83 publications (69 published papers and 14 abstracts) published between January, 1994, and April, 2009, were eligible for inclusion (). After compiling multiple publications of the same study a total of 78 studies were included. Most studies (55) were published in the past 6 years, with the largest numbers of publications being in 2008 (12) and 2007 (13).
Of those studies included, 30 were from congregate settings, ten from voluntary counselling and testing settings, three from programmes preventing mother-to-child transmission (PMTCT), 16 from antiretroviral therapy and medical clinics, ten contact tracing studies, eight population-based surveys of the prevalence of tuberculosis, and one study among men who have sex with men and injecting drug users. Most studies were from sub-Saharan Africa (41); others were from the Americas (22) and Asia (19).
The proportion of eligible individuals who agreed to participate in these studies was variable, ranging from 40–100% in prisons, 44–100% in voluntary counselling and testing and PMTCT settings, 35–100% in clinic-based settings, 83–100% in contact tracing studies, and 66–100% in population-based surveys (webappendix
). Clear external and internal quality-control procedures for microbiological and radiological investigations were only recorded in three studies in congregate settings, two studies in clinic-based settings, one study in household contacts, and four population-based surveys.
The prevalence of newly diagnosed tuberculosis varied greatly between different countries and specific target populations (). The minimum was 0·01% in a contact-tracing study from Peru and the maximum was 24·7% in an antiretroviral therapy clinic in South Africa. Median NNS varied greatly, ranging from 148 in population-based prevalence surveys to just 12 in voluntary counselling and testing services and in antiretroviral therapy and medical clinics ().
Tuberculosis prevalence and the number needed to screen to identify one new case in different target groups
Results of studies of intensified case finding in prisons, psychiatric hospitals, mines, and refugee camps are presented in the webappendix
. HIV status was established in three of 30 studies and screening strategies varied substantially, with most using an initial questionnaire on the symptoms of tuberculosis followed by diagnostic testing of people suspected to have tuberculosis. Three of six studies in miners combined the screening of symptoms with chest radiology to define those people suspected of having tuberculosis, whereas the other three studies used sputum microbiology irrespective of symptoms.
The median NNS in all prisons was 40 (range 14–833), but was lower in studies in prisons in sub-Saharan Africa (median 28, range 14–55; ). The prevalence of HIV was established in only one study in a Brazilian prison (25%).41
Studies in miners and ex-miners found a median NNS of 43 (range 20–86; ). In one study enrolling only miners infected with HIV, the prevalence of newly diagnosed tuberculosis was 4·9%.46
Of the remaining five studies only one study tested miners for HIV, and reported a prevalence of 27%.47
Most studies in voluntary counselling and testing clinics, PMTCT programmes, and in groups at high risk of infection with HIV (men who have sex with men and injecting drug users)100
identified people suspected of having tuberculosis by screening their symptoms before using diagnostic tests (webappendix
). Sputum microscopy was the only microbiological test available in half of the studies. The median NNS was 12 (range 4–123) in voluntary counselling and testing settings and 44 (range 29–47) in PMTCT settings ().
Intensified case finding was done as part of isoniazid preventive therapy programmes or before antiretroviral therapy was started, in individuals infected with HIV accessing health care or enrolled in home-based care services. Half of the studies used questionnaires on the symptoms of tuberculosis as the first step before microbiological investigations for symptomatic individuals (webappendix
). Median NNS was 12 (range 4–71; ).
Results from contact-tracing studies (webappendix
) are not directly comparable with results from other settings since most studies included children. Microbiological examinations were only done in individuals that were symptomatic or in individuals with positive tuberculin skin tests. Median NNS was 48 (range 7–10 000; ).
Five of eight population-based surveys of the prevalence of tuberculosis in settings with high prevalence of HIV used a step-wise screening approach with screening for the symptom of tuberculosis preceding microbiological examination. The remaining three studies examined the sputum in all individuals irrespective of symptoms (webappendix
). Median NNS was 148 (range 29–5000; ).
summarises data from 47 studies in countries with generalised epidemics of HIV. More than half the studies (27) reported a prevalence of newly diagnosed tuberculosis in the screened population of greater than 3% (NNS less than 33). Excluding population surveys, two-thirds (26 of 39) of the studies reported prevalence of newly diagnosed tuberculosis of greater than 3%.
