Kvasnovsky, Charlotte L MD, MPH*†; Cegielski, J Peter MD, MPH‡; Erasmus, Roshen RN§; Siwisa, N Olga RN§; Thomas, Khulile MD§; der Walt, Martie L van MD, MBA†
Tuberculosis (TB) remains a leading cause of morbidity and mortality worldwide. Multidrug-resistant TB (MDR-TB) is an increasing threat, causing an estimated 489 000 cases worldwide in 2006.1 MDR-TB is defined as TB that is resistant to at least the 2 most important first-line anti-TB drugs, isoniazid (INH) and rifampicin (RIF). Treatment for MDR-TB is less effective, more toxic, and more costly than is treatment for drug-sensitive TB, requiring the use of less active second-line drugs (SLDs). Extensively drug-resistant (XDR) TB is additionally resistant to the 2 most important SLDs, a fluoroquinolone and any injectable SLD [amikacin (AMK), capreomycin (CAP), or kanamycin]. Since first described in 2005, XDR-TB has been diagnosed in 49 countries and accounts for an estimated 4.9% of MDR-TB cases.2
Drug-resistant TB and HIV coinfection were described as a “perfect storm” of illness.3 The first outbreak of XDR-TB to gain public attention was in KwaZulu-Natal Province (KZN), South Africa, where all XDR-TB patients tested for HIV were positive and 52 of 53 cases died rapidly after diagnosis.4 Other reports have confirmed high overall mortality in XDR-TB patients, with significantly worse outcomes for HIV-positive-patients.5,6
In 2006, Eastern Cape Province, located south and west of KZN, had a TB incidence of 705 of 100,000. In 2008, it had the single highest MDR-TB caseload in South Africa, however, second-line drug susceptibility testing (SLDST) to diagnose XDR-TB was not available until late 2006. Medications, outside of those used in the standardized treatment of MDR-TB, were also not available. Eastern Cape Province began treatment in October 2006 using a combination of drugs known to be effective against TB, both first-line drugs used in the treatment of MDR-TB [ethambutol (EMB), pyrazinamide (PZA)] and SLDs that were reintroduced to the province [para-amino salicylic acid (PAS), CAP]. In this study, we present the first province-wide follow-up of consecutive patients treated for XDR-TB in South Africa, so as to examine the effect of HIV infection on 12-month survival, and the possible effectiveness of treatment.
An estimated 341,165 cases of pulmonary tuberculosis were diagnosed in South Africa in 2006 of which an estimated 1.6% of new cases and 6.6% of previously treated cases had MDR-TB.1
MDR-TB has been treated exclusively in the public sector in Eastern Cape since 2000 based on drug susceptibility testing (DST) for RIF, INH, and EMB. Patients with MDR-TB were treated with a standardized regimen of 4 months of AMK, ofloxacin (OFL), ethionamide (ETA), EMB, and PZA, followed by 12 months of OFL, ETA, and EMB. Terizidone (TRZ) was available starting in 2003 as a replacement for patients whose TB was either resistant to EMB or with serious adverse effects from a medication.
In October 2006, after XDR-TB outbreaks in neighboring KZN, Eastern Cape began diagnosing XDR-TB cases through quality-assured SLDST. Treatment for XDR-TB was supplemented with CAP and PAS, whereas use of AMK and OFL was discontinued.
In accordance with provincial policy, patients with XDR-TB were hospitalized at 1 of 2 MDR-TB referral centers until they submitted 3 consecutive sputum specimens that were culture negative for Mycobacterium tuberculosis. After discharge, culture-negative patients were followed at outpatient clinics but received medications though the MDR-TB referral centers. No patient with XDR-TB had resectional surgery during the study period.
TB cases in South Africa had access to voluntary counseling and testing and antiretroviral therapy (ARVs) since 2003. In accordance with national policy, HIV-positive MDR-TB patients not yet on ARVs were initiated on ARVs either at the completion of the 4-month intensive phase of MDR-TB treatment if they had a CD4 lymphocyte count less than 200 cells per cubic millimeter or once their CD4 lymphocyte count was below 200 cells per cubic millimeter.
