Ethics statement

Ethics approval was received by London School of Hygiene and Tropical Medicine Observational Research Ethics Committee (LSHTM Ethics Ref 29935). Formal consent was not required as publications reviewed were in the public domain.

The methodology for this systematic review is provided in detail in a published protocol describing the process for the core outcome set development for leptospirosis trials (COS-LEP) project [12]. The review was registered on the COMET initiative and PROSPERO database (CRD42023397461). Preferred Reporting Items for Systematic reviews and Meta-analysis (PRISMA) checklists are provided in S1 Appendix and S2 Appendix.

Study search strategy and selection

A search was conducted of the following databases: the Cochrane Database of Systematic Reviews, Cochrane Central Register of Controlled Trials, MEDLINE, EMBASE, LILACS, Web of Science Core Collection, Clinicaltrials.gov, and OpenSIGLE. Detailed search terms by database and all dates of searches are described in S3 Appendix. For each identified study, forward and backward reference searching using Google Scholar was used to identify additional publications for inclusion.

Any study focusing on human leptospirosis that reported patient outcomes across all age groups globally was included. To maximise identification of all previously reported outcomes and ensure broad international representation while minimising the potential for overlooking clinically significant outcomes, all study design types, both descriptive and analytical, were eligible for inclusion. The incorporation of descriptive studies, including case reports and case series, was intentional as, given the neglected status of leptospirosis and likely limitations of existing literature, inclusion of descriptive study designs was considered essential to ensure capture of all previously reported outcomes. This is in line with established Core Outcome Measures in Effectiveness Trials (COMET) Initiative recommendations [10]. Non peer-reviewed studies such as protocols or registry entries (trial or observational) were included as long as planned outcomes were reported. Full manuscripts were prioritised. Abstract-only publications were included if we were unable to obtain the full publication from the publisher or manuscript author, and as a minimum the abstract included outcomes reported by the study [10].

The search was not limited by language, and translations were done either by the reviewing authors or using machine translation (Google Translate). Further guidance and requirements were followed as specified by the COMET Initiative and PRISMA statement [10,13].

Data extraction

We planned to assess studies in batches chronologically by descending years in 5-year frames until saturation in outcome types was achieved. However, all screened studies were included as we did not reach a stage of outcome saturation.

Studies were screened using the Covidence systematic review software ([href:https://www.covidence.org/]https://www.covidence.org/). Title and abstract review, followed by full text review, was performed by two reviewers (NL and ICH) working independently. A third reviewer (TE) was planned to be consulted if there was disagreement between the first two reviewers, but this did not occur. Agreement at these stages was calculated using Cohen’s coefficient kappa to demonstrate the level of agreement between reviewers [14].

The following data was extracted during the full text review (S4 Appendix): study information including year of publication, institution, and location study was conducted, study design, inclusion/exclusion criteria verbatim; study population – demographics, epidemiological risk factors; interventions used and details; outcomes; outcome measurements, instruments, and/or definitions provided by the authors for each outcome; and the study conclusion verbatim.

Quality assessment and outcome reporting bias

Quality assessment was not performed with the intent to exclude studies from the review, but rather to provide an objective analysis of the study quality of leptospirosis literature reporting outcomes. The primary author (NL) assessed the quality of included studies using the QualSyst tool (S5 Appendix), developed for quality assessment of a variety of study designs [15]. The study was assessed using a template with scoring categories of Yes (2 points), Partial (1 point), No (0 points), and N/A. Summary scores are reported as the percentage proportion of the total possible score (i.e., exclusive of N/A scores), with a larger proportion indicative of a higher level of quality.

Selection of outcomes that are reported in any study is often subject to bias, usually based on the result that is being reported. The Outcome Reporting Bias in Trials (ORBIT) tool has been designed to assess for this selective reporting bias (S5 Appendix) [16]. Studies included in this systematic review were assessed using the ORBIT tool by the primary author (NL) to quantify the amount of selective outcome reporting provided they were observational studies (cross-sectional, cohort, or case-control), trials, or meta-analysis. The goal was to assess and demonstrate the risk of reporting bias in included studies, rather than as a basis for study exclusion.

