Treatment of Dientamoeba fragilis: A retrospective Finnish analysis of faecal clearance and clinical cure comparing four antiprotozoal drugs

Background Dientamoeba fragilis (DF), the most common intestinal protozoal pathogen in affluent countries, causes asymptomatic or symptomatic infections with severity ranging from mild to disabling. Currently, many studies of treatment options only have small sample sizes and report results that are partly contradictory. Methods Investigating data retrieved from Helsinki University Hospital and Helsinki City patient records, we searched for the most effective antiprotozoal in treating DF infections. To study microbiological clearance of DF, we collected laboratory results of control samples from patients given one of four commonly used antiprotozoals: doxycycline, metronidazole, paromomycin, or secnidazole. For patients symptomatic prior to antiprotozoal treatment, we also retrieved data on clinical outcomes. Furthermore, we explored factors associated with faecal clearance and clinical cure. Results A total of 369 patients (median age 38) and 492 treatment episodes were included. Paromomycin (n = 297) proved effective (clearance rate 83%), showing strong association with faecal clearance (aOR 18.08 [7.24–45.16], p < 0.001). For metronidazole the rate was 42% (n = 84), for secnidazole 37% (n = 79), and doxycycline 22% (n = 32). In pairwise comparisons, paromomycin outdid the three other regimens (p < 0.001, χ2 test). Faecal clearance was associated with clinical cure (aOR 5.85 [3.02–11.32], p < 0.001). Conclusions Faecal clearance, strongly associated with clinical cure, is most effectively achieved with a course of paromomycin, followed by metronidazole, secnidazole and doxycycline. Our findings will be useful in devising treatment guidelines for adults with symptomatic D. fragilis infection.

Research into the treatment of DF infection remains scarce and sample sizes small.A variety of antimicrobials are known to have efficacy against DF, including 5-nitroimidazole derivatives, aminoglycosides, iodoquinol, and tetracyclines [14], with clearance rates ranging from 55% to 100% [15,16].The few comparisons conducted between the drugs [15,[17][18][19]] mostly look at children, sometimes yielding rather contradictory results.
In 2019, we showed DF to be the most common pathogenic intestinal parasite in the Helsinki Metropolitan Area in Finland [2].We found that clinical improvement was associated with faecal clearance, consistent with some other studies [3,8,10,11].Spurred by our results, we now studied a considerably larger patient cohort, covering both microbiological clearances and clinical successes separately for the four antiprotozoal most commonly used in Finland: doxycycline, metronidazole, paromomycin, and secnidazole.

Study outline
We compared doxycycline, metronidazole, paromomycin, and secnidazole in treatment of DF infection within the Helsinki Metropolitan area between January 2007 and December 2016.First, we identified all patients with DF in the laboratory database and selected those given these medications.Next, to analyse faecal clearance and clinical cure, we explored the pre-and post-treatment data.Microbiological clearances were compared pairwise between the drugs, with separate analyses for subgroups of adults and patients aged <18 years.
The study protocol was approved by the Department of Medicine and the HUSLAB laboratory at the Helsinki University Hospital (HUH) and the Department of Social Services and Health Care in the City of Helsinki.According to the Finnish Medical Research Act, a review by the Ethical Committee was not required, as this study did not involve interventions (Ministry of Social Affairs and Health, Finland: Medical Research Act.2015 Jan 14; 1-5).

Data collection and handling of specimens
From the HUSLAB database we retrieved data on patients with DF in microscopy or PCR.Faecal sample for microscopy was fixed in Ecofix® (Meridian Bioscience, Inc., Cincinnati, United States) immediately after defecation and subjected to modified trichrome staining (later trichrome sample) [20].Faecal samples in eNAT tubes were analysed by RT-PCR (Amplidiag® Stool Parasites test, Mobidiag Ltd, Finland) [21].During the study period, faecal PCR was only sporadically employed in routine diagnostics.

Selection of patient population
We included all patients with medical records available between January 2007 and December 2016, except those with 1) concurrent faecal pathogens other than DF; 2) an active gastrointestinal disease diagnosed; 3) other DF medication than any of the four drugs examined, or simultaneous use of multiple antimicrobials; 4) post-treatment control faecal samples tested by methods other than trichrome or PCR; 5) post-treatment control samples not provided 14-90 days after last dose; or 6) only one trichrome control sample delivered (with PCR, a single control sample sufficed).

Background information and clinical data
From the electronic medical records of HUH and the City of Helsinki, we collected information on age, sex, ethnicity, underlying diseases, and recent travel history (within one year before diagnosis) and results of other stool microbe analyses, positive findings leading to exclusion.All pre-and post-treatment symptom data were retrieved; symptoms not mentioned were interpreted as absent.

