Active Sputum Monitoring Detects Substantial Rate of Multi-Drug Resistant Tuberculosis (MDR-TB) in an HIV-Infected Population in South Africa

Abstract

Background

Tuberculosis (TB) co-infection with HIV is a substantial problem in South Africa. There has been a presumption that drug resistant strains of TB are common in South Africa, but few studies have documented this impression.

Methods

In Phidisa, a joint observational and randomized HIV treatment study for South African National Defence Force members and dependents, an initiative obtained microbiologic TB testing in subjects who appeared to be at high risk. We report results for HIV-infected subjects.

Results

TB was identified by culture in 116/584 (19.9%) of patients selected for sputum examination on the basis of suggestive symptoms. Smear was an insensitive technique for confirming the diagnosis: only 33% of culture-positive patients were identified by smear, with a 0.2% false positive rate. Of the 107 culture-positive individuals with susceptibility testing, 22 (20.6%) were identified to be MDR and 4 (3.7%) became extremely drug resistant tuberculosis (XDR) while under observation. Culture-positive cases with a history of TB treatment had more than twice the rate of MDR than those without, 27.1% vs. 11.9% (p=0.05).

Conclusions

TB is common in this cohort of HIV-infected patients. Smear was not a sensitive technique for identifying culture-positive cases in this health system. Drug susceptibility testing is essential to proper patient management because MDR was present in 20.6% of culture-positive patients. Better management strategies are needed to reduce the development of MDR-TB since so many such patients had received prior antituberculous therapy that was presumably not curative.

Background

Tuberculosis (TB) is well recognized to be a common disease process in many parts of the world [1–3]. Clinicians often make empiric diagnoses without laboratory confirmation in countries with high prevalence rates because clinicians are highly familiar with the disease, they have practiced for many years without extensive laboratory resources, and investment in technology to perform smears, cultures, conventional susceptibility testing or molecular diagnostics may not be feasible.

In South Africa there is considerable discussion about multiple drug resistant TB (MDR) and extensively drug resistant TB (XDR), but few reports document specific resistance rates [2, 4–13]. Current national guidelines in South Africa do not recommend routine tuberculosis drug susceptibility testing for new cases of TB [13–16].

This study reports how frequently TB isolates were resistant to standard antituberculous drugs in a South African cohort of HIV-infected patients in whom TB was frequently treated empirically.

Methods

Project Phidisa is a joint observational and randomized HIV treatment study for members of the South African National Defence Force (SANDF) and their dependents. The study was a collaboration between the National Institute of Allergy and Infectious Diseases (NIAID) and SANDF between January 2004 and March 2008. Participants in the observational study were screened for HIV infection and monitored longitudinally for disease progression. Between January 2004-December 2007, all patients in this protocol identified as HIV-positive were invited to enroll in a treatment study if they were treatment naïve with CD4+ T-cell counts <200 cells/uL or they had an AIDS defining illness. The treatment study randomized patients to one of four active therapy arms at one of 3 urban and 3 rural military bases in South Africa [17, 18]. This aspect of Project Phidisa assessed the impact of four different antiretroviral regimens for treatment of drug-naïve HIV-infected individuals. Currently, an observational-only study continues to collect data.

Sputum Initiative

An initiative was undertaken during the fourth year of Phidisa to obtain sputum from subjects, in either the observational or randomized component of the study, who appeared to be at especially high risk for TB based on self report or clinically observed weight loss, chronic cough, or fever. Staff were trained to identify patients with cough for more than 2 weeks, substantial weight loss, or unexplained fever. Such patients were encouraged to produce at least one expectorated or induced sputum sample. Specimens were obtained using equipment, environmental control technology, and personnel that were available at each site for clinical care, and varied by locality. Patients who were unable to produce expectorated sputum were induced with hypertonic saline if appropriate facilities were available.

Samples were transported to a commercial reference laboratory (BARC, Lancet Laboratories, http://www.lancet.co.za/) in Johannesburg within 24 hours of being produced and evaluated by direct microscopy (both Ziehl-Nielsen (ZN) and Auromine O (AO) staining), conventional culture (BACTEC, Becton Dickinson), and PCR (Haines and Light Cycler 1.5). Certain studies could not be performed on some days when equipment was not functioning. Results of all tests were reported promptly to the ordering health care provider following standard clinical practice.

