Diagnostic accuracy of symptom screening to rule out tuberculosis among human immunodeficiency virus–infected individuals undergoing screening prior to isoniazid preventive therapy was evaluated. Symptom screening had poor sensitivity, especially among those on antiretroviral therapy. Without culture-based screening, the risk of inadvertently prescribing isoniazid monotherapy is high.
Abstract
Background. Current symptom screening algorithms for intensified tuberculosis case finding or prior to isoniazid preventive therapy (IPT) in patients infected with human immunodeficiency virus (HIV) were derived from antiretroviral-naive cohorts. There is a need to validate screening algorithms in patients on antiretroviral therapy (ART).
Methods. We performed cross-sectional evaluation of the diagnostic accuracy of symptom screening, including the World Health Organization (WHO) algorithm, to rule out tuberculosis in HIV-infected individuals pre-ART and on ART undergoing screening prior to IPT.
Results. A total of 1429 participants, 54% on ART, had symptom screening and a sputum culture result available. Culture-positive tuberculosis was diagnosed in 126 patients (8.8%, 95% confidence interval [CI], 7.4%–10.4%). The WHO symptom screen in the on-ART compared with the pre-ART group had a lower sensitivity (23.8% vs 47.6%), but higher specificity (94.4% vs 79.8%). The effect of ART was independent of CD4+ count in multivariable analyses. The posttest probability of tuberculosis following a negative WHO screen was 8.9% (95% CI, 7.4%–10.8%) and 4.4% (95% CI, 3.7%–5.2%) for the pre-ART and on-ART groups, respectively. Addition of body mass index to the WHO screen significantly improved discriminatory ability in both ART groups, which was further improved by adding CD4 count and ART duration.
Conclusions. The WHO symptom screen has poor sensitivity, especially among patients on ART, in a clinic where regular tuberculosis screening is practiced. Consequently, a significant proportion of individuals with tuberculosis would inadvertently be placed on isoniazid monotherapy despite high negative predictive values. Until more sensitive methods of ruling out tuberculosis are established, it would be prudent to do a sputum culture prior to IPT where this is feasible.
There is a high prevalence of previously undiagnosed tuberculosis in individuals infected with human immunodeficiency virus (HIV), especially among those about to start antiretroviral therapy (ART) [1–3]. Identifying tuberculosis before individuals present with symptoms reduces morbidity and mortality and also reduces the spread of tuberculosis [4]. Intensified case finding, along with isoniazid preventive therapy (IPT) and infection control, forms part of the World Health Organization's (WHO) “three I's” intervention strategy to reduce the morbidity and mortality from HIV-associated tuberculosis [5, 6]. WHO recommends screening for tuberculosis “at every visit to a health facility or contact with a health worker.” Ruling out tuberculosis by intensified case finding is particularly important prior to initiation of IPT in order to offset the risk of isoniazid resistance.
In an attempt to standardize tuberculosis screening for HIV-infected individuals living in resource-constrained settings, the WHO undertook a meta-analysis of individual patient data from studies conducted in high-burden countries to find the most sensitive screening strategy [1,7–16]. The presence of any one of current cough, fever, night sweats, or weight loss had a high negative predictive value and this symptom screening algorithm is now recommended for intensified case finding. It is not known how ART will modify the diagnostic performance of the WHO symptom screening algorithm, as none of the cohorts analyzed to develop the algorithm included patients on ART. This is important as IPT is recommended for individuals on ART in the 2010 WHO guidelines, mostly on the basis of observational data.
Countries most affected by HIV are rapidly scaling up ART, with the WHO guidelines recommending earlier ART initiation. Because most HIV-related resources are being used for ART clinics, pre-ART care is often rudimentary in high-burden countries. Implementing a complex intervention like IPT may be easier in ART clinics. Therefore, it is important to evaluate tuberculosis screening algorithms in individuals on ART.
To investigate the effect of ART on the diagnostic utility of symptom screening to rule out tuberculosis, we compared HIV-infected participants established on ART and those being prepared for ART who underwent intensified case finding for tuberculosis during evaluation for IPT.
METHODS
Study Setting and Participants
The study was conducted at the Ubuntu Clinic in Khayelitsha Site B, Cape Town, South Africa. Ubuntu clinic is an integrated HIV outpatient facility that provides ART. At the time of the study the criterion for starting ART was CD4 count <200 cells/μL or WHO clinical stage 4. Participants attending the clinic are seen every two weeks to monthly during ART preparation, if pre-ART, and then monthly to once every two months for collection of ART medication, if already on ART. Routine assessments at each visit include evaluation for tuberculosis symptoms by the attending clinician, nurse, or lay counselor. For this study, consecutive participants undergoing additional screening for an ongoing pragmatic randomized controlled trial to determine the effectiveness of the combination of IPT and ART to reduce the risk of active tuberculosis in HIV-infected persons (clinical trials registration NCT00463086) were invited to participate. The University of Cape Town research ethics committee approved the trial and its screening phase (REC 013/2007). Written informed consent was provided. The annual tuberculosis case notification rate in the area during the study was approximately 1500 per 100 000 population.