Laboratory Diagnosis and Case Definition
National Health Laboratory Services performs all DST in the public sector. Specimens are cultured initially in mycobacteria growth indicator tube broth. DST is performed by the indirect method on Middlebrook 7H10 agar for INH, RIF, AMK, OFL, ETA, and EMB. Up to July 2007, only patients already on treatment for MDR-TB were eligible for SLDST. After July 2007, patient isolates newly found to be resistant to MDR-TB were also tested against SLD.
XDR-TB was diagnosed when any combination of sputum specimens had M. tuberculosis resistant to at least RIF, INH, an injectable SLD, and a fluoroquinolone. All patients diagnosed with XDR-TB from October 1, 2006, to January 31, 2008, in Eastern Cape Province were included in the study. All patients had at least pulmonary TB disease.
PATIENT SELECTION/INCLUSION CRITERIA
National Health Laboratory Services reported all XDR-TB results to the main MDR-TB referral center in the province, Jose Pearson Hospital in Port Elizabeth, and patients were traced to initiate treatment. Medical records of all patients diagnosed during the catchment period were retrospectively reviewed 12 months after XDR-TB treatment initiation, between June 2008 and March 2009. Three patients were referred from outside sources (1 transfer from another province, 2 diagnosed by private laboratories). Treatment was only available through the public sector. The final analysis was limited to those patients who initiated treatment and had a known HIV status.
MAIN PREDICTOR VARIABLE
A patient was considered HIV positive if the diagnosis was noted in the patient's medical record. A patient was considered to have AIDS if a CD4 lymphocyte cell count less than 200 cells per cubic millimeter before or at XDR-TB treatment start was noted or if the patient was known to be on ARVs.
Clinical information on patient background and past medical history were collected from individual medical records, including previous tuberculosis episodes. Radiographic findings were abstracted from the medical records based on the reading recorded by the treating physician of a chest x-ray obtained within 2 months of XDR-TB treatment initiation. Patient height and weight was measured at treatment initiation. Serious adverse effects of medication were recorded when a drug was documented to have been withdrawn due to a side effect.
TRZ and cycloserine were used interchangeably according to provincial availability and analyzed together as TRZ. Dapsone was included in the initial treatment regimen in 97% of patients but not considered a potential effective drug.1 Patients were treated with individualized regimens composed of the 6 first-line and second-line anti-TB drugs available in the province at the time, CAP, PAS, ETA, TRZ, EMB, and PZA (Fig. 1).
Patients were considered to be on “effective” treatment if they received at least 4 drugs to which their TB could be considered susceptible.1 A drug was considered “effective” if (1) it is recognized as an agent for the treatment of TB7; (2) the patient had either never received it or received it for less than 3 months before XDR-TB treatment; and (3) patient isolates were not found to be resistant to the drug on DST.
Three months was chosen as the maximum previous duration of use because no drugs in a regimen would be considered effective if treatment was failing after 3 months. After this period, the likelihood of acquired resistance is substantial, irrespective of subsequent DST results given the suboptimal specificity of conventional phenotypic DST.
Interim Treatment Outcomes
Patients were assigned to 1 of 2 groups based on their vital status 12 months (365 days) after treatment initiation: either alive and continuing treatment or dead from any cause.8 One patient defaulted at 186 days and was not included in mortality analysis.
Sputum cultures were performed at XDR-TB treatment initiation and every subsequent month to monitor treatment. Patients who began XDR-TB treatment with a positive sputum culture and had 2 consecutive negative sputum cultures taken at least 30 days apart after treatment initiation were considered to have achieved culture conversion. Initial time to sputum culture conversion was calculated as the interval in days between the date of treatment initiation for XDR-TB and the collection date for the first of 2 consecutive negative sputum cultures.
Double data entry was performed on a database created in EpiInfo version 3.3.2 (CDC, Atlanta, GA). Analysis was done with SAS version 9.1 (SAS Institute, Inc., Cary, NC).
