foreign-born status is surprising. Since immigrants were treated as a single group in the articles, regardless of their countries of origin (and whether the countries were at low or high risk of MDR-TB), this result might have been influenced by a dilution of the effect. Studies of immigration as a risk factor for MDR-TB have clearly shown this point 40. Notably, foreign-born status was only studied in Western European countries, whereas no study from the former USSR or Eastern Europe reported data on immigrants.

Unfortunately, factors related to the characteristics of treatment, such as drugs, dose and timing of therapy, could not be analysed because they were not reported in the articles.

Unsuccessful tuberculosis treatment outcomes

Unsuccessful outcomes, defined as failure and treatment interruption, are strong predictors of MDR 36, 42, but other outcomes, such as death, loss to follow-up, transfer-out, or those for whom the outcome is unknown, contribute to the low threshold level of success. A valid estimate of these outcomes is essential in evaluating TB control programmes as well as in suggesting adequate corrections.

In the present review, no single factor explained the heterogeneity of unsuccessful outcomes between studies. The first hypothesis investigated was that misclassification of unsuccessful outcomes could have occurred in the articles, due to the different interpretations given to WHO definitions of these outcomes. Indeed, four studies in the present review did not explicitly report numbers for loss to follow-up or transfer-out 11, 14, 15, 34, but the high percentages (up to 20.0%) of patients with unknown outcomes suggest that these are included in this group. Ten other studies specified those lost and transfer-out but did not report cases with unknown outcomes 9, 12, 19–22, 28, 29, 31, 33. Two studies included transfers-out among the unsuccessful outcomes 16, 19, giving them the same meaning as failure or interruption of treatment.

For these reasons, a definition of unsuccessful outcome was also used that included patients lost to follow-up and transferred-out. This definition gave an estimate nearer to that reported by WHO surveillance, but the heterogeneity remained unexplained. No association was found between unsuccessful outcome and any factor, regardless of the definition used.

A few studies have reported an association between unsuccessful outcome and factors related to the social characteristics of patients. Treatment interruption has been associated with immigrants in Italy 23–27 and Switzerland, where asylum seekers and refugees showed the highest risk 33. Interruption was also associated with homelessness, injecting drugs and alcohol dependence in Hamburg, Germany 19. In Spain, homelessness was a risk factor for transfer-out and HIV-positivity, and injecting drugs for unsuccessful treatment. Finally, one study analysed factors related to treatment and found that no standard therapy in the initial or secondary phase of treatment was associated with unsuccessful outcome or death 22.

Limitations of the present study

The most important limitation of the present study is the large number of articles with missing data on factors analysed; seven gave no information on sex, and 11 none on foreign status. Sensitivity analysis of bias for MDR-TB, the factor influencing outcome, showed no important influence on regression results, although only 12 studies, of the 26 included, reported this information.

Chronic TB cases should have been excluded from the treatment outcomes analysis, as treatment success is known to be less than in those treated with SSCC 1, 43. Unfortunately, few authors reported this explicitly, and only one Russian author clearly defined the patients treated, by including chronic cases 12. This hypothesis seems to be confirmed for studies conducted in the former USSR, where only new TB revealed no heterogeneity. The possible inclusion of chronic cases in other articles could contribute to the low proportion of successes.

Finally, the possible misclassification of new and previously treated cases 44 could explain the present somewhat surprising results, such as the higher percentage of successful outcomes observed in studies which included both new and previously treated cases than in those studying new TB alone.

Conclusions

Since successful TB treatment outcomes fell below the 85% threshold, in most European countries, enhancement of national TB control programmes is desirable.

A threshold of >9% MDR-TB was found to have the greatest negative impact on successful outcome. Further studies, possibly including representative estimates of MDR-TB prevalence, are needed to confirm this result.

Treatment characteristics, such as the drugs used, therapeutic regimens, duration of therapy and outcome of previous treatment, need to be more consistently reported in order to identify the factors related to inadequate treatment and permit improvement of TB control programmes 36, 45.

The different meanings attributed to definitions of unsuccessful outcomes possibly prevented calculation of their pooled estimates and made it difficult to analyse the associated factors. Discussion regarding the difficulties in using World Health Organization definitions of treatment outcomes could help in the comparison of survey results.