Clinical Assessments and Tuberculosis Case Definitions
Participants established on ART (on-ART group) and those being prepared for ART (pre-ART group) were screened for tuberculosis symptoms. Screening occurred between November 2007 and October 2009. The tuberculosis symptom screen at the baseline visit included cough (presence and if duration ≥2 weeks), night sweats, self-reported weight loss (or documented ≥1.5-kg weight loss in a month) [7] and fever [17]. A single sputum sample was obtained by spontaneous expectoration or, if this was unsuccessful, ultrasonic nebulization with hypertonic saline, for smear microscopy (fluorescent) and mycobacterial culture (Bactec MGIT system and Lowenstein-Jensen medium) at the reference laboratory, Groote Schuur Hospital. Culture results were interpreted without knowledge of clinical symptoms. During the study period the reference laboratory reported no cross-contamination: none of 500 dummy sputum samples were false-positive for Mycobacterium tuberculosis (M. Nicol, oral communication, September 2012). Participants who failed to produce sputum were either asked to return the following day (with a sample) or were asked for a sample at their next ART clinic visit where nebulization would be repeated if necessary. A period of up to 1 month for sputum collection was allowed. Unless participants were identified as tuberculosis suspects, further investigations, including a second sputum and chest radiography, were not routinely performed. Participants whose sputum was culture-positive for M. tuberculosis were treated for tuberculosis, regardless of symptoms.
Statistical Analyses
All analyses were performed with Stata/MP software, version 10 (StataCorp, College Station, Texas). Basic description of baseline data and cross-tabulations by ART and tuberculosis diagnosis at screening were done.
Diagnostic Accuracy of Symptoms for Prevalent Culture-Positive Tuberculosis
Within ART study groups, we computed standard test accuracy measures (sensitivity, specificity, and positive and negative predictive values) to detect prevalent tuberculosis using culture positivity as the reference standard. The discriminatory ability of individual tuberculosis symptoms was evaluated by the receiver operating characteristic area or area under the curve (AUC) of sensitivity vs 1 – specificity. Binomial exact confidence intervals were derived for all accuracy estimates.
Selection of Potential Predictors of Tuberculosis
Candidate clinical predictors were determined a priori based on clinical judgment and the published literature and they included age (dichotomized at its median); sex; body mass index (BMI) according to WHO categorization of ≤18.5 (underweight),18.6–25 (normal weight), >25 (overweight) [18]; most recent CD4 count (no older than 6 months and categorized as <200; 200–350; or >350); ART duration ≤3 months or >3 months (if already on ART); prior episode of treated tuberculosis. The early period on ART is associated with a higher mortality mostly owing to diagnosed and undiagnosed tuberculosis [19, 20]. We retrospectively applied the WHO symptom-screening algorithm (the presence of any one of active cough of any duration, fever, night sweats, or weight loss (either self-reported or documented weight loss). Although participants at our ART clinic were reevaluated using a symptom screen during subsequent clinic visits, only results recorded at the initial screening visit were considered for analyses.
Assessment of the Added Value of Clinical Predictors to the WHO Screen
Multivariable logistic regression analysis was performed to develop diagnostic models for culture positive tuberculosis by ART strata. A model with the WHO symptom screen was first derived and its AUC computed. Then the rest of the clinical observations were added and the AUC of the extended model computed. The overall ability of the new multivariable model to discriminate M. tuberculosis culture-positive and -negative participants was assessed by conducting AUC comparisons between the WHO screen only and the extended model. Variables were kept in the final multivariate model if the associated likelihood-ratio test showed a P value of < .20.
All analyses were performed on a dataset with complete data on predetermined predictors as well as culture results, since the objective was to compare the methods rather than to report results for a particular population.
RESULTS
Participants
A total of 1429 participants, 54% on ART, were screened with a tuberculosis symptom algorithm and had sputum culture results available. Numbers invited for screening and those in the final analysis are provided in Figure 1. Participant characteristics overall and by ART group are shown in Table 1. Culture-positive tuberculosis was diagnosed in 126 (8.8%; 95% confidence interval [CI], 7.4%–10.4%) of the 1429 participants, including 84 of 654 pre-ART (13%; 95% CI, 10%–15.7%), and 42 of 775 on ART (5.4%; 95% CI, 3.9%–7.3%).
Figure 1.
Study flow and numbers in analysis. Standards for reporting diagnostic accuracy studies were adhered to (STARD) [27]. Abbreviations: ART, antiretroviral therapy; TB, tuberculosis; WHO, World Health Organization.