Univariate and bivariate analyses were done to show differences between HIV-positive and HIV-negative patients. Categorical data were compared using χ2 or Fisher exact test. Parametric continuous variables were compared using the Student t test and nonparametric data using the Wilcoxon signed-rank test.
For comparison of unadjusted mortality, we used a Kaplan-Meier plot, stratifying by HIV status, and the log-rank test, with death as the event variable and all other outcomes censored at 365 days. Multiple logistic regression was used to determine the association between HIV and vital status 12 months after starting treatment controlling for covariates, such that programs could best assess their needs. All tests were 2 sided, with P < 0.05 considered statistically significant.
The Medical Research Council of South Africa, Eastern Cape Department of Health, and University of Maryland Institutional Review Boards granted approval for the study.
From October 2006 to January 2008, 274 patients were diagnosed with XDR-TB in Eastern Cape Province, of whom, 68 did not initiate XDR-TB treatment (65 patients died before diagnostic results were available and they could be initiated on appropriate treatment, 1 transferred to another province, and 2 were not started on treatment). The HIV status of the 65 patients (23.7%) who died before treatment start was not known. The majority (more than 80%) of HIV-positive patients were diagnosed with HIV before XDR-TB treatment initiation. Eleven patients refused testing for HIV or did not have results available. These patients were similar to the remainder of the cohort (data not shown). Thus, 206 patients started treatment for XDR-TB during the 15-month study period.
Of 195 patients tested, 108 patients (55%) were HIV positive. Seventy-seven patients were known to meet definition for AIDS, including 59 known to be on ARVs. Patients had a median duration of 5 months of MDR-TB treatment before starting XDR-TB treatment [interquartile range (IQR) 1-8], with no difference between HIV-positive and HIV-negative groups (P = 0.13) (Fig. 2).
Twenty-one patients did not have a chest x-rays taken at XDR-TB treatment initiation or a smear result reported. Patient height was missing for 118 of 206 patients and was not included in the analysis nor was body mass index. Patients waited a median of 59 days (IQR 43-65) between sputum specimen submission and XDR-TB treatment initiation.
Interim Outcome Analysis
Overall, 35% of HIV-negative patients died compared with half of HIV-positive patients (P = 0.03) (Table 1). Kaplan-Meier survival analysis showed significantly higher mortality in HIV-positive patients after 3-month, 6-month, and 12-months of treatment (P = 0.03) (Fig. 2). This relationship did not retain significance in the full multivariate model, whereas all other variables that were significant predictors of mortality in the bivariate analysis retained statistical significance in the logistic regression (Table 2).
Among individual bivariate comparisons, there were several predictors of death within a year of treatment initiation. Adjusting for all other covariates, lower weight decreased the odds of survival by 7.5% per kilogram (P < 0.01) (Table 2). The odds of death at 12 months were increased 2-fold for patients smear positive at XDR-TB treatment start (P = 0.01). Each additional month of previous TB treatment, both drug-sensitive and MDR-TB, increased the odds of death by 3% (P = 0.02). Using patients under 25 years old as the referent group, patients 25-42 years of age at XDR-TB treatment start had 3.5 times the odds of early mortality, whereas there was no survival difference for patients older than 42 at treatment start (P = 0.01 and P = 0.15).
In the multivariate logistic regression model, however, there was no significant difference in 12-month survival between patients with and without HIV. However, HIV-positive patients not on ARVs had significantly poorer 12-month survival than did either HIV-positive patients on ARV treatment (P = 0.05) or HIV-negative patients (P < 0.01). There was no difference in 12-month survival between HIV-positive patients on ARV treatment and HIV-negative patients (P = 0.44).
In a reduced multivariate model including only HIV-positive patients not on ARVs and known HIV-negative patients, HIV-positive patients not on ARVs had 2.5 times the odds of death at 12 months (P = 0.04, Table 3).
