Methods
Classification of outcomes and cohorts
For the collection of data on tuberculosis cases notified in 2000, all 51 European countries were requested to classify their outcomes using the six standard categories (‘cured’, ‘completed’, ‘died’, ‘failed’, ‘defaulted’ and ‘transferred’) [TABLE 1] [7]. The first outcome observed within 12 months from start of treatment or diagnosis would be considered definitive. If treatment lasted beyond 12 months for any reason, a case would be classified as ‘other, not evaluated’. Cases lost to follow up were to be classified as ‘defaulted’ (unless fulfilling the conditions for ‘transferred’), and cases diagnosed post mortem were to be classified as ‘died’. Those found to have been wrongly diagnosed as tuberculosis or notified more than once in the same calendar year, as well as those notified from areas not participating in outcome monitoring, were to be excluded from the cohort. Monitoring was limited to new and retreated cohorts of definite pulmonary cases that were culture positive, or smear positive if culture was not available. Data were to be submitted in aggregate form on paper or electronically.

Changes were introduced, beginning with the cohorts of cases reported to European surveillance for the year 2001. Countries were to report outcomes on all the definite cases that had been notified to EuroTB for 2001, including those with unknown previous treatment history. Two additional outcome categories were introduced: ‘still on treatment’ (at 12 months) and ‘unknown’ [TABLE 1]. The ‘still on treatment’ category had already been contemplated in the European recommendations as a way of dealing with previously treated cases failing a full re-treatment course [7]. Instructions on data submission and definitions were developed in English and Russian [8]. Countries were requested to report outcomes in individual format where possible. Participants were invited to give feedback on compatibility between national and recommended definitions.

Other definitions
For the purpose of this article, a new case is defined as a patient with no history of curative, combination antituberculosis treatment or one who has had such treatment for less than four weeks. A retreated case is a patient who had at least one treatment episode lasting four weeks or more before the current notification but not in the same calendar year; a relapse is a retreated case, previously declared cured, and notified again with definite tuberculosis. Multidrug resistance (MDR) refers to resistance to at least isoniazid and rifampicin. ‘Success’ refers to the sum of ‘cured’ and ‘completed’. Countries are grouped in three geographic areas: EU & West (countries of the European Union post-May 2004, plus Andorra, Iceland, Israel, Norway and San Marino), East (countries of the former Soviet Union excluding the Baltic states) and Centre (other countries in the Balkans and Turkey).

Analysis
Outcomes are expressed as the percentage of cases in the respective outcome category divided by all cases included in the cohort. The most recent cohorts reported were used for both numerator and denominator. Data used are those received up to 28 February 2005. For 2000 cohorts, cases classified under ‘other, not evaluated’ were retained in the denominator. Unless stated otherwise, the median of outcomes is used for inter-country comparison. Arithmetic means are used where statistical significance is tested on cases pooled from different countries (P value limit for significance = 0.001). Smear positive cohorts are used for both years in countries where culture positive cohorts were not available.
Completeness of cohorts is calculated as the percentage of definite cases included in outcome monitoring cohorts divided by the number of definite cases previously notified [TABLE 2]. It could exceed 100% if outcome reports included additional cases identified subsequent to initial notification. This commonly occurs after reclassification of cases based on belated retrieval of culture results. Outcome results are discussed for new, definite cases from nationwide cohorts reported in 2001 with 98% completeness or more [TABLE 3]. As completeness tended to be lower in 2000, changes in outcome coding between 2000 and 2001 are discussed solely for countries with >90% completeness in 2000 and reporting more than 10 cases [TABLE 3, countries in bold].

Results
Completeness of cohorts
Whereas 38 of 51 countries submitted outcome data for definite pulmonary cases notified in 2000, the number of countries increased to 42 in 2001. Ten countries did not report outcome information in 2000 or 2001 (Belarus, Croatia, Finland, France, Greece, Luxembourg, Monaco, Spain, Switzerland, Ukraine). In 2000, 19/38 reporting countries had nationwide cohorts with at least 98% completeness, increasing to 31/42 in 2001 [TABLE 2]. The total number of cases included in complete cohorts increased from 25 735 in 2000 to 57 692 in 2001. In 2001, seven countries reported outcome for cases with unknown treatment history, which represented between 1% and 26% of cases reported (1206 cases in total). The number of countries reporting nationwide, complete cohorts increased in all geographic areas. Eleven countries, all from the EU & West, sent individual outcome data.

