Symptom Screening Among HIV-Infected Pregnant Women Is Acceptable and Has High Negative Predictive Value for Active Tuberculosis

Abstract

We evaluated tuberculosis (TB) screening among 799 human immunodeficiency virus (HIV)–infected pregnant women in India. Eleven (1.4%) had active TB. The negative predictive value of screening using cough, fever, night sweats, or weight loss was 99.3%. Tuberculin skin test and targeted chest radiography provided no substantial benefit. TB symptom screening, as recommended by the World Health Organization, is effective for ruling out TB in HIV-infected pregnant women.

Tuberculosis (TB) is a serious global health problem, with one-third of the world’s population infected and ∼9.4 million new active cases occurring each year, an estimated 3.3 million in women [1]. TB in pregnant and postpartum women with human immunodeficiency virus (HIV) infection is particularly devastating and associated with increased maternal and infant mortality as well as HIV perinatal transmission [2–4]. According to new 2010 World Health Organization (WHO) recommendations for TB screening in HIV-infected adults, active TB must be ruled out using specific symptoms (eg, current cough of any duration, fever, weight loss, or night sweats) before isoniazid preventive therapy is initiated [5]. These recommendations were based on studies with few, if any, pregnant women. The objective of our study was to assess the prevalence of TB disease and the acceptance of screening and determine the sensitivity, specificity, negative predictive value, and likelihood ratios of various combinations of TB screening strategies, particularly the additive role of targeted chest radiography, performed around the time of delivery among urban Indian HIV-infected women seeking hospital-based delivery.

METHODS

Study Population

We retrospectively analyzed peripartum TB screening data collected from HIV-infected pregnant women who were enrolled and prospectively followed up until 1 year postpartum between 2002 and 2007 as part of a National Institutes of Health (NIH)–funded phase III prevention of mother-to-child transmission (PMTCT) trial (SWEN trial; Clinical Trials.gov NCT00061321). The eligibility criteria and parent study methods are described elsewhere [6]. We included all 799 women from the trial irrespective of whether their infants were randomized, as well as 41 HIV-infected women in an ancillary cohort who were ineligible to enroll in SWEN but were willing to be prospectively followed up with the same visit schedule and procedures. The study site was an urban 1300-bed public hospital of Byramjee Jeejeebhoy Medical College in Pune, Maharashtra. The study was approved by the Johns Hopkins University and local institutional review boards and ethics committee.

Tuberculosis Screening

HIV-infected women were screened for active TB between 1 week before and 2 weeks after delivery, with 94% screened within 7 days of delivery. TB screening procedures used a standardized symptom assessment that included variables that are now in the 2010 WHO recommended TB symptom screen (current cough of any duration, fever, weight loss, night sweats). We also used expanded criteria, which included additional symptoms (eg, hemoptysis, fatigue) and a targeted physical examination for lymphadenopathy and hepatosplenomegaly. All women received a tuberculin skin test (TST), and TST positivity was defined as ≥5 mm of induration [1]. A shielded chest radiograph was offered if results of the TB symptom screen or TST were positive. We defined cavitary lesions, upper lobe consolidation, hilar lymphadenopathy, miliary shadows, and pleural effusion as radiographic findings that were compatible with a diagnosis of TB. Sputum samples were requested if active TB was suspected on the basis of symptoms, TB-compatible chest radiograph, a positive TST result, and/or physician assessment in the absence of chest radiographic findings (n = 130). Acid-fast bacilli (AFB) testing was performed using Ziehl–Neelsen staining, Lowenstein-Jensen medium. Women with a diagnosis of TB were referred to the local Directly Observed Treatment Short Course treatment center.

Analysis

We defined TB cases as confirmed, probable, or possible TB based on WHO definitions [7]. Hence, all patients with positive TB culture, positive findings at microscopy (AFB or histology), or clinical or radiologic evidence of TB with response to anti-TB therapy were considered TB cases. To account for the fact that we did not evaluate sputum samples or cultures for AFB in all women to look for subclinical TB, we defined TB cases diagnosed within 10 weeks after delivery as prevalent peripartum TB. CD4 cell count and HIV quantitative RNA levels were determined at delivery. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the screening strategy to rule out peripartum active TB were calculated using standard formulas. We specifically evaluated the performance of the combination of symptoms now advocated in the new WHO TB symptom screen. We also calculated the positive and negative likelihood ratios, which indicate the odds of having the disease when a test was positive and of not having the disease when a test was negative, respectively.

RESULTS

Of 799 women screened, the median age was 23 years, and most women were married (96.4%) and were in early stages of HIV disease (WHO stage I/II, 95.0%) with a median CD4 count of 460 cells/mm³ (interquartile range, 316–658) and median log viral load of 3.8 copies/mL (interquartile range, 2.8–4.5) at delivery. Seventy-four (9.3%) had ≥1 of the 4 symptoms (current cough, fever, weight loss, night sweats) included in the new WHO symptom screen. The most common symptoms reported were cough of any duration (6.1%) and fever (4.3%) (Table 1).

Table 1.

