Eligibility criteria and outcome measures adopted in clinical trials of treatments of cutaneous leishmaniasis: systematic literature review covering the period 1991–2015

To document the sources of heterogeneity in outcomes and shortcomings in trial designs reported by previous systematic reviews.


Introduction
The leishmaniases are diseases caused by parasites belonging to the genus Leishmania transmitted to humans through the bite of sand flies: Lutzomyia (New World) and Phlebotomus (Old World) [1]. They are present in 98 countries and cause some 900 000 to 1.3 million cases annually [2,3]. Of the three clinical manifestations (visceral, mucocutaneous, cutaneous), the latter, cutaneous leishmaniasis (CL), is the most prevalent, contributing circa 90% of all cases [3][4][5]. Clinical manifestations of CL, which include most commonly antimonials, pentamidine, miltefosine, amphotericin B, to local treatments using physical therapy or direct application of medications with some level of activity on the parasites in various pharmaceutical preparations [6,8,11]. With very few exceptions, these treatments are more the result of empirical use of existing medications than a planned R&D effort for CL and are often ill-adapted and inconvenient, and some carry significant safety liabilities. Furthermore, the effects of most of these treatments are inconsistently assessed and reported [7][8][9][10][11][12], which means that treatment guidelines are based on weak evidence [1].
To address these fundamental issues of heterogeneity, design and conduct of treatment trials for CL [8,13], which are hampering consistent and effective case management, a guidance document was prepared. Its aim is to provide clinical investigators with guidance for the design, conduct, analysis and report of clinical trials of treatments for CL, while recognising the complexity of the disease, and to enhance the capacity for high-quality trials meeting the requirements of Good Clinical Practice standards [14]. Standardising methods is important to allow between-study comparability and informative meta-analysis, to strengthen the evidence for recommendations on treatment and case management, and ultimately to improve CL case management and control. Inadequate trials may lead to inappropriate conclusions and are an unethical and inefficient use of the limited resources available for neglected diseases such as CL. Consolidating the guidelines to design and conduct of clinical trials for CL will have a positive impact on comparability, interpretation and validity of findings derived of treatment researches, allowing evidence-based decisions and directing patients according to their risk and characteristics. This effort should also be seen in the context of more general ongoing initiatives to improve the quality of reporting of clinical trials Consolidated Standards of Reporting Trials statement (CONSORT) [15][16][17].
Important elements that might account for the apparent heterogeneity of clinical trial outcomes are the characteristics of the treated populations and how treatment effects are measured. We therefore conducted a systematic review of the eligibility criteria and outcome measures adopted in treatment trials of CL conducted during 1991-2015.

Methods
Given that the aim of this systematic review was not to evaluate the efficacy of a specific intervention, the research question was defined in terms of eligibility criteria and cure measures; thus, 'in patients diagnosed with CL (any form) and enrolled in treatment trials (any type of treatment), how do eligibility criteria and outcome measures compare across studies?' Eligibility Inclusion criteria were as follows: (i) studies that include the search terms in title and/or abstract; (ii) clinical trials (all study designs allowed); (iii) original studies investigating the efficacy and safety of treatments (all treatments and routes of administration allowed) of CL (all forms). The search was not restricted by time or language.
We excluded systematic reviews, descriptive studies and analytic studies that only evaluated one group of treatment, trials on clinical manifestations different from CL or complications thereof, co-infection with HIV, and papers reporting insufficient details on methods or papers that were not available in full.

Study selection
After screening out duplicate references, we reviewed all remaining papers in English, Spanish and Portuguese. From the studies meeting the inclusion criteria (see above), two investigators independently extracted and recorded in a specially designed Excel form the study variables (year, language, type of CL, country, clinical phase, randomised, type and number of study arms, sex, age, diagnosis method, specie identification, time of evolution, size, type and number of lesions, anatomical location, pregnant, lactating or childbearing-age women, previous history of CL, previous treatment, laboratory test, electrocardiogram, follow-up time, outcome definitions). To guarantee reproducibility of the data collection, every investigator filled the forms separately. Discrepancies were resolved by consensus or by involving a third person. For papers whose full text was not available, we emailed authors to request it.

Statistical analysis
We report absolute and relative frequencies of the variables of interest by year of publication using 2001 (when the CONSORT guideline was published for the first time).

Description of clinical trials
We identified 75 875 papers published between 1991 and 2015, which became 2238 after removing duplicate articles and 106 after applying the eligibility criteria (Figure 1).

