This review included considerably more information than our previous version, allowing validation of NFTMX and BZD as standard forms of TT. However, the substantial growth of the data available is largely from observational studies. Thus, the studies are at substantial risk of bias and convincing results (i.e. those needed to support the notion that TT is generally beneficial for T. cruzi-infected individuals) would require methodologically robust studies yielding a large and precise effect size and consistent results. This is not the case for the data summarised in this review, as data on clinical efficacy of TT remains imprecise and highly inconsistent.
Summary of main results
This updated systematic review provides three major findings: first, a substantial growth in the information produced on this topic since 2000. Second, more results favouring the efficacy of TT, although hampered by methodological flaws, imprecision and inconsistency. Third, a more reliable estimate of the side effects of conventional drugs.
The first version of this review presented data from five RCTs including 756 participants reporting predominantly parasite-related outcomes. Now, some of those RCTs (i.e. Andrade 2004) have reported extended follow-up periods and provide data regarding a new parasite-related outcome, T. cruzi circulating materials detected by PCR techniques. This review also adds safety data from a relatively large, blinded RCT of TT with BZD in adults (the TRAENA trial, still ongoing, with results on efficacy pending). Further, inclusion of non-RCTs allowed extraction of additional patient-related outcome data from over 3000 individuals. This growth in the amount of information reflects the renewed interest for TT and for resolving the uncertainty around its efficacy.
This data-enhanced review confirms the validity (i.e. proof of concept) of TT as a treatment reducing parasite-related outcomes (including now a positive PCR after treatment). It also suggests that TT reduces both mortality and progress of the associated CCC. However, this potential reduction in outcomes such as ECG abnormalities, progression of CCC and mortality is associated with a number of limitations.
First, our patient-related outcome data on efficacy originate from studies other than RCTs. In seeking this type of information, we accepted the inclusion of studies in which allocation to TT was both non-randomised and open-label. We tried to reduce bias by extracting outcome data that were less prone to influence from assessors (i.e. 'hard' outcomes). Examples included extracting data on severe, rather than mild side effects, and defining progress of CCC as a composite change, rather than an isolated finding, or after at least four years of follow-up. Since the only patient-related efficacy outcome data generated from RCTs were seven cases of new ECG abnormalities in two studies, definitive interpretation of clinical efficacy of TT was not possible on the basis of the currently available data.
Second, in addition to our concerns on validity, there is imprecision and statistical heterogeneity of the combined effect size estimates. For example, mortality data for nitroderivatives (the best-known drugs as valid TT in terms of reducing parasite-related outcomes) still appears to include a risk excess of 56% among individuals receiving active treatment. Furthermore, heterogeneity for mortality or progression of CCC for adults treated with nitroderivatives remains substantial (I2 statistics of 49% for mortality and 66% for progression).
Third, as shown by meta-regression analysis, efficacy data on patient-related outcomes have no relationship to results on serology. In addition, the efficacy of ALLOP was greater than that of nitroderivatives in terms of progress of CCC and mortality, but was smaller in terms of parasite-related outcomes. In summary, no patient-related efficacy outcome data summarised in this review replicate findings from at least one RCT in terms of both statistical significance and homogeneity.
An additional aspect of our findings derived from having access, for the first time, to data from the TRAENA trial (the first report of an RCT testing BZD in more than 600 adults). This study identified reliably the mild side effects and, along with other studies, provided data on severe side effects. While rash was the mild side effect more strongly associated with BZD (20% of treated individuals developed moderate rash, a seven-fold greater incidence than controls), participants had several severe side effects.
Treatment with BZD in TRAENA was associated with fever, lymphadenopathy, arthritis and polyneuritis (around 2.5% versus no participants in the control groups). Combined data showed that TT with both BZD and NFTMX was also associated with hepatitis, at least in the short term. This finding had approximately the same frequency of the other side effects (17 of 578, or 2.9% of individuals exposed to TT). TRAENA investigators stopped TT with BZD in 16% of their participants based on pre-established medical criteria. Overall, 25% of the BZD-allocated participants interrupted their treatment, as compared with 2% in the placebo group. In fact, adherence to TT, a treatment lasting usually eight weeks, was overall 80%. Clinicians recommending TT should expect that one in four to five patients abandon their treatment and at least one severe side effect occurring in every 40 treated patients.
Agreements and disagreements with other studies or reviews
As compared with other reviews in the field, this work has some differences and coincidences. Among different reviews, two publications have appeared as systematic reviews in the field since 2007. The first work focused on the effect of BZD (Perez-Molina 2009), whereas the second review (Bern 2007) aimed at guiding the medical community of the USA to deal with T. cruzi-infected individuals (as a response to the new epidemiological challenges discussed in the Background section of this review).
The review by Perez-Molina also included both RCTs (three studies) and non-RCTs (six studies) testing BZD. These authors included two non-RCTs excluded from this review (Streiger 2004; De Castro 2006), because authors did not record any patient-related outcome or had a follow-up period below four years, or both. In contrast, our review includes two studies testing BZD that these authors did not include (one from Venezuela and one from Brazil), providing data on mortality.
The accompanying meta-analyses on that review showed different results compared with our review for a number of reasons. First, those authors recorded and combined data on parasite-related outcomes (i.e. xenodiagnoses and serology status) that we analysed separately. Second, for consistency across outcomes, we treated any "positive effect" as a risk reduction (i.e. reduction of the risk of remaining seropositive), whereas these authors reported outcomes in the inverse way (i.e. BZD increased the rate of negative seroconversion). More importantly, our review treated patient-related outcomes as reported by authors, whereas we recorded outcome data based on our standard definition of "clinical progression". Despite reaching a similar conclusion, numerical results of the review by Pérez-Molina and colleagues would have been different, had they included the two studies that we did include (both with results not favouring BZD).
The review by Bern and colleagues aimed at guiding practice in the USA has a broader scope, covering aspects such as diagnosis, classification and treatment, focusing on nitroderivatives, the drugs available through the CDC. The authors did not present a quantitative synthesis of their data. They stated "benznidazole and nifurtimox are the only drugs with proven efficacy against Chagas disease", citing a book chapter and an expert, narrative review as supporting references. This review offered guidance with levels of evidence for different clinical situations.
Our work adds to this literature in a number of ways: extracting data from all relevant studies, both RCTs and non-RCTs; comparing cohorts after at least four years of follow-up; and recording data on clinical progress using a uniform, explicit definition. We treated similarly all drugs tested as hypothetical TT. We presented a quantitative synthesis of these data, dividing the information by type of study design. Finally, we reported all data extracted from RCTs, and only those from the non-RTCs with lower risk of ascertainment biases.