The data from our study stresses the need for a rapid TB diagnostic test in HIV patients in Bolivia. The rate of TB in HIV patients with respiratory symptoms was very high (45 %), and almost half (46 %) of these patients died within 60 days. Furthermore, the rate of TB-HIV co-infection in our study alone over a 1.5 year period (n = 48) does not correlate well with the rate of HIV-TB co-infection in all of Bolivia reported in 2013 (n = 170) [1]; this data raises concern that this is a problem of greater magnitude than is currently appreciated. Even more alarming is the prevalence of MDRTB in our population of HIV patients (9 %), which is much higher than the reported MDRTB rates in Bolivia; and that 75 % of those with resistance to one or more drugs died during the duration of our study. A similar study by Kawai et al. in Peru showed that MDRTB-HIV co-infection had a mortality rate above 50 % within two months and was associated with a high rate of continued infectiousness [29]. Current methods of resistance testing in Bolivia can take weeks beyond positive solid culture results, risking significant delays of appropriate treatment. These data strongly reinforce the need for the implementation of a rapid TB diagnostic method with direct drug susceptibility testing in the HIV population in Bolivia. Additionally, the discrepancy between the rates of resistance found in our study compared to previous reports, calls for a dedicated study of antibiotic resistance in Bolivia in HIV patients. Given the increased rates of MDRTB in HIV, the risk of nosocomial spread, and the high mortality associated with this co-infection, if these rates of MDRTB are confirmed in future studies, MDRTB in Bolivia could quickly become a greater threat to public health.
Our study demonstrates that MODS is a valuable method for rapid diagnosis of TB in the HIV population. In this study, the detection yield of MODS was superior to LJ. Additionally, consistent with prior studies based in centralized MODS laboratories, Mtb was observed in MODS in less than one-third of the time of growth in LJ [18, 19, 23]. MODS’ increased diagnostic yield was most profound in the smear-negative HIV patients, a vulnerable group at highest risk for missed or delayed diagnosis. Although smear-negative sputum samples cultured by MODS grew later than MODS smear-positive samples, smear-negative samples by MODS still grew significantly faster than those by LJ.
Our results emphasize the utility of MODS in the ambulatory population. MODS detected almost 30 % more disease than LJ in this population, and LJ did not identify any more disease than ZN alone, supporting a use for MODS in patients with milder presentations.
As this study highlights the need for widespread implementation of a rapid, sensitive TB diagnostic method in HIV patients with direct drug susceptibility testing, it is important to mention the role of GeneXpert as compared to MODS. GeneXpert would also be a welcome alternative to the current standard of care in Bolivia for HIV patients. However, at the current time, without a committed long-term external funding source, its cost ($10/cartridge and $17,000/machine) limits widespread implementation in resource-limited settings like Bolivia [13]. MODS ($3/test) remains a more cost-effective alternative in these settings. Additionally, given MODS performance in smear negative patients as demonstrated in this study and others, and GeneXpert limitations in this domain [14, 15], we would also recommend MODS as a low cost alternative for GeneXpert for smear negative patients.
Our study provides data about sputum scarce HIV patients with respiratory symptoms in a TB endemic area. Despite minimal sputum production, the prevalence of TB in this population trended towards higher than the sputum-productive subjects and the 60 day mortality rate of those TB+ was greater in the sputum-scarce group than the sputum productive patients.
Additionally, we examine the value of the string test as a sample obtainment method for TB in sputum scarce or absent patients. In our sputum-scarce group, the string test detected over 80 % of TB cases, and was not significantly different than the number of TB cases by induced sputum. Importantly, in the 10 patients that could not produce a sputum sample despite induction, the string test detected 4 cases of TB that would have otherwise been missed; thus 40 % of patients that could not produce a sputum sample were diagnosed and received treatment because of the string test. Our diagnostic yield is lower than in previous studies [24], this may be due to lack of neutralization before storage and delayed processing of samples. Despite this, the string test provided an inexpensive, minimally invasive alternative for diagnosis of TB in sputum-scarce and sputum-absent patients without an associated biohazard risk. Given the above data, we suggest that the string test be considered as an alternate sample obtainment method used in conjunction with MODS in HIV patients with little to no sputum production.
Of the patients with both induced and spontaneous sputa (n = 24), although the difference of the results between the two groups were not statistically significant, two patients that were negative for TB by spontaneous sputum were TB+ by induced sputum. Although induction of sputum in the diagnosis of TB is historically more common in the pediatric population [30, 31], more recent studies evaluating the utility in adults in HIV-endemic regions have come to differing conclusions [32–35]. Small studies have promoted the use of induction of sputum by demonstrating an increased diagnostic yield when compared to spontaneous sputum [32–34]; however, a randomized controlled trial of 418 patients in South Africa did not show a significant difference between the results of healthcare worker-instructed versus induced sputum [35]. Furthermore, induction of sputum requires additional equipment, poses a biohazard risk, and is an increased cost to hospitals (US$2.14 vs $7.88) [35].
Additionally, our study is the first in Bolivia to demonstrate that MODS is superior to the current local standards of tuberculosis diagnosis. We showed that a laboratory with minimal prior experience with MODS is able to produce results similar to that of centralized MODS laboratories, as in Peru, where MODS is now the standard of care [23].. Furthermore, the implementation of MODS can be more simplified and affordable in the future, with the recent development of MODS kits and availability of inexpensive inverted microscopes [21–23].
Limitations of our study include our increased time to positivity compared to prior studies in Peru [18, 19]. This increase may have been influenced by delay in processing times of the samples, and the fact that samples were checked less often (due to personnel availability), rather than every day. Although most samples were cultured within 4 days of collection (Median 3.5, IQR 1–6), some were stored for up to 11 days, which may have affected growth of culture. The diagnostic yield of the string test in this study is lower than previous studies [24]; it is important to note our lack of neutralization of the sample before storage as well as delays in processing samples may have decreased string test yield and is a limitation of our study.
Our imperfect reference standard was an additional limitation to optimally evaluate the diagnostic performance of MODS. The performance of MODS would have been more clearly defined if our standard included other sensitive diagnostic methods such as MGIT or GeneXpert rather than a composite of all of the tests being evaluated. This was due to logistical limitations and financial constraints. Although the comparison of MODS to the standard solid culture media for diagnosis of TB in Bolivia provides important data for the community tested, the incorporation of GeneXpert in our study would have been a valuable comparison, had it been a possibility.
Additionally, there is a gap in our data regarding the mortality of the HIV patients that were negative for TB. Regrettably we did not record the mortality data for all patients without TB during the time of the study, thus we are not able to make a comparison in mortality between the TB-positive and TB-negative patients.
In many parts of the world, HIV is no longer a death sentence, but has become a manageable chronic disease due to increased detection and wider availability of effective treatments. The currently undertreated and highly lethal HIV-TB co-infection is a barrier to this becoming a reality in many TB-endemic areas such as Bolivia. Implementation of MODS, or if economically feasible, another rapid diagnostic test such as GeneXpert, should become the standard for diagnosis of TB in HIV populations in resource-scarce settings. This would empower clinicians to treat patients with effective antimicrobials at earlier time points, which would lead to significant effects on both patient mortality and public health through decreased transmission of this deadly disease.