A patient was designated to be a true positive for TB if an organism grew from at least one specimen that was identified by biochemical testing as M. tuberculosis. Susceptibility testing was reported separately for PCR and culture techniques.

Assessment for Presence of M. tuberculosis

Sputum was concentrated by centrifugation at 3,000 rpm for 15 minutes. Duplicate smears were prepared for direct microscopy. For light microscopy, one slide was stained by Ziehl-Nielsen (ZN) technique. For fluorescence microscopy, the other slide was stained with Auromine O (AO). For both slides, 100 high power fields were read before a smear was declared negative. For a smear to be considered positive, at least two organisms had to be seen.

All specimens are cultured by inoculating concentrated sputum in liquid medium in BBL MGIT bottles using the BACTEC MIT 960 system (Becton Dickinson, Sparks, MD). Cultures were held for 6 weeks before reporting the sputum as negative. Colonies were identified as M. tuberculosis by PCR using Light Cycler 1.5 and confirmed by Hain Lifescience genotype MTBDRplus. TB identification by PCR was also done on the sputum sample.

Assessment for Resistance

Drug susceptibility using broth dilution was determined for 4 drugs per sample: Ethambutol, Isoniazid, Rifampicin and one of either Streptomycin or Pyrazinamide, according to availability in the BACTEC system. Individuals with resistance to both Isoniazid and Rifampicin on culture were classified as having multi-drug resistant (MDR) TB and had further culture sensitivity testing for 3 second line drugs: Ethionamide, Kanamycin, and Ofloxacin. Extreme drug resistant (XDR) TB was defined as MDR plus resistance to ofloxacin plus either streptomycin or kanamycin.

Drug susceptibility for Isoniazid and Rifampicin using molecular approach was determined using Hain Lifescience genotype MTBDRplus (Hain Lifescience, Nehren, Germany).

Statistical Analysis

Subject characteristics, including age, gender, CD4 count, and HIV RNA from the Phidisa baseline are assessed. For the analysis of the TB identification and drug sensitivity, only those specimens that had complete culture results and that were collected from an individual known to be HIV-infected at time of specimen collection were included. For individuals found to be culture-positive for TB, the record was examined for evidence of prior history of TB and for prior history of TB treatment. Fisher’s exact test is used to test for difference in MDR rates based on known TB history.

Assessment for Presence of M. tuberculosis

Using culture as the gold standard, sensitivity and specificity for AO smear, ZN smear and PCR were calculated. For this calculation, if subjects had multiple specimens, only the first culture-positive and first culture-negative tests were included in the analysis. Sensitivity and specificity were also examined separately for subjects whose CD4 level was ≤ 50, between 50 and 200, and >200; only subjects with a CD4 result taken within 90 days prior to and no later than 30 days following the sputum collection date were considered for this analysis. The Cochran-Armitage trend test was used to see whether there was a significant trend for decreasing sensitivity with increasing CD4. Test agreement between culture and PCR were considered separately for smear-negative and smear-positive samples.

Assessment for Resistance

The rate of resistance for each drug tested, the rate of MDR TB, and the rate of XDR TB are reported. If a person had more than one set of susceptibility tests, then the only sensitivity results taken from the first culture-positive sample are included in the comparison of PCR and culture diagnostic susceptibility testing. The agreement between PCR and culture was assessed for INH, RIF and MDR resistance. Additionally, for all culture-positive samples, a chart review was done to determine any prior history of TB or TB treatment within the last three years.

PCR Drug Sensitivity Validation

After the Sputum Initiative had been ongoing for about 18 months, independent validation of a subset of drug sensitivity data was done. Susceptibility results were assessed for accuracy by sending isolates to a second laboratory, SA Medical Research Council (MRC) in Overport South Africa that did extensive work for research studies, and retesting the isolates in a blinded fashion. MRC (HAIN MTBDRPluc PCR) did the molecular diagnostic susceptibility testing from cultured samples sent in MGIT bottles (except for one in a Versatrek bottle). The thirty samples were a convenience sample; that is, all frozen samples that were still in storage at the time of validation were reassessed by MRC.

