The Effect of Helminth Infections and Their Treatment on Metabolic Outcomes: Results of a Cluster-Randomized Trial

Abstract Background Helminths may protect against cardiometabolic risk through effects on inflammation and metabolism; their treatment may be detrimental to metabolic outcomes. Methods In a cluster-randomized trial in 26 Ugandan fishing communities we investigated effects of community-wide intensive (quarterly single-dose praziquantel, triple-dose albendazole) vs standard (annual single-dose praziquantel, biannual single-dose albendazole) anthelminthic treatment on metabolic outcomes, and observational associations between helminths and metabolic outcomes. The primary outcome, homeostatic model assessment of insulin resistance (HOMA-IR), and secondary outcomes (including blood pressure, fasting blood glucose, lipids) were assessed after 4 years' intervention among individuals aged ≥10 years. Results We analyzed 1898 participants. Intensive treatment had no effect on HOMA-IR (adjusted geometric mean ratio, 0.96 [95% confidence interval {CI}, .86–1.07]; P = .42) but resulted in higher mean low-density lipoprotein cholesterol (LDL-c) (2.86 vs 2.60 mmol/L; adjusted mean difference, 0.26 [95% CI, −.03 to .56]; P = .08). Lower LDL-c levels were associated with Schistosoma mansoni (2.37 vs 2.80 mmol/L; −0.25 [95% CI, −.49 to −.02]; P = .04) or Strongyloides (2.34 vs 2.69 mmol/L; −0.32 [95% CI, −.53 to −.12]; P = .003) infection. Schistosoma mansoni was associated with lower total cholesterol (4.24 vs 4.64 mmol/L; −0.25 [95% CI, −.44 to −.07]; P = .01) and moderate to heavy S. mansoni infection with lower triglycerides, LDL-c, and diastolic blood pressure. Conclusions Helminth infections improve lipid profiles and may lower blood pressure. Studies to confirm causality and investigate mechanisms may contribute to understanding the epidemiological transition and suggest new approaches to prevent cardiometabolic disease. Clinical Trials Registration ISRCTN47196031.