An XDR-TB treatment regimen including all 6 available drugs was used in 18 of 206 patients, whereas 155 of 206 received at least 5 of these (Fig. 1). First-line drugs EMB and PZA were considered effective in 11.3% of patients, whereas medications used in second-line regimens for MDR-TB were considered effective in 29.7%-68.7% of patients (ETA and TRZ, respectively). PAS and CAP were considered effective in 85.1% and 100% of patients.
Overall, 36 patients (18%) experienced a severe adverse effect to at least 1 anti-TB medication. Gastrointestinal adverse effects resulted in the withdrawal of ETA and PAS in 29 and 4 patients, respectively. Other severe adverse effects were linked to PZA in 1 patient and PAS in 3 patients. HIV-negative patients experienced severe adverse effects at 3 times the rate of HIV-positive patients, 31% as compared with 9.3% (P < 0.01). ARV use had no effect on rate of adverse effects among HIV-positive individuals (P = 0.33).
Primary Versus Acquired Drug Resistance
Of 195 patients, 36 (18%) had an isolate tested for resistance to AMK or OFL before XDR-TB diagnosis that was sensitive to at least 1 of these drugs. These patients either acquired additional drug resistance or were re-infected with a strain of XDR-TB. Thus, of the total 274 cases of XDR-TB diagnosed, 112 of 274 patients (41%) had primary XDR-TB and 36 of 274 (13%) likely acquired XDR-TB during treatment for MDR-TB.
Of the 195 patients, 190 were culture positive at treatment start. The 5 culture-negative patients remained culture negative for the study duration. Sixteen patients (8.4%) converted to negative sputum culture in a median of 143 days (IQR 90-207.5 days). There was no difference in the frequency of culture conversion between HIV-positive (10 of 108, 9.2%) and HIV-negative (6 of 87, 6.9%) patients (P = 0.55).
Before the emergence of XDR-TB in 2005, patients and providers in Eastern Cape Province did not have access to SLDST. Starting in October 2006, 274 patients were diagnosed with XDR-TB in the province, one of the highest notification rates in the world.1 Of these, 65 patients (23%) did not survive long enough to start treatment. Among the 206 patients who started treatment for XDR-TB, HIV-positive patients had similar 1-year survival to HIV-negative patients, in contrast to those HIV-positive patients not on ARVs. Nevertheless, 1-year mortality of XDR-TB patients was very high: 95 of 206 patients (46.1%) initiating treatment and 160 of 274 patients (58.4%) diagnosed with XDR-TB did not survive 1 year.
Low body weight was the single strongest predictor of 12-month mortality. Excluding patients less than 18 years old at the time of diagnosis who could be considered not to have reached an adult weight did not alter the results (data not shown). Indeed, children had higher survival than did older patients in the study cohort (75% survival), a finding that strengthens the association between decreasing weight and early mortality. Studies have linked low body weight and poor outcomes in both TB and HIV.9 Interventions targeting nutritional status and low body weight should be evaluated urgently.
Patients with XDR-TB were treated with an average of 5.2 drugs, whereas the WHO recommends the use of at least 4 drugs to which the patient's TB is considered susceptible.1 Of those total drugs, patients received an average of only 1.7 drugs considered “effective”. Given overall poor outcomes and that treatment with an “effective” regimen was not found to be significant in the univariate or the multivariate model, it is possible that this recommendation may not be practicable in XDR-TB. There may not be enough potent, tolerable, and “effective” drugs to fill a regimen to reliably allow for culture conversion and cure for patients with XDR-TB.
This analysis attempted to move beyond simply counting the number of drugs in a regimen, by assessing the potential effectiveness of each drug within a regimen. This may be a more useful measure of how beneficial a given drug may be, especially in settings where extended first-line and second-line DST are not available. It does not, however, take into account differing intrinsic tuberculocidal activity of medications nor can it replace more complete SLDST. Although it is not a controlled comparison, other settings offer patients a wider variety SLDST, treat patients with anti-TB medications to which their organisms are phenotypically susceptible, and report better survival than Eastern Cape.10,11
The experience of Eastern Cape Province highlights the importance of expanding the availability of quality-assured SLD susceptibility testing and judicious use of SLDs in TB programs. Of 274 patients diagnosed with XDR-TB, 46% were treated with SLDs for greater than 1 month before SLDST was ordered. These patients may have developed SLD resistance during treatment, reemphasizing the need for diligent SLDST in Eastern Cape, to ensure patients receive appropriate treatment in a timely manner.