Data synthesis

All summary statistics were performed on Stata (Stata/IC 15.1). Data describing study design, study population, treatments, and quality/risk of bias assessments were summarized with descriptive statistics. A narrative synthesis was conducted of reported outcomes, which were then gathered based on terminology into group names and then further grouped into domains [10]. As outcome group names and domains were dependent on actual outcomes identified, these were not defined a priori. Outcome domains were initially recorded as reported. Thematically similar domains were combined and mapped to the outcome framework proposed by the COMET Initiative [10]. Outcome groupings and COMET domain mapping was validated by consensus amongst all authors. We considered presenting outcomes by subgroup if there were defined populations targeted by a particular outcome, but this did not apply for this review. Any disagreements regarding outcome classification would have been resolved first by discussion between study authors, and where necessary, at a project steering committee meeting, but this did not apply for this review. Reported outcomes measures were summarized first by grouping thematically, and then categorizing based on the type of outcome. Outcomes were plotted on a sunburst chart to visually illustrate their reporting frequency. Similarly, reported antibiotic treatment types were plotted to a treemap chart.

General characteristics

Consistent with the global distribution of disease, publications from 63 countries across most continents were included (Fig 2). The country with the most publications was India (14.4%; 43/298) followed by Brazil (9.7%; 29/298). Descriptive studies comprised 58.4% (174/298) of included studies, of which case reports formed the majority (81.0%; 141/174) of the group. Observational studies were the second most frequent included study type, of which almost half were of retrospective cohort design (47.1%; 33/70). Interventional studies were the third most frequent study type, with randomised controlled trials being the most frequently used design within this category (79.2%; 19/24).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Number of publications reported Leptospirosis treatment outcomes by country.

Created using QGIS 3.34.11-Prizren and Natural Earth (naturalearthdata.com).

https://doi.org/10.1371/journal.pntd.0013651.g002

Most included studies were published in the last 15 years (67.8%; 202/298). Prior to 2000, the Latin American region produced the largest proportion of included studies (~30%; S7 Appendix). The most common study design was case reports (37%) followed by interventional studies (33%). In subsequent periods, South Asia became the leading region, contributing 21% of included studies between 2000 – 2019, and 25% from 2020 onwards. Case reports remained the most common study design followed by observational studies (respectively 53% and 23% between 2000 – 2019, followed by 35% and 29% from 2020 onwards).

There were 25,876 participants described across all studies exclusive of systematic reviews (to avoid duplication of counts), protocols, registries, modelling, and economic evaluations (Table 1). In studies reporting population descriptions (38.1%; 114/298), half described endemic settings assessing participants drawn from the general population (50.0%; 57/114).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of included studies.

https://doi.org/10.1371/journal.pntd.0013651.t001

Reported outcomes

There were 172 discrete outcomes reported across all studies (Fig 3 and S6 Appendix). The median number of outcomes per study was 3 (range 1–12). There were 55.0% (164/298) studies reporting 1–3 outcomes, 32.9% (98/298) reporting 4–6 outcomes, 7.7% (23/298) reporting 7–9 outcomes, 3.7% (11/298) reporting 10–12 outcomes, and 0.7% (2/298) reporting 13–15 outcomes. Included outcomes were categorized across all five COMET taxonomical core areas.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Sunburst chart demonstrating discrete outcomes and frequency of reporting by number of studies across all COMET core areas.

Made using plotly on Python 3.13.