Table 1.
Clinical Characteristics of Screened Cohort by Antiretroviral Therapy Status
| Characteristics | Total (N = 1429) | Pre-ART(n = 654) | On ART (n = 775) |
|---|---|---|---|
| Age ≥ 35 years | 45.8 (654) | 42.4 (277) | 48.7 (377) |
| Male | 26.3 (376) | 28.6 (187) | 24.4 (189) |
| Female | 73.7 (1053) | 71.4 (467) | 75.6 (586) |
| WHO BMI category | |||
| ≤18.5 (underweight) | 3 (41) | 4 (26) | 2 (15) |
| 18.6–25 (normal weight) | 47 (669) | 52 (339) | 42.6 (330) |
| >25 (overweight) | 50 (719) | 44 (289) | 55.5 (430) |
| Median CD4+ count, cells/μL (IQR) | 209 (145–331) | 165 (117–209) | 289 (194–435) |
| CD4+ count < 200 | 46.8 (669) | 71.1 (465) | 26.3 (204) |
| CD4+ count 200–350 | 29.9 (427) | 23.1 (151) | 35.6 (276) |
| CD4+ count >350 | 23.3 (333) | 5.8 (38) | 38.1 (295) |
| Prior tuberculosis | 38.6 (552) | 21.9 (143) | 52.8 (409) |
| Median months on ART (IQR) | … | … | 12 (4–28) |
| Early ART (duration <3 months) | … | … | 14.0 (135) |
| Viral load <400 copies/mL | … | … | 88.7 (673/759) |
All data are presented as % (No.) unless otherwise specified.
Abbreviations: ART, antiretroviral therapy; BMI, body mass index; IQR, interquartile range; WHO, World Health Organization.
Frequency of Tuberculosis Symptoms or Signs in Screened Cohort by ART Status
Individual symptoms of tuberculosis were more frequently reported by pre-ART than on-ART participants, as shown in Table 2. Including cough of any duration, as per the WHO symptom screen, rather than cough >2 weeks increased the proportion of participants positive for any tuberculosis symptom from 11.8% to 14.4%. Pre-ART participants more commonly had a positive WHO screen (23.7%) compared to those on ART (6.6%; P < .0001).
Table 2.
Frequency of Tuberculosis Symptoms or Signs in the Screened Cohort by Antiretroviral Therapy Status
| Tuberculosis Symptom/Sign | Total (N = 1429) | Pre-ART (n = 654) | On-ART (n = 775) |
|---|---|---|---|
| Current cough | 8.8 (125) | 13.5 (119) | 4.8 (37) |
| Cough ≥2 week duration | 5.7 (82) | 10.6 (69) | 1.7 (13) |
| Night sweats | 5.0 (71) | 10.1 (66) | 0.6 (5) |
| Fever | 0.6 (9) | 1.4 (9) | … |
| Weight lossa | 7.8 (112) | 14.8 (97) | 2 (15) |
| Any 1 tuberculosis symptomor sign positive (cough ≥2 weeks) | 11.8 (168) | 21.2 (139) | 3.7 (29) |
| WHO screen—any 1 tuberculosis symptom/sign positive (current cough) | 14.4 (206) | 23.7 (155) | 6.6 (51) |
All data are presented as % (No.).
Abbreviations: ART, antiretroviral therapy; WHO, World Health Organization.
aEither self-reported or documented.
Diagnostic Utility of Symptoms for Prevalent Culture-Positive Tuberculosis
The diagnostic utility of symptom screening for tuberculosis is shown in Table 3. Symptoms had poor sensitivity for culture-positive tuberculosis, even when using the WHO screen. The sensitivity of symptom screening was higher for the pre-ART than the on-ART group. Individual symptoms with the highest sensitivity for culture-positive tuberculosis among pre-ART participants were weight loss (32.1%, 95% CI, 22.4%–43.2%), cough of any duration (31%; 95% CI, 21.3%–42%), night sweats (23.8%; 95% CI, 15.2%–34.3%), and fever. The WHO screen had a sensitivity of 47.6% (95% CI, 36.6%–58.8%) and a specificity of 79.8% (95% CI, 76.3%–83%) for pre-ART participants, whereas for those on ART the sensitivity was 23.8% (95% CI, 12.1%–39.5%) and the specificity was 94.4% (95% CI, 92.5%–96%). Multivariable analyses confirmed the effect of ART on sensitivity to be statistically significant even after controlling for CD4+ count or BMI (not shown). Using a symptom screen including cough of any duration, as in the WHO screen, rather than cough ≥2 weeks, as used more traditionally for screening for tuberculosis, increased the sensitivity for culture-positive tuberculosis by 2.4% in pre-ART participants and by 7% in those on ART. Of total culture-confirmed cases, 76.2% and 52.4% in the on-ART and pre-ART groups, respectively, would inadvertently receive IPT (1 – sensitivity). The posttest probability of culture-positive disease following a negative WHO screen (1 – negative predictive value) was 8.9% (95% CI, 7.4%–10.8%) for pre-ART and 4.4% (95% CI, 3.7%–5.2%) for those on ART. The ability of the WHO screen to discriminate those with and without tuberculosis was 64% (AUC 95% CI, 58%–69%) for the pre-ART group and 59% (AUC 95% CI, 53%–66%) for the on-ART group.
