key: cord-0310050-k8t1qlni
authors: Saraswati, K.; Maguire, B. J.; McLean, A. R.; Singh-Phulgenda, S.; Ngu, R. C.; Newton, P. N.; Day, N. P.; Guerin, P. J.
title: Systematic review of the scrub typhus treatment landscape: Assessing the feasibility of an individual participant-level data (IPD) platform
date: 2021-04-01
journal: nan
DOI: 10.1101/2021.03.31.21254664
sha: 1a03f019899cdf41e2586b112572407c8a634498
doc_id: 310050
cord_uid: k8t1qlni

Background: Scrub typhus is an acute febrile illness caused by intracellular bacteria from the genus Orientia. It is estimated that one billion people are at risk, with one million cases annually mainly affecting rural areas in Asia-Oceania. Relative to its burden, scrub typhus is understudied, and treatment recommendations vary with poor evidence base. These knowledge gaps could be addressed by establishing an individual participant-level data (IPD) platform, which would enable pooled, more detailed and statistically powered analyses to be conducted. This study aims to assess the characteristics of scrub typhus treatment studies and explore the feasibility and potential value of developing a scrub typhus IPD platform to address unanswered research questions. Methodology/Principal Findings: We conducted a systematic literature review looking for prospective scrub typhus clinical treatment studies published from 1998 to 2020. Six electronic databases (Ovid Embase, Ovid Medline, Ovid Global Health, Cochrane Library, Scopus, Global Index Medicus), ClinicalTrials.gov, and WHO ICTRP were searched. We extracted data on study design, treatment tested, patient characteristics, diagnostic methods, geographical location, outcome measures, and statistical methodology. Among 3,100 articles screened, 127 were included in the analysis. 12,079 participants from 12 countries were enrolled in the identified studies. ELISA, PCR, and eschar presence were the most commonly used diagnostic methods. Doxycycline, azithromycin, and chloramphenicol were the most commonly administered antibiotics. Mortality, complications, adverse events, and clinical response were assessed in most studies. There was substantial heterogeneity in the diagnostic methods used, treatment administered (including dosing and duration), and outcome assessed across studies. There were few interventional studies and limited data collected on specific groups such as children and pregnant women. Conclusions/Significance: There were a limited number of interventional trials, highlighting that scrub typhus remains a neglected disease. The heterogeneous nature of the available data reflects the absence of consensus in treatment and research methodologies and poses a significant barrier to aggregating information across available published data without access to the underlying IPD. There is likely to be a substantial amount of data available to address knowledge gaps. Therefore, there is value for an IPD platform that will facilitate pooling and harmonisation of currently scattered data and enable in-depth investigation of priority research questions that can, ultimately, inform clinical practice and improve health outcomes for scrub typhus patients.

Scrub typhus is a zoonotic disease transmitted during feeding by larval mites from the genus 78

Leptotrombidium [1] . It is caused by obligate intracellular bacteria from the genus Orientia [1] . Orientia 79 tsutsugamushi, the main causal species, is prevalent within the so-called "tsutsugamushi triangle"an 80 area bordering Pakistan in the west, Far-East Russia in the northeast, and North Australia in the south 81 to diagnose clinically because it lacks reliable distinguishing features that differentiate it from other 92 febrile illnesses prevalent in the endemic areas, such as typhoid fever, dengue fever and malaria [1, 6] . 93 Therefore, laboratory tests play an essential role in diagnosis. The gold standard for diagnosis is the 94 indirect immunofluorescence assay (IFA), which requires sophisticated laboratory facilities and skilled 95 personnel [6] . Antibiotics, including doxycycline, tetracycline, chloramphenicol, and azithromycin, are 96 the current mainstays of treatment [1, 7] . The treatment efficacy of other antibiotics such as rifampicin 97 has also been studied, although using it in areas with a high prevalence of tuberculosis is not 98 recommended [8] . Delayed response to antibiotic treatment has been reported [9], although the 99 contribution of antimicrobial resistance is uncertain [10] . National treatment recommendations vary or 100 are simply not stated, and the absence of WHO recommendations illustrates the lack of high-quality 101 supportive evidence for optimal treatments [7, 11] . 102 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 1, 2021. identify the most at-risk population, the regional distribution, and the optimal treatments to optimise 106 disease control efforts, which to date rely on limited evidence. Despite its mortality and morbidity 107 impact, scrub typhus is not considered in the current WHO list of Neglected Tropical Diseases (NTD). 108 109 Existing data collected from clinical trials and longitudinal patient observational studies are the sources 110 of our current evidence-based knowledge. However, there have been no recent reviews of the existing 111 literature, and we hypothesise that in the context of scarce data, they are an underutilised source of 112 information to address research priorities and bridge knowledge gaps. This systematic review aims to 113 assess the characteristics of scrub typhus treatment studies and explore the need for and the feasibility 114 of answering knowledge gaps through pooling scrub typhus individual patient data. 115 116