https://doi.org/10.1371/journal.pntd.0013651.g003

The COMET core area of physiological/clinical was the most frequently reported core area (Fig 4; 74.5%; 222/298), with 14 COMET domains included in this core area. The number of outcome types per domain ranged from 1 – 28. All included outcomes were classified as benefit associated except for one – “Association of inflammatory marker with onset of JHR”. The COMET domain of renal and urinary included 13 outcome types reported in 26.5% (79/298) of studies. The COMET domain of general outcomes had 15 outcome types reported in 20.5% (61/298) of studies. The COMET domain of infection and infestation had 28 outcome types reported in 20.1% (60/298) of studies. The COMET domain of hepatobiliary included 8 outcome types reported in 16.8% (50/298) of studies. The COMET domain of respiratory, thoracic, and mediastinal included 9 outcome types reported in 16.4% (49/298) of studies. The COMET domain of blood and lymphatic system included 14 outcome types reported in 14.4% (43/298) of studies. The COMET domain of nervous system included 7 outcome types reported in 9.1% (27/298) of studies. The COMET domain of cardiac included 4 outcome types reported in 5.0% (15/298) of studies. The COMET domain of gastrointestinal included 6 outcome types reported in 3.0% (9/298) of studies. The COMET domain of immune system included 3 outcome types reported in 2.4% (7/298) of studies. The COMET domain of pregnancy, puerperium, and perinatal included 10 outcome types reported in 1.7% (5/298) of studies. The COMET domain of eye included 2 outcome types reported in 0.7% (2/298) of studies. The COMET domains of ear as well as musculoskeletal/connective tissues both included one outcome type reported in 0.3% (1/298) of studies respectively. In studies published prior to 2000, the most commonly reported outcomes came from the COMET renal domain (37.0%; S7 Appendix). This pattern persisted between 2000 – 2019 (28.4%), however since 2020 infection outcomes have become the most frequently reported (28.6%),

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Sunburst chart demonstrating discrete outcomes and frequency of reporting by number of studies across COMET core area of physiological/clinical.

Made using plotly on Python 3.13.

https://doi.org/10.1371/journal.pntd.0013651.g004

The COMET core area of resource use was the second most frequently reported core area (Fig 5; 59.4%; 177/298). There were three COMET domains included in this core area. The number of outcome types per domain ranged from 2 – 20. All included outcomes were classified as benefit associated. The COMET domain of hospital included 8 outcome types reported in 53.7% (160/298) of studies. The COMET domain of need for further intervention included 20 outcome types reported in 24.8% (74/298) of studies. The COMET domain of economic included 2 outcome types reported in 2.0% (6/298) of studies. Although the COMET domain of hospital outcomes remains consistently reported on, the proportion of outcomes describing the need for further intervention became increasingly more commonly reported in recent years – 24.7% after 2020 compared to 7.4% prior to 2000 (S7 Appendix).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Sunburst chart demonstrating discrete outcomes and frequency of reporting by number of studies across COMET core area of resource use.

Made using plotly on Python 3.13.

https://doi.org/10.1371/journal.pntd.0013651.g005

The COMET core area of death was the third most frequently reported core area (Fig 6; 56.0%; 167/298), with 5 outcome types included in the sole domain which were all classified as benefit associated. The proportion of studies reporting death has remained consistently around 55% throughout the included published years (S7 Appendix).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Sunburst chart demonstrating discrete outcomes and frequency of reporting by number of studies across COMET core area of death.

Made using plotly on Python 3.13.

https://doi.org/10.1371/journal.pntd.0013651.g006

The COMET core area of adverse events was the fourth most frequently reported core area (Fig 7; 11.4%; 34/298), with 8 outcome types included in the sole domain which were all classified as harm associated. The proportion of studies reporting adverse events has remained consistently between 10% - 15% throughout the included published years (S7 Appendix).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 7. Sunburst chart demonstrating discrete outcomes and frequency of reporting by number of studies across COMET core area of adverse events.

Made using plotly on Python 3.13.

https://doi.org/10.1371/journal.pntd.0013651.g007

And finally, the COMET core area of life impact was the fifth most frequently reported core area (Fig 8; 4.0%; 12/298). There were four COMET domains included in this core area. The number of outcome types ranged from 1 – 5. All included outcomes were classified as benefit associated except for “Development of antimicrobial resistance” in the domain of delivery of care. The COMET domain of global quality of life included 1 outcome type reported in 2.0% (6/298) of studies. The COMET domain of personal circumstances included 2 outcome types reported in 1.7% (5/298) of studies. The COMET domain of delivery of care included 5 outcome type reported in 1.0% (3/298) of studies. The COMET domain of cognitive functioning included 2 outcome types reported in 0.3% (1/298) of studies. The proportion of studies reporting adverse events has remained consistently between absent or low throughout the included published years (S7 Appendix). Notably, outcomes in the life impact core area were the most infrequently reported, representing a key gap in the existing literature that future trials should seek to address.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 8. Sunburst chart demonstrating discrete outcomes and frequency of reporting by number of studies across COMET core area of life impact.