Table 3.
Diagnostic Accuracy of Screening Tests Stratified by Antiretroviral Therapy Status
| Tuberculosis Symptoms and Signs in Total Cohort and by ART Status | Total With Symptom(s) | TB Cases With Symptom(s) | Sensitivity % (95% CI) | Specificity % (95% CI) | PPV % (95% CI) | NPV % (95% CI) | AUC (95% CI) |
|---|---|---|---|---|---|---|---|
| Total cohort TB prevalence = 8.8% (95% CI, 7.4%–10.4%) | |||||||
| Current cough | 125 | 31 | 24.6 (17.4–33.1) | 92.8 (91.2–94.1) | 24.8 (17.5–33.3) | 92.7 (91.2–94.1) | 59 (55–63) |
| Cough for ≥2 weeks | 82 | 26 | 20.6 (13.9–28.8) | 95.7 (94.5–96.7) | 31.7 (21.9–42.9) | 92.6 (91–93.9) | 58 (55–62) |
| Night sweats | 71 | 20 | 15.9 (10–23.4) | 96.1 (94.9–97.1) | 28.2 (18.1–40.1) | 92.2 (90.6–93.6) | 56 (53–59) |
| Fever | 9 | 5 | 4 (1.3–9) | 99.7 (99.2–99.9) | 55.6 (21.2–86.3) | 91.5 (89.9–92.9) | 52 (50–54) |
| Weight lossa | 112 | 32 | 25.4 (18.1–33.9) | 93.9 (92.4–95.1) | 28.6 (20.4–37.9) | 92.9 (91.3–94.2) | 60 (56–64) |
| Any 1 TB symptom or sign positive (chronic cough) | 168 | 45 | 35.7 (27.4–44.7) | 90.6 (88.8–92.1) | 26.8 (20.3–34.2) | 93.6 (92.1–94.9) | 63 (59–67) |
| WHO screen: any 1 TB symptom/sign positive (current cough) | 206 | 50 | 39.7 (31.1–48.8) | 88 (86.1–89.7) | 24.3 (18.6–30.7) | 93.8 (92.3–95.1) | 64 (60–68) |
| Not on ART at screening TB prevalence = 13% (95% CI, 10%–15.7%) | |||||||
| Current cough | 88 | 26 | 31 (21.3–42) | 89.1 (86.3–91.6) | 29.5 (20.3–40.2) | 89.8 (87–92.1) | 60 (55–65) |
| Cough for ≥2 weeks | 69 | 24 | 28.6 (19.2–39.5) | 92.1 (89.6–94.2) | 34.8 (23.7–47.2) | 89.7 (87–92.1) | 60 (55–65) |
| Night sweats | 66 | 20 | 23.8 (15.2–34.3) | 91.9 (89.4–94) | 30.3 (19.6–42.9) | 89.1 (86.3–91.5) | 58 (53–63) |
| Fever | 9 | 5 | 6 (2–13.3) | 99.3 (98.2–99.8) | 55.6 (21.2–86.3) | 87.8 (85–90.2) | 53 (50–55) |
| Weight lossa | 97 | 27 | 32.1 (22.4–43.2) | 87.7 (84.7–90.3) | 27.8 (19.2–37.9) | 89.8 (86.9–92.2) | 60 (55–65) |
| Any 1 TB symptom or sign positive (chronic cough) | 139 | 38 | 45.2 (34.3–56.5) | 82.3 (78.9–85.3) | 27.3 (20.1–35.5) | 91.1 (88.3–93.4) | 64 (58–69) |
| WHO screen: any 1 TB symptom/sign positive (current cough) | 155 | 40 | 47.6 (36.6–58.8) | 79.8 (76.3–83) | 25.8 (19.1–33.4) | 91.2 (88.3–93.5) | 64 (58–69) |
| On ART at screening TB prevalence = 5.4% (95% CI, 3.9%–7.3%) | |||||||
| Current cough | 37 | 5 | 12 (4–7.3) | 95.6 (93.9–97) | 13.5 (4.5–28.8) | 95 (93.2–96.4) | 54 (49–59) |
| Cough for ≥2 weeks | 13 | 2 | 5 (1–16.2) | 98.5 (97.3–99.2) | 15.4 (2–45.4) | 94.8 (92.9–96.2) | 52 (48–55) |
| Night sweats | 5 | 0 | … | 99.3 (98.4–99.8) | … | 94.5 (92.7–96) | |
| Fever | 0 | 0 | … | … | … | … | … |
| Weight lossa | 15 | 5 | 12 (4–25.6) | 98.6 (97.5–99.3) | 33.3 (11.8–61.6) | 95.1 (93.4–96.5) | 55 (50–60) |
| Any 1 TB symptom/sign positive (chronic cough) | 29 | 7 | 16.7 (7–31.4) | 97 (95.5–98.1) | 24.1 (10.3–43.5) | 95.3 (93.5–96.7) | 57 (51–63) |
| WHO screen: any 1 TB symptom/sign positive (current cough) | 51 | 10 | 23.8 (12.1–39.5) | 94.4 (92.5–96) | 19.6 (9.8–33.1) | 95.6 (93.8–97) | 59 (53–66) |
Abbreviations: ART, antiretroviral therapy; AUC, area under the curve of sensitivity versus 1 minus specificity; CI, confidence interval; NPV, negative predictive value; PPV, positive predictive value; TB, tuberculosis; WHO, World Health Organization.