Tuberculosis Screening Criteria in Peripartum Women With Human Immunodeficiency Virus Infection

Criteriona	Total population, no. (%) (n = 799)b	Women with TB, no. (%) (n = 11)
Clinical symptoms
Any cough	49 (6.1)	5 (45.5)
Cough for >2 weeks	7 (0.9)	2 (18.2)
Fever	34 (4.3)	5 (45.5)
Weight loss	5 (0.6)	0 (0.0)
Night sweats	0 (0.0)	0 (0.0)
Hemoptysis	1 (0.1)	0 (0.0)
Clinical signs or staging
Lymphadenopathy	12 (1.5)	2 (18.2)
Hepatosplenomegaly	8 (1.0)	2 (18.2)
WHO clinical stage 3 or 4	40 (5.0)	10 (91)
TST positive (n = 778)	164 (21.1)	6 (55)
Abnormal chest radiograph (n = 188)
Any abnormality	38 (20.3)	3 (33.3)c
Abnormality compatible with TB	17 (9.1)d	2 (22.2)c

^aTB, tuberculosis; TST, tuberculin skin test; WHO, World Health Organization.

^bThe total sample included 799 women except where otherwise indicated (for TST and chest radiography).

^cOf 11 women with diagnoses of TB, 9 underwent chest radiography, and 3 had chest radiographic abnormalities; 2 of the 3 had a TB-compatible radiograph (evidence of hilar lymphadenopathy with consolidation in 1 and miliary pattern in 1), and 1 had evidence of cardiomegaly with perihilar congestion, which was not compatible with TB.

^dFifteen women did not have diagnoses of TB but had chest radiographic findings compatible with TB, including cavitary lesions in 3, consolidation or nodules with hilar lymphadenopathy in 6, military pattern in 1, 1-sided pleural effusion in 1, upper lobe consolidation in 2, and hilar lymphadenopathy in 2.

Of 799 women, 21 (2.6%) did not have a TST performed or read within 72 hours. Of 778 women with a TST, 164 (21.1%) were TST positive; TST positivity did not vary significantly by CD4 cell count (P = .95). A total of 230 women were eligible for chest radiography (on account of a positive TST [68.3 %], positive symptom screen [28.7 %], or both [3.0 %]). Of these, 187 (81.3%) underwent chest radiography. Thirty-eight (20.2 %) radiographs were abnormal, and 17 (9.0%) had an abnormality compatible with TB (Table 1). Of 130 women referred for sputum evaluation, 107 (82.3%) provided sputum samples for AFB smears and culture (TB-compatible chest radiograph, n = 17; no available chest radiograph but TST or symptom screen positive, n = 43; negative chest radiograph but symptoms or physical examination findings present, n = 47).

After all evaluations were completed, 11 cases (1.4%) of active TB were identified (Figure 1); 5 were smear negative but confirmed at culture, 4 were smear positive for AFB, and 2 met the WHO definition of clinical or radiologic TB with response to treatment. Of these 11 cases, 5 (45.4%) were positive only at WHO symptom screen (1 woman reported cough, and 4 reported both cough and fever), 5 (45.4%) had a positive TST result but negative TB symptom screen (2 had evidence of extrapulmonary TB, 1 had an abnormal chest radiograph, and 2 were asymptomatic but with cultured-confirmed diagnoses), and 1 (9.1%) was positive at both WHO symptom screen and TST. A TB-compatible chest radiograph was identified in 2 of 9 women who underwent chest radiography; both reported a positive symptom screen.

Figure 1.

Flow chart showing number of women screened and results. Women were eligible for chest radiography (CXR) if they had either positive tuberculin skin test (TST) results or a positive symptom screen (n = 230). Women were referred for sputum evaluation for acid-fast bacilli (AFB) if active tuberculosis (TB) was suspected based on symptoms, TB-compatible chest radiograph, a positive TST results and/or physician assessment in the absence of chest radiographic findings (n = 130). All women were followed up for development of TB disease and defined as TB case patients if TB was diagnosed within 10 weeks after delivery. Active TB was defined as confirmed (culture), probable (positive at AFB-smear), or clinical or radiologic (with response to TB treatment), according to World Health Organization definitions [7]. Abbreviation: HIV, human immunodeficiency virus.

Specificity (90.9%), NPV (99.3%), and negative likelihood ratio (4.9) were excellent when the WHO recommended symptom screening was used alone. Although the sensitivity was greatest (100%) when WHO symptom screening was used with the TST, the NPV was only marginally higher, and specificity, not surprisingly, decreased to 71% (Table 2). A subset analysis on patients with more advanced HIV disease (CD4 ≤350 cells/mm³ or WHO clinical categories 3 and 4) provided largely comparable results for the total population, with marginally lower specificity and NPV.

Table 2.