Made using plotly on Python 3.13.

https://doi.org/10.1371/journal.pntd.0013651.g008

Outcome measures

There were 92 individual outcome measures which could be extracted from included studies, summarized in Table 2. These were categorized into 13 thematic areas, and then further sub-categorized by whether the measures were categorical or continuous.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. List of outcome measures thematically grouped and categorized by measure type.

https://doi.org/10.1371/journal.pntd.0013651.t002

The median number of measures per thematic area was 3 (range 1–17). “Time measures of disease, organ failure, or treatment” and “renal outcomes” were the two thematic areas that had both categorical and continuous measures, and both contained the greatest number of measures by category. No patient reported outcome measures were identified. The wide variation in measure types across domains highlighted the absence of standardised outcome measurement in human leptospirosis research.

Quality assessment and risk of bias

The inter-rater reliability during title and abstract screening was calculated as a Cohen’s kappa score of 0.71, indicative of substantial agreement. During full text review this was calculated as 0.92, indicative of excellent agreement.

Quality and risk of bias assessments by study category are provided in Table 3. The overall median QualSyst quality score for studies was 0.70 (interquartile range, IQR; 0.50 – 0.83). The median quality scores for interventional studies were the lowest amongst study types, with contributing factors being QualSystem tool category “No” scores in the sections of characterisation of subject/comparator (50%; 12/24), blinding of investigators and participants (62.5%; 15/24, and 66.7%; 16/24 respectively), sample size specification (66.7%; 16/24), and control of confounders (91.7%; 22/24). The median quality for observational studies were the second lowest amongst study types, with contributing factors being QualSyst tool category “No” scores in the sections of sample size specification (35.7%; 25/70), and control of confounders (60.0%; 42/70).

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Quality assessment and risk of bias results by study type.

https://doi.org/10.1371/journal.pntd.0013651.t003

Only 34.4% (103/299) of studies could be analysed using the ORBIT scoring, which included interventional, observational, and systematic reviews with meta-analysis study designs. For interventional studies, 33.3% (8/24) were judged to have no risk of bias, 54.2% (13/24) a high risk of bias, and 12.5% (3/24) a low risk of bias. For studies with an observational design, 37.1% (26/70) were judged to have a risk of no bias, 20% (14/70) a high risk of bias, and 42.9% (30/7)) a low risk of bias. Finally, among systematic reviews with a meta-analysis component, 66.7% were judged to have no risk of bias, 22.2% (2/9) a high risk of bias, and 11.1% (1/9) a low risk of bias.

Antibiotic interventions

Antibiotics were reported as an intervention in 74.5% (222/298) of studies. A summary of all reported antibiotics is included in Fig 9 and S8 Appendix. Cephalosporins (49.1%; 109/222), penicillins (48.6%; 108/222), and tetracyclines (41.4%; 92/222) were among the most frequently reported antibiotic classes. Ceftriaxone, penicillin, and doxycycline were the most frequent antibiotics used from the three aforementioned antibiotic classes respectively. Just under half (44.1%; 98/222) of studies reported using alternative antimicrobial regimens, and only 21 studies (9.5%; 21/222) reported on azithromycin or chloramphenicol – the only two non-β lactam/non-tetracycline antibiotics with existing randomised controlled trial evidence [17,18]. Almost half of the studies (46.8%; 104/222) reported on antibiotics which have no previous existing trial evidence to support their use in the treatment or prevention of leptospirosis [7,9]. Prior to 2000, the most commonly referenced antibiotic class was penicillin (68.2%; S7 Appendix) followed by tetracyclines (50.0%). Between 2000 – 2019, reporting shifted towards penicillin (49.3%) and cephalosporins (46.7%). Since 2020, cephalosporins (70.0%) and tetracyclines (60.0%) have been the most commonly referenced antibiotic classes.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 9. Treemap plot of antibiotics reported by studies.