aEither self-reported or documented.
Value of Additional Predictors to the WHO Screen
In multivariable analyses, BMI and CD4+ count were independent predictors of tuberculosis in both ART groups, and ART duration of <3 months was an additional predictor among those on ART (Table 4). Addition of BMI and CD4+ count to a predictive model with the WHO screen alone increased the AUC yield to 74% (95% CI, 69%–80%; Figure 2A). However, CD4+ count may not always be available, so we added BMI alone to the WHO screen model, which gave an AUC of 71% (95% CI, 66%–76%; Figure 2A). Among those on ART, adding BMI, CD4+ count, and duration of ART <3 months or more improved discrimination of the WHO screen model by 11%, AUC of 70% (95% CI, 60%–79%; Figure 2B). Addition of BMI alone to the WHO screen improved the AUC to 69% (95% CI, 61%–78%).
Table 4.
Predictors of Culture Positive Tuberculosis Stratified by Antiretroviral Therapy Status
| Characteristic | Pre-ART |
On ART |
||
|---|---|---|---|---|
| Unadjusted OR (95% CI) | Adjusted OR (95% CI) | Unadjusted OR (95% CI) | Adjusted OR (95% CI) | |
| Demographics | ||||
| Age <35 years | Ref | Ref | ||
| Age >35 years | 1.3 (.9–2.1) | 1.0 (.5–1.8) | ||
| Sex | ||||
| Female | Ref | Ref | ||
| Male | 1.5 (.9–2.4) | 1.4 (.7–2.8) | ||
| WHO BMI category | ||||
| ≤18.5 (underweight) | 4.4 (1.9–10.1) | 3.8 (1.7–8.8) | 2.1 (.4–9.7) | 2.3 (.5–11.1) |
| 18.6–25 (normal weight) | Ref | Ref | Ref | Ref |
| >25 (overweight) | 0.3 (.2–.6) | 0.3 (.2–.6) | 0.5 (.3–1.0) | 0.6 (.3–1.2) |
| Degree of immunosuppression | ||||
| CD4+ count <200 cells/μL | 6.9 (1.0–51) | 5.8 (.8–43.4) | 2.7 (1.2–5.7) | 2.0 (.9–4.5) |
| CD4+ count 200–350 cells/μL | 2.6 (.3–21.2) | 2.2 (.3–18.3) | 1.2 (.5–2.7) | 1.1 (.5–2.5) |
| CD4+ count >350 cells/μL | Ref | Ref | Ref | Ref |
| Duration on ART | ||||
| ART for >3 months | … | … | Ref | Ref |
| Early ART ≤3 months | … | … | 2.5 (1.3–4.9) | 1.9 (.9–4.0) |
| History of tuberculosis | ||||
| No prior tuberculosis | Ref | Ref | ||
| Prior treated tuberculosis | 0.9 (.5–1.6) | 0.7 (.4–1.3) | ||
Adjusted ORs are from the best multivariate clinical model.
Abbreviations: ART, antiretroviral therapy; BMI, body mass index; CI, confidence interval; OR, odds ratio; WHO, World Health Organization.
Figure 2.
Value of additional predictors to the World Health Organization (WHO) screen in pre–antiretroviral therapy (ART; A) and on-ART (B) participants. A, Pre-ART group: WHO screen only area under the curve (AUC) = 64% (95% confidence interval [CI], 58%–69%); WHO screen and body mass index (BMI) = 71% (66%–76%); WHO screen plus BMI and CD4+ count AUC = 74% (95% CI, 69%–80%). B, On-ART group: WHO screen only AUC = 59% (95% CI, 53%–66%); WHO screen plus BMI and ART duration AUC = 69% (95% CI, 61%–78%). WHO screen plus BMI, CD4+ count, and ART duration AUC = 70% (95% CI, 60%–79%).