Performance Characteristics of Tuberculosis Screening Combinations for Total Population and Patients With Advanced Human Immunodeficiency Virus Disease

	Total population, %; Patients with advanced HIV disease,a %
Screening criteria	Sensitivity	Specificity	PPV	NPV	Positive LR	Negative LR
WHO symptomsb alone	54.5; 54.5	90.9; 88.8	7.7; 17.6	99.3; 97.8	6.0; 4.9	4.9; 0.5
WHO symptoms or expanded criteriac	63.6; 63.6	89.7; 87.6	8.0; 18.4	99.4; 98.2	6.2; 5.1	0.4; 0.4
WHO symptoms or TST positivityd,e	100; 100	71.0; 70.0	4.7; 12.9	100; 100	3.4; 3.3	0.0; 0.0
WHO symptoms or abnormal chest radiograph	55.6; 55.6	73.0; 67.5	9.4; 16.1	97.0; 93.1	2.1; 1.7	0.6; 0.7
WHO symptoms or TB-compatible chest radiographf	50.0; 50.0	64.6; 67.4	5.8; 13.9	96.7; 92.8	1.4; 1.5	0.8; 0.7
WHO symptoms, expanded criteria, or abnormal chest radiograph	66.7; 66.7	69.7; 63.8	10.0; 17.1	97.6; 94.4	2.2; 1.8	0.5; 0.5
WHO symptoms, expanded criteria, or TB-compatible chest radiograph	55.6; 55.6	79.8; 76.3	12.2; 20.8	97.3; 93.8	2.7; 2.3	0.6; 0.6

Abbreviations: HIV, human immunodeficiency virus; LR, likelihood ratio; NPV, negative predictive value; PPV, positive predictive value; TB, tuberculosis; TST, tuberculin skin test; WHO, World Health Organization

^aAdvanced HIV disease was defined as CD4 counts ≤350 cells/mm³ or WHO clinical category 3 or 4 disease.

^bWHO symptoms included cough of any duration, fever, weight loss, or night sweats.

^cExpanded criteria include ≥1 of the following: fatigue, hemoptysis, lymphadenopathy, or hepatosplenomegaly.

^dTST positivity was defined as induration ≥5 mm in diameter.

^eSensitivity was 100% because TST was correlated with extrapulmonary TB, abnormal chest radiograph, or “subclinical” TB (absence of signs or symptoms of TB but positive culture).

^fTB-compatible chest radiographic findings included cavitary lesions, upper lobe consolidation, hilar lymphadenopathy, miliary shadows, and pleural effusion.

DISCUSSION

Our study has several important findings. First, among peripartum HIV-infected women seeking public hospital–based delivery in an urban Indian setting, we found a 1.4% prevalence of active TB among all screened and a 3.4% prevalence among those with positive TST results. This is lower than the 2% prevalence Kali et al [8,9] identified among all screened and the 11% prevalence Nachega et al [8,9] identified among TST-positive women in South African HIV PMTCT programs. This may in part be because pregnant women in India tend to have less advanced HIV infection and are, on average, younger compared with HIV-infected pregnant women in Sub-Saharan Africa [6]. This also explains the low PPV values for all of our screening combinations, similar to those in most TB screening studies of HIV-infected adult populations that have a TB prevalence of ≤5% [10].

Second, our results support the rationale for the new revised WHO TB symptom screen (cough of any duration, fever, weight loss, or night sweats) to rule out TB in HIV-infected adults [5] and confirms its value in HIV-infected peripartum women residing in a highly TB endemic region. Furthermore, regardless of HIV disease stage, we did not find that the presence of ≥1 other non-WHO screening clinical symptoms or signs added much value to the ruling out of TB.

Third, we addressed the role of TST and targeted chest radiography in peripartum women who had either a positive symptom screen or positive TST. The addition of TST and targeted chest radiography were of marginal to no added value and must be weighed against the challenges of implementing these tests effectively in busy resource-constrained programs. A decision analysis assessing addition of chest radiography to symptom screen did not find this to be of any additional value in Botswana [11]. A meta-analysis of 12 TB screening studies, however, did find a small incremental value for chest radiography, which increased sensitivity by ∼11% but increased NPV only marginally, by 1% [10].

Our study did have some limitations. We did not perform chest radiography or culture for AFB in all screened women but rather implemented a targeted approach; therefore, our ability to detect active TB could have been underestimated [12]. However, we followed up all women after screening to monitor for TB disease and therefore were able to assess the performance of symptom screening at delivery. The generalizability of our results may be limited, because our study took place in a large urban Indian public hospital where the antenatal prevalence of HIV infection was 3%. Nonetheless, our findings are consistent with a large meta-analysis of TB screening studies that included a variety of HIV-infected adult African and Asian nonpregnant populations [10] and, importantly, they reflect the realities of TB screening in busy antenatal programs in public hospitals.

Given the adverse outcomes associated with active TB in pregnant and postpartum women, the time has come to rapidly scale up TB screening and integrate TB prevention strategies in HIV PMTCT programs. Our study confirms the importance of using the simple, new WHO symptom screen to rule out TB in the setting of HIV PMTCT programs. Ruling out active TB is necessary prior to the initiation of isoniazid preventive therapy, a TB prevention strategy that has been shown to be highly efficacious in numerous trials of HIV-infected persons, particularly those who are TST positive [13].