Abbreviations follow European Committee on Antimicrobial Susceptibility Testing nomenclature except for *: CTR Ceftiaxone, CTA Cefotaxime, CTZ Ceftazidime, CEP Cefepime, CET Cefalothin, *CFP Cefoperzone, CIX Cefixime, *3CS Third generation cephalosporin (not specified), *CSNS Cephalosporin (not specified), PEN Penicllin, AMO Amoxicillin, FLU Flucloxacillin, AMP Ampicillin, DOX Doxycyline, TET tetracycline, OXY Oxytetracycline, *OMC Omadacycline, MIN Minocycline, AZI Azithromycin, CLA Clarithromycin, ERY Erythromycin, *SPI Spiramycin, *MANS Macrolide (not specified), CIP Ciprofloxacin, LEV Levofloxacin, MOX Moxifloxacin, *FLNS Fluoroquinolone not specified, PIT Piperacillin-Tazobactam, AMC Amoxicillin-Clavulanic Acid, AMS Ampicllin-Sulbactam, *PIS Piperacillin-Sulbactam, MER Meropenem, IMI Imipenem, *CANS Carbapenem not specified, VAN Vancomycin, AMI Amikacin, CHL Chloramphenicol, MET Metronidazole, TRS Trimethoprim/sulfamethoxazole, *BLSN β-Lactam not specified, LIN Linezolid [19].

https://doi.org/10.1371/journal.pntd.0013651.g009

Organ-specific and other interventions

Immunosuppressive therapy was reported and specified across 41 different studies; 53.7% (22/41) case reports/series, 17.1% (7/41) interventional, 14.6% (6/41) observational, 9.8% (4/41) reviews, and 4.9% (2/41) trial/study protocol or registries. Additionally, immunosuppressive therapy was reported but not specified in 1 systematic review, 1 modelling study, and 5 case reports. Reported immunosuppressive treatments include methylprednisolone (58.5%; 24/41), prednisolone (31.7%; 13/41), dexamethasone (22.0%; 9/41), hydrocortisone (17.1%; 7/41), corticosteroid (non-specific; 4.9%; 2/41), cyclophosphamide (4.9%, 2/41), and cyclosporin (2.4%; 1/41).

Higher level clinical care or organ supportive therapy was reported by 44.0% (131/298) of studies. Reported interventions included intensive treatment unit (ITU) admission (18.1%; 54/298); respiratory supportive measures including high-flow nasal oxygen (0.7%; 2/298), non-invasive ventilation (4.0%; 12/298), and intubation/mechanical ventilation (19.8%; 59/298); and extracorporeal supportive therapies including plasma exchange (3.4%; 10/298), dialysis/haemofiltration (25.8%; 77/298), and extracorporeal membrane oxygenation (3.4%; 10/298).

Reported blood pressure support strategies included intravenous fluid resuscitation (15.1%; 45/298) and vasoactive inotropes (19.5%; 58/298). Diuretic treatment was an additional renal supportive treatment reported in some studies (4.4%; 13/298). Transfusion of blood products was reported in 12.8% (38/298) studies, and organ transplant in 0.3% (1/298).

Other less commonly reported interventions included health promotion strategies (0.7%; 2/298), N-acetyl cysteine as liver supportive care (0.3%; 1/298), and surveillance-only strategies (0.3%; 1/298).

Quality and risk of bias assessments

The QualSyst tool provides more granular detail in the final quality assessment for design types such as interventional or observational analytical studies [15]. This is because factors such as study population characteristics, blinding, sample size calculation, and control of confounders are more relevant to these study designs and are critical to the accurate measurement and validity of the study’s stated outcomes. Despite this, it is notable that both interventional and observational studies scored poorly on these criteria when assessed. Weaknesses in these areas may have introduced outcome reporting bias in certain studies by affecting which outcomes are reported or how outcomes are described, consequently influencing the outcomes included in this review.