Compatibility of period of observation and outcome categories
Romania and 20 countries from the EU & West submitted feedback on their coding experience in 2001. Twelve countries (57%) stated that they applied a 12 month maximal observation period, while in the others this was longer (three countries) or not defined. Fourteen countries (67%) reported no incompatibilities between outcome categories proposed and those in national use. Three countries (14%) noted differences with one category while four countries differed in more than one category. ‘Cured’ was not always differentiated from ‘completed’ (four countries), ‘failed’ was sometimes defined differently, or was not available as a category (three countries), and ‘defaulted’ was sometimes applied in a different way (three countries). A number of countries could distinguish between death from tuberculosis or from other causes. One country reported that an outcome could be changed within the 12-month period if, for example, a defaulter resumed treatment after an interruption.

Classification of outcomes in 2001 and changes from 2000
Among nationwide, complete cohorts of new cases in 2001 [TABLE 3], ‘success’ ranged from 54% to 100% (median: 76%). ‘Died’ was more frequent in the EU & West compared with the Centre and East (means: 9% versus 4%, P<10-6). In general, the number of ‘unknown’ was inversely proportional to the total of ‘defaulted’ and ‘transferred’. In 20 countries that reported fewer than 2% of cases as ‘unknown’, cases overall were classified more often as ‘transferred’ or ‘defaulted’ than in the 12 countries with a higher proportion of ‘unknown’ (means: 8% versus 5%; P<10-6). ‘Failed’ was rarely reported in the EU & West (<1%) in contrast to the Centre (3%) and East (8%). Conversely, ‘still on treatment’ was more commonly reported in the EU & West (1%; country range: 0%-15%) than in the Centre and East (0%; 0%-9%).
In 2001, 15 of 31 countries reporting outcomes had cases classified as ‘still on treatment’ (1%-15%) and 15 as ‘unknown’ (1-30%), with higher proportions in both categories amongst retreated cases (data not shown). Three types of shifts in outcome coding could be discerned in 2001 cohorts when compared to 2000 [TABLE 3]
a) ‘other, not evaluated’ shifted to ‘still on treatment’ in Estonia, Latvia and Portugal;
b) ‘other, not evaluated’ shifted to ‘unknown’ in Austria, and possibly in Sweden where this shift was accompanied by an increase in ‘still on treatment’ and a drop in ‘success’;
c) ‘defaulted’ shifted to ‘unknown’ in Ireland.

Discussion
Changes to the outcome monitoring methodology introduced in 2001 were meant to enhance inter-country comparability and ensure that all definite pulmonary cases would be monitored and assigned an outcome. Cases with unknown previous treatment history, or who were still on treatment at 12 months, would be retained in the calculation of cohort completeness. Ensuring completeness would reduce the likelihood of selection bias when reporting outcomes. In countries reporting nationwide outcome data, cases notified in areas or units not participating in monitoring would be classified as ‘unknown’ and kept in the denominator for the calculation of outcome percentages. Reducing the proportion of ‘unknown’ would then become an intermediate goal to improve coverage.
The increase in the proportion of countries submitting nationwide cohorts from 37% to 60%, which more than doubled the size of complete cohorts, is an important achievement in European tuberculosis surveillance. However, sustaining or improving upon this achievement in future is not assured, especially in certain Eastern countries where reporting systems are not yet stable. The definition of a retreated case is not harmonised, particularly in countries of the former Soviet Union, and has at times changed in the interim [9]. This precludes conclusive discussion of outcomes among retreated cases. For many countries, the compatibility between recommended and national outcome monitoring parameters is not known. In countries providing information, the period of observation was not standardised, and this limits inter-country comparison, since chances of success may vary with the duration of evaluation. Another possible source of bias when comparing national programmes is the absence of a lower time limit for defining treatment completion, which may therefore be expected to vary substantially if drug regimens are not standardised. Likewise, ‘success’ may improve if outcome is changed after the case first satisfies the definition of another outcome category (eg, reclassification of defaulters). There is evidence that ‘defaulted’, ‘transferred’ and ‘unknown’ tend to be used interchangeably, thus reducing the possibility of meaningful comparison of these categories at European level. Having a sub-category of ‘died’ for cases dying directly from tuberculosis rather than a concurrent cause could be useful in programme monitoring [7] but this would require a harmonised definition of which cases to include.
The shift observed from ‘other, not evaluated’ to the ‘still on treatment’ category was anticipated, since the former category was reserved for cases on prolonged treatment. In Portugal, where drug resistance is low, this shift has largely been caused by the continued use of long term chemotherapy regimens for non-MDR tuberculosis (A Fonseca Antunes, personal communication, 11 May 2005). In Estonia, however, ‘still on treatment’ cases were mostly MDR (data not shown), and a similar explanation would be likely for Latvia, another Baltic state with a high MDR burden [10]. In Lithuania, the proportion of ‘still on treatment’ in 2001 was more modest than in neighbouring Baltic states despite similar MDR levels [11]. This shift was not observed in other former Soviet countries probably because MDR cases were mostly classified as ‘failed’ both in 2000 and 2001. Such differences may represent variability in patient access to drug-susceptibility testing and appropriate chemotherapy.
Where access to laboratory testing is good, MDR cases are commonly identified ahead of the fifth month of treatment and embarked on long term medication, making it more likely that they are classified as ‘still on treatment’ at 12 months rather than ‘failed’. In much of western Europe, ‘failed’ is rarely used, because the follow-up bacteriological information required to define this category is often not captured by surveillance systems. In the new definitions for outcome monitoring in MDR cases, ‘failed’ is reserved for cases who are bacteriologically positive at a much later stage in the course of their second line treatment [12]. Until such time as second line treatment becomes widely available in all European countries, the category ‘failed’ will have to be retained. As more countries develop the capacity to rapidly diagnose drug resistance and to change over to second line regimens, the ‘still on treatment’ option will have a wider utility, and the ‘failed’ category will become less important.