Prevalence of tuberculosis among individuals screened in different settings in countries with generalised epidemics of HIV
The screening strategies varied widely across the studies. Symptom screening was used in all but one of the prison studies, screening of mine workers invariably included chest radiography, and all contact-screening studies used symptom screening and assessment of responses to the tuberculin skin test. Symptom questionnaires were diverse, ranging from any kind of respiratory symptoms to various durations of productive cough plus or minus weight loss, night sweats, fatigue, fever, and haemoptysis. Similarly, the number and timing of sputum samples varied substantially.
In 12 studies that included only individuals infected with HIV, sputum examination was done irrespective of symptoms. Nine of these studies were done in antiretroviral therapy services, medical clinics, or home-based HIV care programmes and the remaining three were done in voluntary counselling and testing centres. All but one of these studies did both smears and cultures. 17 studies screened individuals infected with HIV with symptom-based questionnaires preceding sputum examination of people suspected to have tuberculosis. These studies were done in antiretroviral therapy and medical clinics (seven studies), voluntary counselling and testing centres (seven), and PMTCT programmes (three). Of these 17 studies, seven only did sputum smears whereas ten did both smears and cultures. In the subset of studies in which both smears and cultures were done, the prevalence of newly diagnosed tuberculosis in individuals infected with HIV was slightly lower (weighted median 9·8%) in studies only investigating preselected people suspected to have tuberculosis on the basis of symptom screening compared with studies doing sputum smears and cultures on everyone irrespective of symptoms (weighted median 11·6%).
Because the strategies and yield of intensified case finding were very heterogeneous between studies done in different countries and patient groups, we sought to identify variables independently associated with the yield of tuberculosis by use of metaregression analysis ().
Univariate analysis of factors potentially affecting yield of intensified case screening in populations with mixed or unknown HIV status
Factors potentially affecting yield of intensified case screening in populations with individuals infected with HIV
Of the 78 studies included in the analysis, some target populations included only individuals infected with HIV (30 studies), whereas the remainder included individuals with and without HIV (congregate settings, contact screening, and population-based prevalence studies). Metaregression analysis was therefore stratified with regards to HIV status of the screened population.
In the univariate analysis of studies with patients with mixed or unknown HIV-status (48 studies), national prevalence of tuberculosis and HIV, screening strategy of microbiological investigations in all individuals, and availability of culture were not associated with yield of screening ().
By contrast, univariate analysis of studies that only included individuals infected with HIV (30 studies) found that both the use of symptom prescreening and the country prevalence of tuberculosis were associated with the detected yield of tuberculosis (). However, the availability of culture and the national prevalence of HIV were not associated with the detected yield of tuberculosis. Multivariate analysis showed that microbiological sputum examination (smear or culture) on all individuals without prior selection on the basis of screening for symptoms detected an additional four cases per 100 individuals screened (p=0·05). Furthermore, an increment in country prevalence of tuberculosis of 100 cases of tuberculosis per 100 000 population was associated with an additional yield of one case per 100 individuals screened (p=0·03).
Restricting the analysis to studies in individuals infected with HIV routinely doing both sputum smears and cultures as part of the microbiological investigations (22 studies) showed similar results with regards to effect estimate for symptom screening (slope 3·6, 95% CI −0·9 to 8·0).
Estimates of national prevalence of tuberculosis are infrequently on the basis of prevalence survey data, they are instead derived from estimates of the incidence of tuberculosis. In view of this and that national estimates of incidence of tuberculosis are more readily available, we repeated the metaregression analysis with the inclusion of estimated incidence of tuberculosis rather than estimated prevalence. In this model, an increment in country incidence of tuberculosis of 100 cases of tuberculosis per 100 000 people was associated with an additional yield of 0·7 case per 100 individuals screened (p=0·04).
Only three studies reported treatment outcomes of individuals diagnosed with tuberculosis during intensified case finding. In a population-based study in South Africa, only 13 (56%) of 23 people actively detected with tuberculosis completed treatment.97
Among 24 women infected with HIV in India diagnosed with tuberculosis after giving birth, 21 started treatment for tuberculosis and 17 (70%) were either cured or still receiving treatment at time of analysis.62
In the Côte d'Ivoire, of 134 prisoners diagnosed with tuberculosis, 99 (74%) were cured, 32 (24%) died, and in three (2%) treatment failed.28
None of these studies had data for treatment outcome available for those diagnosed passively.
None of the studies included in this systematic review did costing or cost-effectiveness analyses.