There is remarkable outcome reporting bias across the major study types identified in this review. In our assessment, the proportion of studies at-risk of bias does not decrease as more rigorous study designs are used. Both interventional and observational analytical studies reported either a high or low risk of bias in about two thirds of included studies respectively. It is particularly notable that just over half of included trials were at high risk of outcome reporting bias – though perhaps not surprising given that trials were often single centre and underpowered.

These assessments were undertaken to provide context regarding the characteristics and reporting practices of included studies, rather than to inform study exclusion. Consistent with the objectives of the wider COS-LEP project, all reported outcomes were retained to ensure comprehensive outcome identification. Decisions regarding exclusion or prioritisation will be undertaken in subsequent consensus-building stages of the COS-LEP project [12].

Outcome reporting

We demonstrate in this review the amount of heterogeneity in reported outcomes across the various different types of leptospirosis studies. This inconsistency has been previously highlighted in other healthcare research fields, and poses a significant challenge to synthesizing the necessary information to guide patient management or wider leptospirosis health policy.[22]

We identified that outcomes in the physiological/clinical core area were the predominant efficacy-related outcomes across all study types, particularly in the domains of renal, respiratory, infectious, and general outcomes. The common theme in the outcomes between these domains was the prevention of different manifestation of severe leptospirosis and/or clinical deterioration, highlighting the importance placed on managing these as a treatment goal. This is also reflected in the second most commonly reported core area of resource use – where the top five outcomes (complete medical discharge from health services, length of hospitalisation, ITU admission, the need for dialysis, and ITU length of stay) were all strongly influenced by severe leptospirosis disease.

The inclusion of case reports and series in this review likely led to an increase in reporting of outcomes of priority to healthcare professionals. However, the inclusion of these types of descriptive studies was intentional and pre-defined to ensure comprehensive identification of outcomes reported in the literature consistent with established COMET methodology for outcome mapping in core outcome set development [10,12]. Case reports and series often report rare, unexpected, or clinically important outcomes which may not be captured in other study designs with less clinical focus. Restricting inclusion to analytical studies could have introduced a selection bias towards more investigator-selected endpoints, and risked overlooking outcomes relevant to clinical practice.

Regardless of this limitation, it is notable that outcomes that would be important to other stakeholders such as patients and policymakers, including quality-of-life or treatment cost-effectiveness, were infrequently reported. Despite the growing recognition in clinical trial design of the need for treatment evaluation to extend beyond exclusive clinical endpoints and incorporate patient-reported outcomes of interest, these were concerningly absent from many included studies [23]. Quality of life, an indicator of treatment benefit from the patient perspective, is increasingly becoming an important measure to gauge treatment effect on patients, but was only extracted in this review from two trial protocols [24]. Similarly, there were a limited numbers studies that assessed economic outcomes or cost effectiveness, despite their critical importance to national or institutional governing bodies and decision-makers seeking to implement best practice guidelines. Taken together, these findings suggest that the current treatment evidence base is insufficiently aligned to wider stakeholder priorities. Addressing this gap in patient-reported outcomes will be essential to the COS-LEP development process, and will be explored through a dedicated qualitative study engaging patients and healthcare providers to identify outcomes of greatest relevance.

Harms outcomes accounted for a small proportion of outcomes reported across all studies - 11.7% (35/298). Of these only 17.1% (6/35) were reported in randomised controlled trials. The lack of safety outcome reporting in trials is a wider methodological issue in trial design which can contribute to reporting biases and misrepresentation of an intervention as a safe or preferable option regardless of intervention efficacy [25]. Assessing harms outcomes can provide a broader overview of the risk-benefit of an intervention, for example linking closely to quality of life or treatment adherence [26]. A distinct harms outcome that can occur following antimicrobial treatment of spirochaetal infections such as leptospirosis is a Jarisch-Herxheimer event, a temporary systemic febrile inflammatory reaction [27]. Despite this being of high clinical relevance to health professional and patients, it is only reported in 4.0% (11/298) of studies. Given the lack of reported qualitative studies seeking patient or healthcare provider perspectives on leptospirosis treatment outcomes, exploration of harms outcomes of interest to these groups will be vital to the construction of a future COS.