Faecal clearance and clinical cure
The dosages of all four regimens were recorded.For patients with several medications, each course was included as one episode.
Faecal clearance and clinical cure were evaluated separately and analysed for associations.Clearance was considered successful if at least two post-treatment trichrome samples (or one PCR sample) had been recorded as negative and none of the other possible control samples proved positive (trichrome or PCR).Only episodes with control samples provided 14-90 days after the last treatment day were included.
Clinical cure was defined as complete symptom resolution.In analyses of the rates of clinical cure, we only included patients initially symptomatic.

Statistics
The distribution of continuous variables was described by medians and interquartile ranges.For categorical variables distributions and 95% J.-P.Pietilä et al. confidence intervals were determined.The χ 2 test was used for comparisons; the Fisher's exact test was applied for small sample sizes.Significance levels were Bonferroni-adjusted in multiple comparison situations.Factors associated with faecal clearance and clinical cure were determined by a multivariable binary logistic regression model.Statistical analyses were conducted using SPSS software version 25 (IBM, New York, United States).

Subject group
We identified 2716 patients with a positive DF finding in trichrome or PCR samples recorded in the HUSLAB database over the study period.After exclusions (1811 for missing medical records; 422 for insufficient control sampling; 58 for other pathogens; 48 for no treatment; 8 for no regimens studied here), 369 patients with DF comprised the study population (Fig. 1).

Demographics and microbiological samples
Of the 369 patients, 63% were female and 88% of Finnish ethnic origin.The median age was 38.0 years (with no differences between genders), and 26% were aged <18 years; 54% had travelled abroad the preceding year (Table 1).Of all patients, 88% (323/369) were symptomatic (Supplementary Table 1).For other stool pathogens and apathogenic parasites, see Supplementary Tables 2 and 3. Supplementary Table 4 presents the demographic data in more detail.

Treatment episodes and control samples
The 369 DF patients had been given antiprotozoals for a total of 739 episodes, 492 of which were included in the treatment analyses (for exclusions, see Fig. 1).Paromomycin was used in 60%, metronidazole 17%, secnidazole 16%, and doxycycline in 6% of the episodes.The dosages proved consistent, and for 86% (422/492) of the treatment episodes all data were available (Supplementary Table 4).
The median time was 30 days (IQR 25-43) for first control sample after last treatment day and 49 days (IQR 30-77) for the second (Supplementary Table 4).In subgroup analysis, the first control samples were provided slightly earlier by those given paromomycin than those receiving the other regimens.However, no difference was seen in the median time of the second control samples (Supplementary Table 4).
Of the treatment episodes, 73% (361/492) were reported for adult patients and 27% (131/492) for those aged <18 years; 53% (69/131) of the latter included secnidazole.The regimens used varied by age group, with secnidazole commonly administered to those <18 years, but rarely to adults.Doxycycline, metronidazole, and paromomycin were used almost entirely for adults (Supplementary Table 4).
In our binary regression model, paromomycin treatment was associated with faecal clearance (p < 0.001).Furthermore, metronidazole use and male gender appeared to be associated, but without reaching statistical significance (p-values 0.07 and 0.08, respectively).No association was found for the other factors studied (age, sex, ethnic origin, apathogenic parasite carriage, foreign travel, pre-existing conditions) (Table 2).
Despite significant demographic differences between the various treatment groups (Supplementary Table 4), the results of the best fitted model did not provide any significant evidence of interactions between treatment groups and other variables (Table 2).
In subgroup analyses by age group, a significant association was seen for adults (OR 11.55 [4.08-32.69],p < 0.001).Among those aged <18 years, clearance rate and clinical cure appeared to be associated, but the difference did not reach statistical significance (OR 2.94 [0.87-9.95],p = 0.082).Unfortunately, due to small sample sizes, regression analysis by individual antiprotozoals was not feasible for separate scrutiny of

Discussion
To explore the various antiprotozoals used in treating DF infections, we scrutinized faecal clearance rates and association with clinical cure.We report two major results: 1) in clearing out the pathogen, paromomycin outdoes three other antiprotozoals, doxycycline, metronidazole, and secnidazole; 2) enabled by a large cohort, our previous finding of an association between faecal clearance and clinical cure is further reinforced, especially for adults.

Overall faecal clearance
To our knowledge, our study scrutinises in a single design the largest data to date on Dientamoeba fragilis episodes treated with paromomycin (n = 297), secnidazole (n = 79), and doxycycline (n = 32).The clearance rate for paromomycin exceeds those for doxycycline, metronidazole, and secnidazole (p < 0.001 against each).This confirms previous reports of paromomycin's superiority over metronidazole [16,17,19].
Our lower clearance rates could be ascribed to bigger sample sizes and, partly, timing of controls: we collected specimens 34 and 45 days (medians) after the last dose, while, as exemplified by one study of secnidazole, control sampling has in many research designs been timed for such days as 7 and 14 [8].Indeed, at least with metronidazole among day care-aged children, the clearance rate has been reported to decrease from 62.5% at two weeks to 24.9% at eight weeks after treatment [6] pointing to lower rates in later control samplingand speaking against spontaneous eradication of DF.The decrease in clearance may reflect substantial reinfection rates rather than recrudescence: day-care aged children are at high risk of DF carriage, and the pathogen appears to be easily transmitted within families [25].