We provide new information on the impact of high HIV coinfection rates on XDR-TB, including the observed benefit of ARV treatment regardless of CD4 lymphocyte count. Programs achieving the highest cure rates reported less than 1% HIV coinfection in XDR-TB patients, as compared with 55% coinfection in Eastern Cape. This should account for part of the high mortality in our population. However, we found no difference in mortality between HIV-positive patients treated with ARVs and HIV-negative patients. Restricting the multivariate model to HIV-positive patients not on ARVs and HIV-negative patients, HIV-positive patients had 2.5 times the odds of early mortality as did HIV-negative patients (P = 0.04, Table 3). Similar results were found in comparing HIV-positive patients on ARVs and those not on ARVs (data not shown), highlighting the crucial importance of ARVs in the comanagement of HIV and XDR-TB.
Our results shed light and perspective on a bleak situation, that of extremely high mortality of patients with XDR-TB in a setting with high HIV coinfection.4 The emergence of XDR-TB in South Africa parallels that of MDR-TB in the 1990s. In New York, patients with MDR-TB initially had very high mortality, especially those patients coinfected with HIV.12 As awareness of drug resistance increased, patients were better managed on more appropriate treatments, leading to improved outcomes.13 South Africa became aware of the emergence of resistance to SLDs in a similarly startling manner and has improved access to SLDST, availability of drugs, and infection control measures nationwide.
Overall, we saw low rates of culture conversion within the first year of treatment, 8.4% of those who started treatment. Treatment programs in Peru and Russia have shown higher rates of culture conversion and eventually cure in patients with XDR-TB, with cure rates of up to 60%.10,11 These programs treated patients with more drugs on average, including drugs to which the patients were known to be resistant (RIF, OFL) or likely to be resistant (drugs in the same class as drugs with known resistance such as moxifloxacin and levofloxacin). Resectional surgery was also available, which may be a great benefit especially to patients in whom TB remains limited to 1 lung.14
During this study's duration, anti-TB drugs were added to the regimens of XDR-TB patients, including high-dose INH, amoxicillin/clavulanic acid, and moxifloxacin. Moxifloxacin is effective in vitro against 20%-40% OFL-resistant M. tuberculosis, although its effectiveness in vivo is not known.15
Until April 2010, protocol in South Africa was to delay initiation of ARVs in all HIV-positive patients until the initial 4-month or 6-month intensive phase of MDR-TB treatment. In our study, HIV-positive patients, even those patients already on ARV treatment, had fewer adverse effects from XDR-TB treatment, suggesting that dual therapy can be tolerated. Furthermore, one retrospective analysis found that initiation of ARVs within the intensive phase of treatment for drug-sensitive TB significantly reduced overall morbidity and mortality.16 Guidelines that went into effect across South Africa in April 2010 will allow for the initiation of ARVs in any HIV-positive patient initiating XDR-TB treatment, regardless of CD4 lymphocyte count.
The HIV-positive patients not on ARV treatment had a median CD4 lymphocyte count of 215 cells per cubic millimeter (IQR 67-355). Our data suggest that these patients might have benefited from early ARV treatment. With continued integration of TB and HIV services in South Africa, and rapid polymerase chain reaction-based testing for anti-TB drug resistance, more patients will be diagnosed in time to initiate appropriate XDR-TB treatment (van der Walt, PhD, MBA, personal communication from 2010). Given better outcomes for those HIV-positive patients on ARVs, the evidence supports earlier initiation of ARV therapy.
© 2011 Lippincott Williams & Wilkins, Inc.