Given the large number of outcomes identified at this early stage, a key consideration will be focusing the final outcome set on measures that are feasible to implement in low-resource endemic settings. Outcomes and measures selected must be practically achievable within the constraints of local health systems where leptospirosis burden is greatest, including limitations in laboratory infrastructure, clinical staffing, and research data collection capacity. This consideration will be explicitly incorporated into the COS-LEP Delphi and final consensus process.

The policy and public health implications of establishing a COS extend beyond trial methodology. Constructing a standardised outcome set is an essential step towards producing trials that can strengthen the evidence base needed to update WHO treatment guidelines, which currently rely on limited and methodologically weak trial data. Consistent outcome reporting across future trials would support antimicrobial stewardship efforts by enabling reliable comparative evaluation of treatment regimens, and inform procurement and resource allocation decisions by health institutions in endemic settings. The process of defining core outcomes and their measurements would also represent the first formal standardisation of certain leptospirosis-specific outcomes, such as pulmonary involvement and Jarisch-Herxheimer reactions, neither of which has been consistently defined across studies and whose precise characterisation is important for both clinical practice and cross-trial comparability.

Outcome measure

We compile the first comprehensive description of outcomes measures for human leptospirosis disease reported in the literature. The most frequently reported outcome measure category includes 25 chronological time measures categorized across continuous and categorical measures. Even within this category there are varying lengths of time. For example, the measure of severe disease manifestation within a pre-defined time period included ≤ 48 hours, ≥ 48 hours, and  ≤ 21 days following treatment. Depending on the organ system involved, an outcome may not be captured within these time periods. This discrepancy highlights the need to standardise measures used to enable better comparability of results.

We demonstrate that even in the most frequently reported outcome domain there is a wide inconsistency in types of measure. For example, within the domain of renal and urinary outcomes alone there are least 12 different measures consisting of routinely available diagnostics to more experimental tests. We did not identify any patient reported outcome measures in included studies.

Multi-organ involvement is a critical sequela of leptospirosis and a major cause of disease severity.[5] As previously noted, a precise definition of this outcome will be essential in future trials. We report several measures used to capture this outcome, ranging from mortality alone to composite parameters. Composite outcome measures can more accurately capture disease manifestation across one or multiple organ systems, and the use of established scoring criteria offers the advantage of validated, semi-quantitative systems. For example, the modified sequential organ failure assessment score is an established criteria that has been previously reported to guide leptospirosis therapy, and which offers pragmatic and clinically relevant endpoints that would be useful in large Phase III trials [28–30].

Acute kidney injury (AKI) is a common manifestation in leptospirosis disease [31]. Clinical endpoints in trials focused on AKI have commonly relied on measuring small short-term changes in creatinine levels or composite endpoints of long term major adverse kidney events, with several limitations in both approaches [32]. Previous leptospirosis studies have used a range of renal outcome measures to capture AKI events, including biochemical thresholds or the presence of oliguria/anuria, all generally aligned with internationally accepted criteria [33–35]. In addition to the immediate disease process, progression to chronic kidney disease (CKD) after acute leptospirosis infection has been previously documented [36]. According to KDIGO guidelines, CKD is defined as abnormalities of kidney structures or function, present for a minimum of 3 months, with implications for health [37]. Future trial will need to incorporate short and long-term renal outcomes to comprehensively capture this domain.

Other outcomes measures of organ involvement are less well defined and will require a consensus on both measurement and stratification of clinical importance. Of the remaining organ system outcome domains reported in the Physiological/Clinical core area, only 5 had defined associated outcome measures. Pulmonary complications, ranging from acute respiratory distress syndrome (ARDS) to pulmonary haemorrhage, represent severe manifestations associated with an elevated mortality risk [38]. Thrombocytopenia is another well-documented complication linked to clinical bleeding risk, though platelet count thresholds have varied across studies [39]. Reversible hepatic injury is a hallmark of severe leptospirosis, but while associated with an increased mortality risk in observational studies, it rarely progresses to liver failure [40]. Pancreatitis, the primary gastrointestinal complication with reported outcome measures, is an emerging syndrome of interest that often necessitates intensive care management [41]. Neurological manifestations are diverse, ranging from a meningoencephalitis to paralysis, and are also recognized complications of leptospirosis [42].