In conclusion, outcome should be reported for all definite pulmonary cases notified, regardless of treatment history. The 12-month maximum period of observation should be applied for the classification of all outcomes. Cases treated beyond 12 months and having MDR tuberculosis (identified at start or during the current treatment episode) would form the subject of continued monitoring with a longer period of observation (24-36 months).
The eight outcome categories proposed can be used for national outcome monitoring. Owing to the incomplete differentiation of ‘cured’ from ‘completed’, and to the non-uniform use of ‘defaulted’, ‘transferred’ and ‘unknown’ in classifying cases lost to follow up, analysis of outcome monitoring at European level and inter-country comparison should be based on five categories: ‘success’, ‘death’, ‘failed’, ‘still on treatment’ and ‘others’. European countries should further standardise their parameters for tuberculosis outcome monitoring in order to enable a more meaningful comparison of programme performance between countries and over time. In the West, where tuberculosis patients are older and deaths are thus expected to be higher, it is all the more imperative to bolster patient follow up if countries are to approach the 85% success target.
The WHO and EuroTB should continue working together to harmonise monitoring methodology, promote the evaluation of control programmes and support countries to provide nationwide, complete data. In order to better understand the determinants of outcome, collection of tuberculosis notification data on an individual case basis should be promoted.

† Andrea Infuso, EuroTB scientific coordinator, died suddenly on September 20, 2005. This Euroroundup is a posthumous publication.

Acknowledgements
We acknowledge the contribution of European national correspondents (listed) and other national surveillance staff supplying the data: M Coll Armangué (Andorra), H Hafizi (Albania), M Safarian (Armenia), JP Klein (Austria), I Mammadova (Azerbaijan), M Wanlin, A Aerts (Belgium), Z Dizdarevic, B Stefanovic (Bosnia & Herzegovina), D Stefanova (Bulgaria), I Gjenero Margan (Croatia), P Mavrides (Cyprus), L Trnka (Czech Republic), P Andersen (Denmark), V Hollo (Estonia), W Haas (Germany), D Kozma (Hungary), T Blöndal (Iceland), J O’Donnell (Ireland), D Chemtob (Israel), D Caraffa de Stefano (Italy), GB Rakishev (Kazakhstan), AS Alisherov (Kyrgyzstan), J Leimans (Latvia), E Davidaviciené (Lithuania), L Simonovska (Macedonia F.Y.R), A Pace Asciak (Malta), V Burinski (Republic of Moldova), P van Gerven (Netherlands), E Heldal, B. Winje-Askeland (Norway), M Korzeniewska-Kosela (Poland), A Fonseca Antunes (Portugal), E Corlan, E Ibraim (Romania), M Perelman (Russian Federation), A Sorcinelli (San Marino), D Popovac (Serbia & Montenegro), E Rajecova (Slovakia), J Sorli (Slovenia), V Romanus (Sweden), UI Sirodjiddinova (Tajikistan), E Kibaroglu (Turkey), BD Jumaev (Turkmenistan), J Watson, B Smyth, J McMenamin (United Kingdom), GT Uzakova (Uzbekistan). Other contributors to discussion on the subject have included P Barboza (EuroTB), J Veen (KNCV) and D Bleed and C Dye (WHO), as well as past and present members of the EuroTB Advisory Committee not mentioned above, namely L Clancy, M Diez, F Drobniewski, M Forssbohm, H Rieder, P Ruutu, and R Zaleskis.