DISCUSSION
In this large cohort of pre-ART and on-ART HIV-infected individuals, the WHO symptom screening algorithm had a negative predictive value of >90% but lower sensitivity than reported for clinics in high-tuberculosis-prevalence settings [1]. A negative predictive value >90% does not mean symptom screening is a useful way to rule out tuberculosis, as illustrated by the fact that most participants with culture-confirmed tuberculosis in our cohort had none of the 4 tuberculosis symptoms. Posttest probabilities of disease therefore remained high. A novel feature of our study was the assessment of the performance of symptom screening by ART status, with lower sensitivity (23.8% vs 47.6%), but higher specificity (94.4% vs 79.8%) in the on-ART compared with the pre-ART group. The sensitivity of symptom screening at the initial baseline visit was most reduced among those on ART when the traditional cough ≥2 weeks duration was used rather than cough of any duration as used in the WHO algorithm. Addition of BMI to the WHO symptom screen significantly improved discriminatory ability in both ART groups, which was marginally further improved by adding CD4 count. Consistent with some studies from sub-Saharan Africa, the prevalence of previously undetected culture-confirmed tuberculosis was high at 5.4% and 13% in the on-ART and pre-ART groups, respectively [2, 3, 21–24].
In the recent WHO-led meta-analysis, symptom screening had higher sensitivity (90.1% in clinical settings [1]) than we found. A likely explanation for the lower sensitivity of symptom screening in our cohort is that the clinic where our study was conducted practices active case finding, therefore prior screening for tuberculosis symptoms would likely have been carried out on patients evaluated for the study. Prior screening has a marked effect on the sensitivity of the WHO algorithm. In the WHO meta-analysis, patients from studies where prior screening was not conducted were associated with >10-fold increase in the odds of a truly positive WHO screen result compared with patients from studies where prior screening had been conducted [8–10].
Our study had some limitations. First, as discussed above, prior screening of our participants may have affected the performance of the WHO symptom screen. Our results will be generalizable to clinic settings where regular active case finding is implemented, as recommended by WHO, but not to other settings where prior tuberculosis screening has not been done. Second, the new suggested WHO symptom screen was applied retrospectively; as a result, we included weight loss as either self-reported or documented. This limits generalizability of our results to settings that consider self-reported or documented weight loss when actively screening patients for disease. Third, we only performed one sputum culture at baseline and chest radiography was not part of screening. Confirmation with a second culture is desirable, especially in asymptomatic participants in whom higher false-positive results are to be expected. However, false-positive cultures were unlikely in our study as the laboratory had no cross-contamination in a survey carried out during the study period. The evidence for the additional contribution of chest radiography in evaluating individuals prior to IPT remains limited [1]. Last, we conducted a complete case analysis excluding 661 participants with missing data (Figure 1). About half were excluded for missing sputum culture results. This might have underestimated the positive predictive value of symptom screening, but the discriminatory ability of the multivariable model will unlikely be affected by missing predictors as the proportions missing were small. Those with missing data were not different from those in the final analysis with respect to proportions on ART or other predictors of tuberculosis.
Our finding that symptom screening had poor sensitivity, especially for those on ART, in an HIV clinic practicing regular tuberculosis symptom screening has important public health implications. Based on a few studies of individuals who were not on ART, asymptomatic patients with M. tuberculosis on sputum culture may progress to symptomatic disease [10, 11, 21, 25]. The degree to which ART modifies progression of such asymptomatic cases is currently unclear. However, frequent screening may not be performed because patients often have long intervals between clinic visits (eg, those stable on ART and those not yet qualifying for ART), or because clinics may omit screening as they are often overburdened. Furthermore, using the WHO symptom screen prior to giving IPT in clinics practicing regular screening would inadvertently place a significant proportion of patients with tuberculosis on isoniazid monotherapy, which may result in isoniazid resistance. Of total culture-confirmed cases in the on-ART and pre-ART groups, 76.2% and 52.4%, respectively, would inadvertently receive IPT and there was a 9.1% and 4.4% posttest probability of culture-positive tuberculosis among the WHO screen–negative individuals in the pre-ART and on-ART groups, respectively. The WHO recently recommended IPT implementation for patients on ART in high-burden countries as it is perhaps easier to add isoniazid to patients on other therapies who are being regularly followed up than to implement IPT in patients not yet qualifying for ART. Yet patients on ART appear to be at particularly high risk of being asymptomatic with positive sputum cultures. Consideration of BMI and/or CD4 count would improve the ability of the WHO symptoms screen to discriminate those with and without tuberculosis. Further studies are necessary to develop more sensitive methods to rule out tuberculosis, especially for patients on ART prior to giving IPT. The role of new diagnostics such as the urine lipoarabinomannan lateral flow and Xpert® MTB/RIF assays in evaluating patients ahead of IPT also needs to be explored [26].