Search strategy 118 We conducted a systematic literature review following the Preferred Reporting Items for Systematic 119

Reviews and Meta-Analyses (PRISMA) statement (S1 Checklist) [14] . The review is registered in the 120

International Prospective Register of Systematic Reviews (PROSPERO) under the reference 121 CRD42018089405. Six electronic databases (Ovid Embase, Ovid Medline, Ovid Global Health, 122

Cochrane Library, Scopus, Global Index Medicus), ClinicalTrials.gov, and WHO ICTRP were searched 123 using a search strategy detailed in the supporting information (S2 Appendix). The search was limited 124 to publication dates from January 1998 up to February 2020 due to the significantly reduced likelihood 125 of gathering individual participant-level data with time since publication [15] . No language restriction 126 was applied at the search stage or initial screening of literature for eligibility. The reference lists of 127 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 1, 2021. ; https://doi.org/10.1101/2021.03.31.21254664 doi: medRxiv preprint systematic reviews and meta-analyses (including Cochrane reviews) identified during the systematic 128 search were screened for any additional studies that met the inclusion criteria to ensure comprehensive 129 inclusion of articles. Melbourne, Australia). When conflicts arose, a third researcher (BM) was consulted. Inclusion was not 147 limited to any minimum quality assessment to ensure a complete assessment of the research landscape. Data Capture) database [16] . Extracted data included details on study design, treatment tested, patient 152 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 1, 2021. ; https://doi.org/10.1101/2021.03.31.21254664 doi: medRxiv preprint characteristics, diagnostic methods, geographical location, outcome measures, and statistical 153 methodology. Data extraction was performed by one of three reviewers (KS or SS or RN). For each 154 publication, a second alternative reviewer (KS or SS or RN) cross-checked every data variable entered 155 by the first reviewer for quality control to identify missed or discrepant information. The third reviewer 156 resolved any data interpretation differences that arose between the two reviewers. 157

For the purposes of this review, we use the term "study arm" to refer both to a study arm from a clinical 159 trial and patient cohorts from observational studies where treatment is administered. across the primary and update searches, we screened the title and abstract of 3,100 articles (Fig 1) . 171

There were 660 full texts assessed for eligibility. From these, 533 articles were excluded, among those 172 357 were case reports or case series. We found 20 papers not in English: four in Korean and 16 in 173

Chinese. We could not access the full text of six papers despite extensive attempts. We included 127 174 relevant studies in the analysis. There were seven interventional studies and 120 observational studies 175 included (S3 Appendix). 176 177 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 1, 2021. ; https://doi.org/10.1101/2021.03.31.21254664 doi: medRxiv preprint The methods used to diagnose scrub typhus varied substantially across the included studies (Table 1) . 210

The most commonly used diagnostic method was enzyme-linked immunosorbent assay (ELISA) 211 (n=65/127, 51%). Indirect immunofluorescence assay (IFA), which is typically considered as the gold 212 no details on the diagnostic methods used. For interventional studies, IFA was the most commonly used 220 diagnostic method (n=4/7, 57%). While for observational studies, the most commonly used diagnostic 221 method was ELISA (n=65/120, 54%). The proportion of studies using PCR and ELISA showed an 222 increasing trend over time (Fig D, S4 Appendix) . is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 1, 2021. There was substantial heterogeneity in treatment, with 14 types of antibiotics administered in the 235 included studies. Doxycycline was the most commonly administered antibiotic (94/141 study arms with 236 8,062 participants), followed by azithromycin and chloramphenicol, which were used in 55 (5,855 237 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 1, 2021. ; https://doi.org/10.1101/2021.03.31.21254664 doi: medRxiv preprint participants) and 20 (1,801 participants) study arms, respectively (Fig 4) . All other antibiotics were 238 administered to less than ten study arms and less than 1,000 participants. The dosage regimen administered in the studies varied substantially. We identified 52 described 243 combinations of drug, dose, and frequency. When the duration of antibiotic treatment was also 244 considered, there were 66 unique combinations of drug, dose, frequency and duration. There were ten 245 dosage regimens reported for doxycycline: one with loading dose and nine with no loading dose. Two 246 study arms administered 200 mg doxycycline followed by a 7-day course of 100 mg twice daily or 247 every 12 hours. Among the reported dosage regimens with no loading dose, the most often used 248 doxycycline regimen was 100 mg twice daily (n=23/35, 64%) (Fig 5) . There were six dosage regimens 249 reported for azithromycin: two with loading dose and four with no loading dose. One study arm used 250 20 mg/kg/dose loading dose, followed by 10 mg/kg/dose for two days. One study arm administered 1 g 251 followed by 500 mg once daily for two days. Within the dosage regimens without loading dose, the 252 most commonly administered azithromycin regimen was 500 mg once daily (n=7/13, 54%) (Fig 5) . For 253 chloramphenicol, there were three identified dosage regimens (administered in one study arm each): 254 500 mg four times daily, 500-1000 mg every six hours, and 25-50 mg/kg/day four times daily. with 106 (10,794 participants), 99 (10,103 participants), 93 (9,688 participants), and 87 studies (7,895 261 participants) reporting these outcomes, respectively (Fig 6, Table in S4 Appendix) . They were also the 262 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 1, 2021. ; https://doi.org/10.1101/2021.03.31.21254664 doi: medRxiv preprint most commonly assessed outcomes among the observational studies with 100 (10,211 participants), 95 263 (9,770 participants), 87 (9,105 participants) and 80 (7,251 participants) studies, respectively. 264