Noncommunicable diseases accounted for 72.3% of deaths globally in 2016 [1]. Ischemic heart disease and cerebrovascular disease are the highest contributors to this mortality and morbidity [1]. Metabolic disorders such as type 2 diabetes mellitus (T2D) and dyslipidemia are important cardiovascular disease risk factors. In 2017 an estimated 425 million people had diabetes worldwide and the number is projected to increase-disproportionately in low-and middle-income countries (LMICs)-to 629 million by 2045 [2]. Dyslipidemia, which results in atherosclerosis, is the leading risk factor for myocardial infarction and stroke [3,4].
Chronic, low-grade, obesity-driven inflammation has been implicated in the etiology and progression of insulin resistance and T2D [5]. Atherosclerosis results from a chronic inflammatory state mediated by macrophages in the vascular subendothelium combined with high levels of lipids in the systemic circulation [6]. Lipids are utilized by and involved in mediation of immune processes; therefore, inflammation may influence lipid levels in the circulation [7]. Modulation of these inflammatory processes may delay the development of T2D and atherosclerosis.
More than a billion people are helminth-infected worldwide, predominantly in LMICs [8]. Helminths have coevolved with humans over millions of years so that they survive in the human host but mostly cause only subtle harm. To ensure survival, helminths modify type 2 and regulatory immune responses in the human host [9]. With these immunomodulatory properties, helminths may be able to annul the inflammation that results in metabolic disorders and therefore confer protection [10]. Studies in experimental animals support this hypothesis. Mice infected with Schistosoma mansoni [11], Nippostrongylus brasiliensis [12], and Litomosoides sigmodontis [13] had decreased insulin resistance and improved glucose tolerance compared with uninfected mice. Infection of mice with S. mansoni cercariae [14] and intraperitoneal inoculation of S. mansoni eggs [15] resulted in lower blood lipid levels than in uninfected/ uninoculated controls. In mice, S. mansoni infection also reduced development of atherosclerotic lesions [14], and ES-62, a product from the filarial nematode Acanthocheilonema vitae, caused a 60% reduction in aortic atherosclerotic lesions [16].
In humans, published studies investigating the effect of helminths on metabolic outcomes are few and observational, except for a single trial. In China, history of Schistosoma japonicum infection was associated with lower levels of fasting blood glucose, glycated hemoglobin (HbA1c), and insulin resistance, and lower prevalence of diabetes and metabolic syndrome [17], whereas chronic infection with this helminth was associated with lower serum triglycerides and total cholesterol [18]. Among Australian Aboriginals, Strongyloides stercoralis infection was associated with a lower likelihood of diabetes [19], and treatment of Strongyloides worsened glycemic status [20]. Among the Tsimane of the Bolivian Amazon, helminth infection was inversely associated with serum lipid levels [21]. Chronic infection with the trematode Opisthorchis felineus was associated with lower serum cholesterol levels and diminished atherosclerosis in a postmortem study in Russia [22]. In Indonesia, infection with soil-transmitted helminths (STHs) was associated with lower insulin resistance and body mass index (BMI) [23]. The only trial to date, conducted in Indonesia, showed no effect of albendazole treatment for STHs on insulin resistance, BMI, or waist circumference overall, but showed increased insulin resistance in individuals infected with STHs at baseline [24].
Observational studies cannot demonstrate causality and are prone to bias from unmeasured confounders. The only previous trial was done in a setting with high prevalence of STH. However, other parasitic infections such as S. mansoni may have a different impact on metabolic outcomes because of the different location of adult worms and inflammatory and regulatory elements of their life cycles. We therefore aimed to investigate, in a setting of high S. mansoni prevalence, the effect of helminths and their treatment on metabolic outcomes in individuals aged ≥10 years. We also conducted observational analyses to assess cross-sectional associations between helminths and metabolic outcomes.

Study Setting and Design
This was a 2-arm, cluster-randomized trial of community-wide intensive vs standard anthelminthic treatment conducted in 26 Lake Victoria fishing villages of Koome subcounty, Mukono district, Uganda (ISRCTN47196031). This population is ethnically diverse with the majority from central Uganda (of Bantu ethnic origin). There is some acculturation as different ethnic groups live together and use Luganda as the lingua franca. The main economic activity is fishing with a minority involved in agriculture. Villages are well defined, geographically separate, and close to the lake. The residents live in temporary housing made from wood (locally obtained) and roofed with iron sheets or plastic. Houses are built close together. The island communities trade with mainland communities (both nearby and far) in fish and general merchandise.
The protocol has been published elsewhere [25]. In brief, 13 villages were randomized to intensive and 13 to standard intervention. Intensive treatment comprised praziquantel 40 mg/ kg single-dose administered using an extended height pole (which extended treatment to include preschool age groups) and albendazole 400 mg triple-dose, both quarterly. Standard treatment comprised single-dose praziquantel (40 mg/kg) annually and single-dose albendazole (400 mg) biannually. Trial interventions started in September 2012. Originally, the trial was designed to investigate the effects of 3 years' intervention on allergy-related outcomes and helminth-related morbidity, but interventions were extended for an extra year, allowing us to assess their impact on metabolic outcomes.

Randomization and Masking
This was an open trial. Villages were randomized in a 1:1 ratio using restricted randomization to ensure balance for village size, previous community-wide anthelminthic treatment, and distance from the subcounty health center [25].