456
given to 36.2% (n = 88; 5497 patients (47.8%)), consisting of: 22.2% (n = 54) local applications of heat or cold or intralesional injections, 14% (n = 34) topical treatments such as creams or ointments; 12.8% (n = 31) combined treatments. Of the remaining groups, seven (2.9%) received no treatment and two (0.8%) did not specify the route of administration (Table 1). The majority of these studies 74% (n = 78) were published after the launch of CONSORT in 2001, enrolling 9011 patients (78%); 50% of the studies had been conducted by 2008 and 50% of patients enrolled by 2011 (Table 1 and Tables 2 and 3 present all inclusion and exclusion criteria identified by this systematic review classed by trial category, taking into account their date of publication (before or after 2001).
Age was an eligibility criterion in 68.9% (n = 73) of the trials. Of these, 32 defined the minimum age for enrolment, which included for 15 studies (20.5%) preschool-aged children (2-5 years old), for seven (9.6%) school-aged children (6-12 years), for six (8.2%) adolescents (13-17 years) and for four (5.4%) adults only (18 years or more). Two studies (2.7%) had only an upper limit which was <18 and ≤60 years. An age range was defined 50.7% (n = 37) of the studies, which was between 2 and 88 years; 5 and 12 years were the most frequent lower limit (7 (9.6%) studies each); 60 years was the most common upper limit (21.9% (n = 16) of studies). Two studies (2.7%) included children without specifying the age range. The proportion of studies enrolling children under 12 years of age increased after the year 2000 ( Table 2).
Type of lesions was considered in 13.2% (n = 14) of studies; all included ulcerated lesions, but while nine (64.3%) also allowed all other types of lesions, five (35.7%) restricted inclusions to ulcers. Of note, one study limited enrolment to ulcerated lesions for only one of the two topical treatment groups (local application of liquid solution composed of thioxolone and 3 mg benzoxonium chloride + cryotherapy) ( Table 2) [93].
Number of lesions was defined in 46.2% (n = 49) of studies, ranging from 1 to 20; 79.6% (n = 39) allowed participants with no more than five lesions, and one 458  required at least 6 [40]. One study restricted inclusion to patients with 'very few' lesions [30] ( Table 2).

Exclusion criteria
Anatomical location was restricted in 29 studies (27.4%), 75.8% (n = 22) of which excluded cases with facial lesions (close to or on the nose, eyes, lips and/or ears) ( Table 3). Pregnant and lactating women were excluded in 74 of the 100 studies that enrolled both men and women, of which 30 (40.5%) evaluated systemic therapies, 28 (37.8%) evaluated local therapies, 15 (20.3%) included both routes of administration, and one study did not specify the route of administration; six required contraception for inclusion of women of childbearing age, and of these, four evaluated systemic therapies, one topical and one both routes of administration (Table 3).
Altered laboratory values and ECG were exclusion criteria in 54 (50.9%, of which 22 (40.7%) specified haematology, liver and renal functions) and 15 (14.2%) studies, respectively. All studies that included ECG had antimonial treatment in at least one arm (Table 3).
Other exclusion criteria such as mucosal involvement, known hypersensitivity to the drugs used in the study, severe underlying disease and/or different clinical manifestations of CL were applied in 55.7% of the studies. Table 4 presents follow-up times and outcome measures according to the categories reported in the different studies and according to whether they had been conducted before or after the publication of the CONSORT statement.

Treatment outcome assessment
Follow-up time (reported in 105 studies) ranged from 1 to 24 months, counting from either the beginning or the end of treatment. Overall, 6 months was the most common duration, adopted in 32 studies (30.5%), of which 65.6% from the end of treatment; 50% of the studies conducted after 2001. Only one study did not report duration of follow-up [67] (Table 4a).

CONSORT guidelines
Of the 78 studies (73.6%) published after 2001, 62 (80%) appear to follow the CONSORT guidelines on reporting on methods for assigning patients to treatment, explanation of rationale, eligibility criteria, interventions, statistical methods. Though not included in the CON-SORT statement, we analysed articles for reporting on ethics and found that 95% of studies did so (Tables 5 and Fig 4).   [113] 462