Conversely, clinical response, fever clearance, and fever clearance time were the most commonly 265 assessed outcomes among interventional studies (7 studies each, 644 participants). that even within the tsutsugamushi triangle, information is lackingmany countries did not have any 284 data points, e.g. Myanmar and Pakistan. Substantial heterogeneity was observed in diagnosis methods, 285 antibiotic treatment regimens, and observed outcomes. 286 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 1, 2021. To our knowledge, this is the first systematic review exploring the scrub typhus treatment study 300 landscape. We conducted an extensive search in electronic databases and clinical trial registries. We 301 also included observational studies to ensure a more comprehensive identification of IPD. We flagged 302 ongoing studies to allow evaluation of what kind of data would be available in the future. Although no 303 language restrictions were applied during the search, we could not extract data from articles in Korean 304

and Chinese due to limited translation resources. This review only included peer-reviewed published 305 studies. We also specified the search to start in 1998 due to the small probability of obtaining individual 306 patient data from earlier studies [15] . This process might introduce selection bias, and there may be 307 more IPD available in ongoing investigations and grey literature such as government surveillance data 308 that were not identified in this search. Due to resources restrictions, only one reviewer screened the title 309 and abstract. This might introduce classification error in the selection stage. However, we were able to 310 deploy two reviewers in the full-text screening and data extraction stage. 311 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 1, 2021. This review highlights the heterogeneity in the methods and reporting of scrub typhus treatment studies. 312

There are various treatment regimens administered and outcomes observed. Due to differences in 313 definitions used, it would be challenging to conduct a useful aggregated data meta-analysis without 314 strong assumptions and access to additional data and information. In this situation, curation and 315 standardisation of IPD would help in pooling available data [22, 23] . outcomes that should be measured and reported, as a minimum, in all clinical trials in specific areas of 337 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 1, 2021. In the future, further quality assessments of clinical trials may be useful to identify the source of 342 heterogeneity and identify any barriers to develop more standardised methods and reporting. 343

Formulation of standardised definition, procedures to assess treatment efficacy, and reporting through 344 a consensus-based approach within the research community is an important next step. For data sharing 345 to become normalised, mechanisms need to be put in place to ensure a mutually beneficial relationship 346 between people who re-use the data and those who collect them [23] . Engagement and cooperation with 347 the scrub typhus research community and investment in data infrastructure to implement standards and 348 facilitate data sharing and meta-analysis are of utmost importance to establish an IPD platform. There is still considerable uncertainty about the optimal treatment for scrub typhus, reflected by the 360 highly heterogeneous set of treatment regimens used in the studies reviewed over the last 22 years. 361

Despite the low number of scrub typhus treatment trials published, there is likely a reasonable amount 362 of data available with key information to address this uncertainty. This review further supports the case 363 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 1, 2021. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 1, 2021. ; https://doi.org/10.1101/2021.03.31.21254664 doi: medRxiv preprint

Preferred reporting items for systematic reviews 446 and meta-analyses: the PRISMA statement

The availability of research data declines rapidly with article age

The REDCap 452 consortium: building an international community of software platform partners

the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted

R: A language and environment for statistical computing. R Foundation for 455

456 18. RStudio Team. RStudio: Integrated Development for

Diagnostic accuracy of a loop-mediated isothermal pcr assay for detection of Orientia tsutsugamushi 462 during acute scrub typhus infection

Guidance for individual participant data meta-469 analysis

Beyond open data: 472 realising the health benefits of sharing data

Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: 475 guidelines for reporting observational studies

Statement: updated guidelines for 478 reporting parallel group randomised trials

CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)The copyright holder for this preprint this version posted April 1, 2021. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 1, 2021. ; https://doi.org/10.1101/2021.03.31.21254664 doi: medRxiv preprint