Treatment strategies

This systematic review provided an opportunity to compile the first comprehensive list of antimicrobial, immunosuppressive, and other organ-supportive treatments used in the management of human leptospirosis. These data are invaluable to inform discussions relating to a control arm regimen in future clinical trials, and the design of a systematic assessment of treatment options for optimal use. Importantly, these findings are presented to contextualise outcome variability and research practices, rather than to evaluate comparative effectiveness.

Current, although now somewhat outdated, World Health Organisation (WHO) guidance recommends that antibiotic choice be guided by timing of presentation and disease severity [43]. For mild disease, recommended regimens include oral antibiotics such as amoxicillin, ampicillin, doxycycline, or erythromycin. For severe cases, intravenous penicillin is preferred. Third generation cephalosporins or quinolones are noted as potential alternatives. Despite existing previous published international and national guidance on choice of antimicrobial, this review demonstrates that this is not reflective of real-world use of treatments. This is possibly due to a wide range of factors including availability, drug quality assurance, suitable health facilities to dispense and administer medication, direct and indirect costs, wider antimicrobial stewardship concerns, and uncertainty of diagnosis. We could not reliably extract information on treatment regimen composition, and although many listed antibiotics were not given as monotherapy, the dosing regimen information was incomplete and variable between studies. Although the evidence of antimicrobial resistance in human leptospirosis remains unclear due to methodological issues with antimicrobial susceptibility testing, it is generally accepted that there are no emerging resistance issues with the organism [44,45]. Consequently, data on resistance patterns was not systematically reported in included studies, if at all.

Cephalosporins, penicillins, and tetracycline antibiotic classes accounted for the majority of reported antimicrobial regimens in this review - in line with WHO recommendations as well as national guidance from countries where leptospirosis is a public health concern [46–48]. However, there is great variation of within-class choice, method of delivery, and dosing. We show that over half of the studies report the use of alternative antibiotics, and nearly 20% describe therapies with no supporting trial evidence or to which leptospirosis may demonstrate a potential intrinsic resistance [7,45]. These findings suggest an incoherence in antimicrobial strategy at all levels, with profound implications for standardising clinical guidelines, resource planning, and wider antimicrobial stewardship considerations.

Other non-antimicrobial strategies, such as organ-supportive therapies and non-pharmacological interventions, were reported with varying frequency. Renal replacement therapy was among the most frequently described, reported in nearly 20% of studies. Although findings from observational studies suggest a possible survival benefit, this has not been confirmed in randomised clinical trials [49]. Respiratory supportive strategies such as mechanical ventilation often employing lung-protective ventilation were reported at a similar frequency, but likewise lacks validation from trial evidence.

Despite the limited evidence supporting immunosuppressive treatment for leptospirosis there were several studies, which also included randomised controlled trials, that explored the use of different immunosuppressive agents [8]. Similar to antimicrobial regimens, information on dosing and choice of agent was variable between studies.

Limitations and deviations from protocol

Despite our methodology not limiting study inclusion by language, there may still exist studies not published or indexed in English which will have been missed. We have however demonstrated data capture across 63 countries. There may have been unintended reviewer bias introduced during screening of publications, particularly more narrative-type studies, however we demonstrate a high degree of inter-reviewer agreement in our analysis.

Machine translation was used for a small proportion of non-English studies. While this may have introduced inaccuracies in the interpretation of nuanced clinical terminology, the majority of included studies were published in English (87.9%). Furthermore, data extraction focused on discrete outcome names, definitions, and study characteristics rather than interpretative content, limiting the potential impact of translation errors on findings.

There were no qualitative studies identified during this review, although this was expected in our published protocol. We did not perform subgroup analysis as per our protocol as the majority of studies either did not describe this in sufficient detail or omitted this information. We performed a limited comparison of outcomes and treatments by region and publication time period. These questions, along with more granular analyses of treatments and patient characteristics at the individual level, require further investigation through dedicated descriptive or analytical multi-national epidemiological studies.