In conclusion, our study demonstrates that the WHO symptom screen has poor sensitivity, especially in the on-ART group, in a clinic where regular tuberculosis screening is practiced. Discriminatory ability of symptom screening was improved by adding CD4 count and/or BMI. Until more sensitive methods of ruling out tuberculosis are established, it would be prudent to do a sputum culture prior to IPT when this is feasible.
Notes
Acknowledgments. We thank individuals and staff at the Ubuntu Clinic in Khayelitsha as well as the ART-IPT study core team. We also thank the Provincial Government of the Western Cape and City of Health, particularly Drs Virginia de Azevedo and Gio Perez, for integrating the ART-IPT study into clinic data and patient systems.
Author contributions. M. X. R., R. J. W., J. R. G., and G. M. designed this evaluation. All authors interpreted the data. M. X. R. analyzed the data and wrote the first draft; all authors helped to revise the paper. The authors had full access to all study data and were solely responsible for the decision to submit for publication.
Financial support. This work was supported in part by the Department of Health of South Africa, the Wellcome Trust, and Médecins sans Frontières, which provided additional staff to Ubuntu Clinic. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. M. X. R. and R. J. W. were supported by the Wellcome Trust (084323, 084670, 088316) and the European and Developing Countries Clinical Trials Partnership (EDCTP). R. J. W. was additionally supported by the European Union (SANTE/2005/105-061-102), the EDCTP (IP.07.32080.002), and the Medical Research Council of the United Kingdom (U.1175.02.002.00014.01).
Potential conflicts of interest. All authors: No reported conflicts.
All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.
References
- 1.Getahun H, Kittikraisak W, Heilig CM, et al. Development of a standardized screening rule for tuberculosis in people living with HIV in resource-constrained settings: individual participant data meta-analysis of observational studies. PLoS Med. 2011;8:e1000391. doi: 10.1371/journal.pmed.1000391. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Bassett IV, Wang B, Chetty S, et al. Intensive tuberculosis screening for HIV-infected patients starting antiretroviral therapy in Durban, South Africa. Clin Infect Dis. 2011;51:823–9. doi: 10.1086/656282. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Mtei L, Matee M, Herfort O, et al. High rates of clinical and subclinical tuberculosis among HIV-infected ambulatory subjects in Tanzania. Clin Infect Dis. 2005;40:1500–7. doi: 10.1086/429825. [DOI] [PubMed] [Google Scholar]
- 4.Dowdy DW, Chaisson RE, Maartens G, Corbett EL, Dorman SE. Impact of enhanced tuberculosis diagnosis in South Africa: a mathematical model of expanded culture and drug susceptibility testing. Proc Natl Acad Sci U S A. 2008;105:11293–8. doi: 10.1073/pnas.0800965105. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Stop TB Partnership. Geneva, Switzerland: World Health Organization; 2011. Guidelines for intensified tuberculosis case-finding and isoniazid preventive therapy for people living with HIV in resource-constrained settings. Available at: http://whqlibdoc.who.int/publications/2011/9789241500708_eng.pdf. Accessed 7 December 2011. [Google Scholar]
- 6.Granich R, Akolo C, Gunneberg C, et al. Prevention of tuberculosis in people living with HIV. Clin Infect Dis. 2011;50(suppl 3):S215–22. doi: 10.1086/651494. [DOI] [PubMed] [Google Scholar]
- 7.Mohammed A, Ehrlich R, Wood R, Cilliers F, Maartens G. Screening for tuberculosis in adults with advanced HIV infection prior to preventive therapy. Int J Tuberc Lung Dis. 2004;8:792–5. [PubMed] [Google Scholar]
- 8.Lewis JJ, Charalambous S, Day JH, et al. HIV infection does not affect active case finding of tuberculosis in South African gold miners. Am J Respir Crit Care Med. 2009;180:1271–8. doi: 10.1164/rccm.200806-846OC. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Corbett EL, Zezai A, Cheung YB, et al. Provider-initiated symptom screening for tuberculosis in Zimbabwe: diagnostic value and the effect of HIV status. Bull World Health Organ. 2010;88:13–21. doi: 10.2471/BLT.08.055467. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Corbett EL, Bandason T, Cheung YB, et al. Epidemiology of tuberculosis in a high HIV prevalence population provided with enhanced diagnosis of symptomatic disease. PLoS Med. 2007;4:e22. doi: 10.1371/journal.pmed.0040022. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Ayles H, Schaap A, Nota A, et al. Prevalence of tuberculosis, HIV and respiratory symptoms in two Zambian communities: implications for tuberculosis control in the era of HIV. PLoS One. 2009;4:e5602. doi: 10.1371/journal.pone.0005602. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Cain KP, McCarthy KD, Heilig CM, et al. An algorithm for tuberculosis screening and diagnosis in people with HIV. N Engl J Med. 2010;362:707–16. doi: 10.1056/NEJMoa0907488. [DOI] [PubMed] [Google Scholar]