Association between faecal clearance and clinical cure
Faecal clearance was associated with clinical cure, our finding according with numerous previous investigations [1,2,[8][9][10][11][12].Our results contradict a double-blinded randomised paediatric study reporting no association between clearance and clinical cure [6].However, methodological criticism points to those results not having been conclusive [26].Nor was any association reported between DF carriage and symptoms in some later studies among children [27][28][29], the data consistent with high asymptomatic carriage rates for this age group [7,30].To further explore potential differences between adults and children, we separately analysed the data within the two age groups.This scrutiny revealed an association between faecal clearance and clinical cure that reached statistical significance among adults, but not patients aged <18 years.
It is noteworthy that for some patients with successful faecal clearance, symptoms only wane graduallya pattern also seen in giardiasis.Since we classified eased symptoms as clinical failure, the clinical cure rates probably remain underestimates.

Background for guidelines
Our data serve as background information for devising international/local treatment guidelines.While no international guidelines care currently available, the Center for Disease Control and Prevention has suggested paromomycin, iodoquinol, and metronidazole as the primary medication [31].In the Netherlands, one clinical guideline favours clioquinol as first-line drug [5] despite its associated central nervous system adverse events.Since the current data have already been implemented into the Finnish guidelines, paromomycin is recommended as first-line treatment, whenever available.This choice is supported by studies showing it to be well tolerated [32].If paromomycin is not available, metronidazole or secnidazole are, despite their lower efficacy, the drugs of choice.Treating patients with no symptoms is not generally warranted.As obvious, despite eradication of DF, not all patients will be cured: abdominal complaints maypartly or fullyhave some other aetiology.On the other hand, early treatment of symptomatic D. fragilis patients may help to avoid more extensive diagnostic examinations.

Limitations
Because of our retrospective design, data on all gastrointestinal pathogens were not available for all participants.However, since prolonged diarrhoea is particularly characteristic of parasitic pathogens, the most common of them (Giardia and Cryptosporidium spp.) were by default covered for all participants using either trichrome samples or faecal PCR test.On the other hand, in prolonged diarrhoea viral aetiology is presumed rare [33]; bacterial aetiology had been excluded for 43%.We could not include a 'no treatment' control group, because such patients usually lacked control samples.As microbiological results may impact symptom evaluation, we recommend the clinical approach adopted by our department: results are given only after the clinical outcome has been recorded.
We acknowledge that fixation-trichrome staining is less sensitive than PCR [34,35].Comparisons between the various regimens and individuals were nevertheless valid, for the same method was used in preand post-treatment analyses.Furthermore, as the clearance rates between episodes with two or three control samples did not differ significantly (data not shown), we chose not to require the recommended three control samples [36] for inclusion.

Conclusions
Our study shows a significant association between faecal clearance and clinical cure, a finding strongly encouraging medical therapy for symptomatic patients.Demonstrating its superior efficacy over metronidazole, secnidazole and doxycycline, our results support the use of paromomycin as the first-line antiprotozoal.Doxycycline should not be considered a drug of choice at all.

Table 1
Demographics of 369 patients diagnosed with Dientamoeba fragilis infection between 2007 and 2016 in Helsinki and Uusimaa Hospital District, Finland.

Table 2
Regression analysis of factors associated with faecal clearance and clinical cure after individual antiprotozoal treatment episodes.All treatment x other variable interactions were non-significant (p > 0.07) in clinical cure analysis.In clearance analysis age × treatment interaction had a p < 0.001 but was left out of our model for clinical insignificance; all treatment x other variable interactions were non-significant (p > 0.07).
a Faecal clearance analysis is missing 13 paromomycin and 3 secnidazole episodes, as there were no data to show, whether the patients were symptomatic or not before medication.bAll patients were symptomatic prior to treatment.c Successful faecal clearance was defined by negative post-treatment samples.

Table 3
Clinical outcome after successful faecal clearance of symptomatic D. fragilis infection using doxycycline, metronidazole, paromomycin, or secnidazole.Test of significance with χ 2 test.(Comparison between episodes with faecal clearance and no clearance.) Only symptomatic patients' treatment episodes were included.Asymptomatic patients were omitted from analyses.The control samples totalled 1110, of which 98.9% (1098/1110) were trichrome stained and 1.1% (12/1110) analysed by faecal PCR.aP-values by 2-sided Fisher's Exact test.