RESULTS

There were 832 sputum samples taken from 613 HIV-infected patients May 2007 – Dec 3 2008. From these samples, 785 specimens (94.4%) from 584 individuals had complete results. Of the 47 specimens missing culture tests, 42 were indeterminate, mostly due to overgrowth, and the remaining tests were not completed due to administrative or lab errors. There were 116/584 (19.9%) individuals that had at least one culture result positive for TB.

Table 1 presents the Phidisa baseline characteristics of patients in the Sputum Initiative cohort.

Table 1.

PHIDISA baseline characteristics for 584 HIV+ individuals in the Sputum Initiative cohort.

Characteristics
Age (years) (mean, SD)	35.9 (6.0)
Male (n,%)	398 (68%)
CD4 (cells/µL)(median) (25th, 75th percentiles)	144 (68,276)
HIV RNA (copies/mL) (median) (25th, 75th percentiles)	130853 (40300, 283500)

For the calculation of PCR and smear test accuracy, data from 512 first negative tests and 116 first positive tests were analyzed. Table 2a and 2b present the degree of agreement among smear, culture and PCR. Using culture as the gold standard, the sensitivity (percent of culture-positives that tested positive) and specificity (percent of culture-negatives that tested negative) are shown for PCR and smear. PCR had a low positive predictive value (52/99=52.5%), due to the moderate specificity (419/466=90%) and largely culture-negative population that was tested. The positive predictive value for AO and ZN smear was 97.4% (38/39).

Table 2.

Table 2a: Sensitivity and specificity of PCR and smear compared to culture.
	Culture Positive	Culture Negative
	N = 116 (%)	N = 512 (%)
PCR Positive	52 (50.5)	47 (10.1)
PCR Negative	51 (49.5)	419 (89.9)
	103	466
	Culture Positive	Culture Negative
	N = 116 (%)	N = 512 (%)
AO Smear Positive	38 (33.0)	1 (0.2)
AO Smear Negative	77 (67.0)	511 (99.8)
	115	512
	Culture Positive	Culture Negative
	N = 116 (%)	N = 512 (%)
ZN Smear Positive	38 (34.9)	1 (0.2)
ZN Smear Negative	71 (65.1)	480 (99.8)
	109	481

Table 2b: Sensitivity and specificity of PCR compared to culture for AO smear-positive samples (N=39).
	Culture Positive	Culture Negative
	N = 38* (%)	N = 1 (%)
PCR Positive	34 (91.9)	0 (0)
PCR Negative	3 (8.1)	1 (100)
	37	1

^*PCR results are missing for 1 smear/culture positive.

ZN, AO, and PCR had decreasing trends for sensitivity with increasing CD4 levels, but these trends were not significant. There was no variation in specificity. (Table 3)

Table 3.

Sensitivity and Specificity of Direct Sputum Tests According to CD4 Count.

		Sensitivity of Direct Sputum Tests
CD 4 Count	# Culture Positive N=105	ZN	AO	PCR
≤ 50	22	45% (9/20)	42.9% (9/21)	66.7% (12/16)
51–200	55	33.3% (18/54)	32.7% (18/55)	47.1% (24/51)
>200	28	25.0% (6/24)	21.4% (6/28)	45.8% (11/24)
		Specificity of Direct Sputum Tests
CD 4 Count	# Culture Negative N=477	ZN	AO	PCR
≤ 50	52	100% (48/48)	100% (52/52)	87.5% (42/48)
51–200	193	99.5% (181/182)	99.5% (192/193)	88.2% (150/170)
>200	232	100% (216/216)	100% (232/232)	92.0% (196/213)

Differences between CD4 groups did not reach statistical significance. Specimens in this analysis limited to 105/116 culture-positive and 477/512 culture-negative specimens from individuals with a CD4 count taken within 90 days prior and 30 days following culture.