Discussion
This systematic review provides an overview of how participants were selected and treatment effects were assessed in therapeutic trials of OWCL and NWCL published between 1991 and 2015. Overall, we found 106 trials, which enrolled 11 531 participants in 243 treatment groups. These studies were conducted in 24 countries, which correspond to approximately one-fourth of the countries endemic for CL worldwide [1, 3], being collectively responsible for one-third of the global estimated current burden of CL [124]. This landscape analysis scrutinises the range of criteria used by investigators to select participants and to assess how treatment works in order to account more accurately for the main sources of the heterogeneity in trial outcomes reported by previous systematic reviews [12]. Inconsistent methodologies have been identified as the reason why CL treatment guidelines are based on weak evidence. The present review includes 61% [12], 47% [8] and 80% [11] of papers included in previous reviews, and 53 more.
Drawing generalisable conclusions and making treatment recommendations is no easy task, as CL is not just one disease. The paucity and fragmentation of information make this task all the more difficult. Not counting 38 articles that could not be recovered, just over 100 trials and 11 000 patients studied in almost a quarter of a century is not much for a disease that would have affected some 25 million people during the that period. To this must be added the complexity of this disease in terms of the causative Leishmania species and the resulting differences in the natural history and evolution of disease, as well as response to treatment. Here, 68 and 38 were on OWCL and NWCL, respectively, and the causative species was identified in less than one-fifth of the studies (12 NWCL and eight OWCL). Also, 30 treatments were tested in 243 treatment groups (115 systemic, 88 topical and 31 combined; unknown or untreated for the remaining 9). Together, these elements explain the fragmentation of the evidence produced by these studies and the resulting paucity of effective treatments that can be recommended for use with enough confidence that they will work.
The CONSORT statement introduced a set of criteria aimed at improving the quality of clinical research reports [15][16][17]. As approximately two-thirds of both studies and participants were from articles published after 2001, one would expect quality to have improved since. Of note, half of the studies were conducted by 2006 and half of the participants recruited by 2010, meaning that they tend to concentrate in more recent years. The  [26,36,108] 1075 [43,44,52,53,57,62,64,86,87,91,98,103] 1 An affirmative answer in any of the categories, was considered compliance with the criteria exclusion. 2 Due to that in many of the studies were considered more of one anatomical location, for this exclusion criteria, sample size/total studied (%) was not calculated. 3 A negative answer in this specific item was considered as compliance with the exclusion criterion.  Complete re-epithelisation WITHOUT any activity signs 10 26 Complete re-epithelisation WITHOUT relapse -9 Complete re-epithelisation WITHOUT any activity signs + negative parasitology test Complete re-epithelisation AND negative parasitology test   [25,81,113,119] From beginning of treatment 129/2520 (5.1) [19,39] 2969/8975 (33.1) [59-61, 63, 64, 66, 70, 72, 73, 75, 82, 85-87, 93, 103, 105, 114-117] After cure     [15][16][17] and second (2010) [125] version, shows an increase in the reporting on most criteria such as patient allocation, sample size calculation, treatment allocation and ancillary analysis. In addition, approximately two-thirds and one-third of the papers, respectively, were structured with clear sections and subsections for materials and methods (ethical statement, design, participants, treatments, etc.) and results (baseline data, efficacy, safety outcomes, etc.). Despite this positive trend, overall we found that studies adopted a range of eligibility criteria and outcome measures and that basic requirements in the definitions were not always present or varied across the studies.
As for demographics, the admissible age was specified in just over two-thirds of the studies, in which varying age ranges were defined; the proportion of children enrolled increased after 2001; 94% of studies specified the sex of participants. Equal opportunities were offered to both genders in 95% of cases, limited to non-pregnant, non-lactating women in 74% of these.
Reassuringly, parasitological confirmation was required in all studies but one. However, there were inconsistencies as to elements related to the natural history of disease which would affect response to treatment [4]. Species identification was required in only 19% of studies (11.1% and 32.5% of those on OWCL and NWCL, respectively), but was not conducted on all patients enrolled. Previous duration of illness was defined in just 468 under half of studies, mostly published after 2001. This parameter is important with respect, on the one hand, to the tendency of forms like those caused by L. major and L. mexicana to self-heal over a certain period of time [126,127], and on the other hand, to the severity, size and number of lesions, which could increase with time in non-self-healing species.
It is also very difficult to compare studies in terms of lesion characterisation, as only 13% defined the type of lesion, 28% their size and 46% their numberthe latter would be expected to determine also the choice of the route of administration (systemic or topical). It is estimated that in general, 90% of cases present with fewer than five lesions [128].
When comparing the eligibility and outcome criteria found in this review with those proposed by Olliaro et al.
2013 [14], we found only partial consistency. Entry criteria taken into account in most of the trials were as follows: demographic characteristics (age and sex), parasitological confirmation, and exclusion of pregnant and lactating women as well as patients who had already been treated for the ongoing episode of CL, those with lesions close to mucous membranes and/or on the face, those with hypersensitivity to study drug and those with different clinical manifestations to CL. Other, yet important, criteria that were considered only in a minority of studies were as follows: parasite species identification, previous duration, type, number, location and size of lesions, which were present only in 19%, 43%, 13%, 46%, 27% and 28% of the studies, respectively. Concerning outcome measures, the definition of the primary outcome was described in 90% of studies; for  approximately three-quarters of these, the assessment of cure was based on clinical evaluation (complete re-epithelialisation) as in the guidance document [14]. However, it is difficult to provide a more accurate comparison also because the phase of clinical experimentation (whether pre-registration phases 2 and 3, or post-registration phase 4) was rarely reported in the published papers. In summary, this study provides further evidence of the variable quality of treatment trials in CL; it explores the granularity of the methods and results sections of papers over a 15-year period, and assesses the adequacy of reporting before and after the publication of the CON-SORT statement and its subsequent update. Lack of standardisation, compounded with the small number of trials relative to the magnitude of the disease in its multiple forms, and with the range of treatments tested explains why evidence to inform treatment guidelines is generally weak for CL. While improvements have occurred in the quality of reporting, much remains to be done in adhering to standardised methodologies.
Solving the problem of cutaneous leishmaniasis treatment requires both development of therapeutic alternatives and improvements in the quality of evidence. Standardisation of clinical trial methods for the evaluation of CL is necessary to determine effectivity and safety, to compare studies and strength of the evidence, and ultimately lead to better treatment outcomes.