- 13.Day JH, Charalambous S, Fielding KL, et al. Screening for tuberculosis prior to isoniazid preventive therapy among HIV-infected gold miners in South Africa. Int J Tuberc Lung Dis. 2006;10:523–9. [PubMed] [Google Scholar]
- 14.Shah S, Demissie M, Lambert L, et al. Intensified tuberculosis case finding among HIV-infected persons from a voluntary counseling and testing center in Addis Ababa, Ethiopia. J Acquir Immune Defic Syndr. 2009;50:537–45. doi: 10.1097/QAI.0b013e318196761c. [DOI] [PubMed] [Google Scholar]
- 15.Kimerling ME, Schuchter J, Chanthol E, et al. Prevalence of pulmonary tuberculosis among HIV-infected persons in a home care program in Phnom Penh, Cambodia. Int J Tuberc Lung Dis. 2002;6:988–94. [PubMed] [Google Scholar]
- 16.Lawn SD, Edwards DJ, Kranzer K, et al. Urine lipoarabinomannan assay for tuberculosis screening before antiretroviral therapy diagnostic yield and association with immune reconstitution disease. AIDS. 2009;23:1875–80. doi: 10.1097/qad.0b013e32832e05c8. [DOI] [PubMed] [Google Scholar]
- 17.South African Department of Health. Pretoria: South African Department of Health; 2004. 2004 national antiretroviral treatment guidelines. Available at: http://southafrica.usembassy.gov/media/2004-dohart-guidelines.pdf . Accessed 12 August 2012. [Google Scholar]
- 18.World Health Organization. Geneva, Switzerland: WHO; 1995. Report of a WHO Expert Committee on Physical Status: the use and interpretation of anthropometry. Available at: http://whqlibdoc.who.int/trs/WHO_TRS_854.pdf . Accessed 13 September 2011. [PubMed] [Google Scholar]
- 19.Lawn SD, Harries AD, Anglaret X, Myer L, Wood R. Early mortality among adults accessing antiretroviral treatment programmes in sub-Saharan Africa. AIDS. 2008;22:1897–908. doi: 10.1097/QAD.0b013e32830007cd. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Lawn SD, Myer L, Harling G, et al. Determinants of mortality and nondeath losses from an antiretroviral treatment service in South Africa: implications for program evaluation. Clin Infect Dis. 2006;43:770–6. doi: 10.1086/507095. [DOI] [PubMed] [Google Scholar]
- 21.Lawn SD, Brooks SV, Kranzer K, et al. Screening for HIV-associated tuberculosis and rifampicin resistance before antiretroviral therapy using the Xpert MTB/RIF assay: a prospective study. PLoS Med. 2011;8:e1001067. doi: 10.1371/journal.pmed.1001067. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Ngowi BJ, Mfinanga SG, Bruun JN, Morkve O. Pulmonary tuberculosis among people living with HIV/AIDS attending care and treatment in rural northern Tanzania. BMC Public Health. 2008;8:341. doi: 10.1186/1471-2458-8-341. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Kufa T, Mngomezulu V, Charalambous S, et al. Undiagnosed tuberculosis among HIV clinic attendees: association with antiretroviral therapy and implications for intensified case finding, isoniazid preventive therapy, and infection control. J Acquir Immune Defic Syndr. 2012;60:e22–8. doi: 10.1097/QAI.0b013e318251ae0b. [DOI] [PubMed] [Google Scholar]
- 24.Moore D, Liechty C, Ekwaru P, et al. Prevalence, incidence and mortality associated with tuberculosis in HIV-infected patients initiating antiretroviral therapy in rural Uganda. AIDS. 2007;21:713–9. doi: 10.1097/QAD.0b013e328013f632. [DOI] [PubMed] [Google Scholar]
- 25.Oni T, Burke R, Tsekela R, et al. High prevalence of subclinical tuberculosis in HIV-1-infected persons without advanced immunodeficiency: implications for TB screening. Thorax. 2011;66:669–73. doi: 10.1136/thx.2011.160168. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Lawn SD. Point-of-care detection of lipoarabinomannan (LAM) in urine for diagnosis of HIV-associated tuberculosis: a state of the art review. BMC Infect Dis. 2012;12:103. doi: 10.1186/1471-2334-12-103. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Bossuyt PM, Reitsma JB, Bruns DE, et al. Towards complete and accurate reporting of studies of diagnostic accuracy: the STARD initiative. BMJ. 2003;326:41–4. doi: 10.1136/bmj.326.7379.41. [DOI] [PMC free article] [PubMed] [Google Scholar]