Drug Susceptibility

Figure 1 displays the resistance rates for first line drugs. There were 20/107 subjects who were MDR on initial diagnostic susceptibility testing and 22/107 (20.6%) that were MDR on first or later diagnostic susceptibility testing. XDR was detected on 3/107 (2.8%) individuals on an initial isolate and a fourth was detected 2 months after an initial pansensitive test.

Figure 1.

Culture drug sensitivity results by culture (N=107*).

Results of drug susceptibility testing for 107 isolates, performed by culture based method. Not all isolates were tested for all drugs.

* First line susceptibility was available for 107/116 culture positive individuals. If individuals had more than one sample, results are presented for first diagnostic susceptibility testing.

† Second line drugs were done only for 19/20 individuals that had MDR findings on first positive sample.

+ MDR resistance is defined as resistant to both isoniazid and rifampicin. XDR-TB is defined as MDR plus resistance to ofloxacin and at least one of streptomycin/kanamycin.

** Complete second line test results were missing for 2/20 individuals that were MDR on first diagnostic susceptibility testing. One additional individual was found to be XDR on a subsequent diagnostic susceptibility testing.

Figure 2 shows the history of TB treatment by MDR status, as determined by self-report and chart review. The rate of MDR (alone or with XDR) was 11.9% in patients with no known prior TB treatment history, and 27.1% amongst those with evidence of prior TB treatment (p=0.05).

Figure 2.

Rates of MDR alone and MDR with XDR in TB cases by prior TB treatment history (N=107*).

Rates of prior history of tuberculosis treatment based on patient self-report and patient records for isolates that were drug susceptible, or MDR, or XDR.

* MDR status based on first diagnostic susceptibility testing sample for individuals with multiple samples.

Among the 107 culture-positive samples with diagnostic susceptibility testing, there were 103 that also had PCR drug sensitivity results for INH and RIF. PCR agreed with the culture INH and RIF sensitivity results for 94/103 (91.3%) of the samples. Using only PCR sensitivity to INH and RIF to define MDR, 13/20 (65%) individuals would have been correctly identified as MDR, with no false MDR detections (data not shown).

Since patients received their antituberculous therapy in a different health care system, information was not routinely available regarding the antituberculous drugs prescribed, the duration of therapy, adherence to therapy, or follow-up microbiologic results.

PCR Validation Study

There were 28 TB culture-positive samples sent that MRC lab was able to test by molecular techniques for isoniazid (INH) and rifampicin (RIF) susceptibility testing (diagnostic susceptibility testing). All 28 samples were classified as sensitive to INH by MRC PCR, but 1/28 (3.6%) of these samples was designated as resistant by the study PCR (BARC Laboratory). The overall agreement rate for RIF resistance was 24/28 (85.7%). For RIF, there were 2 samples shown to be resistant by both MRC and the study PCR and 22 shown to be sensitive by both labs; however, there were 4/26 (15.4%) MRC sensitive samples that were classified as resistant by BARC. None of the 28 MRC results were MDR, but 1/28 (3.6%) was classified by BARC as MDR.

Discussion

South Africa has a particularly heavy burden of TB: in 2005 it was estimated that there were 285,000 incident cases of TB in South Africa [1, 18]. Thus strategies to screen for TB, to treat TB effectively, and to prevent transmission are desperately needed.

This study demonstrated that clinicians could identify patients with a high likelihood of TB using general clinical parameters that focused on long-term cough, apparent weight loss, or fever. In these HIV clinics, 19.9% of patients who were asked to provide sputum did in fact have positive sputum cultures for TB. Other HIV clinics in South Africa have reported high yields. Yields of 19–26% have recently been reported in several South African HIV programs which obtained either expectorated or induced sputum on one or more occasions from all patients attending HIV clinics regardless of the presence of suggestive symptoms or signs [18–21]. Thus, in South Africa, active screening of symptomatic and perhaps asymptomatic HIV-infected patients, at regular intervals seems likely to be highly effective.

Many programs in developing countries rely on smear alone to identify patients with confirmed TB. However, smear was an insensitive technique for confirming TB in this South African setting at all CD4+ T cell strata (Table 3): only 33% of culture-positive patients were positive by AO smear. This is consistent with 8–29% rate reported in recent investigations from South Africa and other developing countries involving HIV-infected patients [19, 20, 22–25] marginally below the 45–80% rate reported by the CDC [23] but well below the 93% that some reference laboratories can attain [26]. The performance characteristics of smear might be enhanced by improved training, or the opportunity to spend more time on each specimen. However, other techniques are needed to rapidly identify patients with confirmed TB, especially if the number of organisms is small [3, 22, 25, 26].

A nucleic acid amplification test would be a logical technique to identify more culture-positive patients at the time of their initial visit, and to identify those organisms that were resistant to Isoniazid or Rifampin [3, 23, 27–32]. In the United States, the CDC recently issued updated guidelines supporting such an approach [23].

In the current study, PCR had a low positive predictive value (52.5%), due to the moderate specificity (90%) and largely culture-negative population that was tested (TB prevalence, 19.9%). Thus, as performed in this study, PCR alone was not sensitive or specific enough to be reliable to definitively diagnose a TB-positive individual. On the other hand, in this setting, PCR had a high negative predictive value (89.1%). These PCR results were obtained from a large, widely used commercial laboratory. A research laboratory might have achieved different results.

Most TB programs in South Africa do not use culture and susceptibility testing as part of their TB program due to cost constraints and absence of appropriate technical and personnel resources [1–3, 13–15]. It is well known that MDR and XDR are present in South Africa, yet there are few published data documenting prevalence.

The WHO Global Report on Antituberculosis Drug Resistance in the World has recently reported that Africa recorded one of the lowest median levels of drug resistance worldwide: the MDR rate for Africa was a mean of 2.2% (range 0.0–5.8%) [1, 5, 8, 11, 13]. South Africa, in particular, was reported to have MDR prevalence of 3.1% [1, 5, 8, 13]. There has been much discussion about why these rates were so low, but underreporting has been suspected [11, 33]. In many reports it is not clear whether the data included both treatment naïve and treatment experienced patients.

Other studies reporting national data and regional surveys suggested that MDR-TB rates began to rise in the late 1990’s in South Africa [1–12, 19, 34, 35]. A systematic survey in 2001–2002 of 9 South African provinces, conducted by the Tuberculosis Lead Programme of the Medical Research Council reported a rate of MDR of 1.6% in isolates from patients with newly diagnosed TB (provinces ranged from 0.9% to 2.6%), and 6.6% in patients with previously treated TB (provinces ranged from 3.9%–13.7%) [10, 11]. The Tuberculosis Strategic Plan for South Africa, 2007–2011, reported that from 2004–2007 there were over 11,000 laboratory confirmed cases of MDR in South Africa, many coming from the Western Cape [13]. For 2006, the reported rates for MDR were 0.4% to 2.67% by province. During the same period there were over 800 cases of XDR reported.

In this current HIV+ patient population of geographically mobile, actively employed individuals and their dependents, MDR was present in 20.6% of TB cases. PCR testing of isolates for INH and Rifampin resistance showed a high correlation with culture based results, providing additional confirmation of the accuracy of the results. The presence of MDR in patients who self reported no prior TB treatment was 11.9%, which is consistent with several recent reports from South Africa [6–12, 19]. The presence of MDR in patients who did self report prior TB treatment was considerably higher than previously reported, 27.1%.

HIV-infected patients with TB have a poorer one year survival than HIV-infected patients without TB. MDR and XDR add to mortality incrementally. A recent study in South Africa reported a one year mortality among HIV-TB co-infected patients with MDR of 69%, and a one year mortality with co-infected XDR patients of 82% [12, 15].

In the HIV-infected patient population studied in South Africa, where unrecognized TB was common, more consistent and uniform interventions are needed to screen high risk patients for TB. It seems likely that given the national prevalence of TB, all HIV-infected patients in South Africa should be considered high risk regardless of CD4+ T-cell count or symptomatic status. More rapid and accurate diagnosis and more effective management strategies are needed if the cycle of transmission and poor outcome in the community, in the workplace, and in health care settings is to be broken.