Exposures and Outcomes
The exposures of interest were anthelminthic intervention (for the trial) and helminth infections (for observational analyses). The primary outcome was insulin resistance measured by the homeostatic model assessment of insulin resistance (HOMA-IR) [26], calculated as (fasting serum insulin [μU/ mL] × fasting glucose [mmol/L]) / 22.5. Secondary outcomes were fasting blood glucose, HbA1c, fasting serum lipids (triglycerides, total cholesterol, high-density lipoprotein cholesterol [HDL-c], low-density lipoprotein cholesterol [LDL-c]), blood pressure, BMI, waist circumference, waist-to-hip ratio, and (for the trial) helminth infection status.

Procedures
The effects of helminths and anthelminthic intervention were assessed in a household survey conducted in all study villages after 4 years of intervention. We selected 70 households per village by simple random sampling using Stata version 13.0 software (StataCorp, College Station, Texas) (Supplementary Methods). Permission for household participation was sought from each household head or another adult in the household if the head was absent. Household members aged ≥10 years were invited to participate.
After obtaining written informed consent from each household member (and assent from individuals aged 10-17 years), a questionnaire was administered. Data were collected on sociodemographic characteristics, diet, exercise, lifestyle, and personal and family history of diabetes and hypertension. Participants' blood pressure (BP), weight, height, and waist and hip circumference were measured using standardized procedures (Supplementary Methods). Peripheral venous blood was collected after an overnight fast for measurement of the primary and secondary outcomes (Supplementary Methods).
Each participant was requested to provide 1 stool sample, from which duplicate slides were made and examined independently by 2 experienced technicians using the Kato-Katz method. Stool real-time polymerase chain reaction (PCR) was also used to detect S. mansoni, hookworm (Necator americanus), and S. stercoralis infection (Supplementary Methods).

Statistical Methods
For this survey, we planned to recruit 1950 participants (Supplementary Methods). Statistical analyses were done using Stata version 13.0 software; all analyses allowed for within-cluster correlations. The trial analysis was by intention-to-treat and measured cluster-level differences. For continuous outcomes, means were calculated for each village (cluster-specific means) and the mean of these within each study arm was used as a summary measure of the outcome in that study arm. The t test was used to compare cluster-specific means between arms, with corresponding 95% confidence intervals (CIs) computed. Continuous outcomes with skewed cluster-specific mean distributions were log-transformed before analysis and results back-transformed to give geometric mean ratios. For binary outcomes, we calculated risk ratios by dividing the mean of the cluster-specific proportions in the intensive arm by that in the standard arm. P values were calculated from t tests comparing cluster-specific proportions, and CIs using a Taylor series approximation to estimate standard errors. For all outcomes, the effects of the intervention were additionally adjusted for age and sex using a 2-stage approach [27].
The individual level observational analysis was performed to assess helminths as risk factors for metabolic disease. Potential confounders were identified based on a causal diagram (Supplementary Figure 1). Crude and adjusted associations were estimated using linear regression models fitted for each worm separately. Stata "svy" commands were used to allow for clustering of participants within villages, and for the nonself-weighting survey design due to variable village sizes. For adjusted analyses, only risk factors/confounders crudely associated with the outcome with P value ≤ .15 were included in final models. No corrections were made for multiple testing.

Ethical Considerations
Ethical clearance was granted by the Uganda Virus Research Institute Research ethics committee (reference GC/127/17/01/573), the London School of Hygiene and Tropical Medicine (reference 9917), and the Uganda National Council for Science and Technology (reference HS 2185). Permission to conduct the work was granted by community leaders in all study villages.

RESULTS
Between April and November 2017, 70 households were randomly selected from each village except in 2 villages with fewer households where all were selected. Of eligible households and eligible individuals in those households, 71.3% (1276/1790) and 87.0% (1898/2181), respectively, agreed to participate. We examined 1817 of 1898 (95.7%) participants, and 1686 of 1898 (88.8%) provided blood samples after an overnight fast. Numbers of participants providing data were balanced across trial arms ( Figure 1).
Characteristics of participants are shown in Table 1. This is a young population with a mean age of 31.5 years (standard deviation [SD], 11 years) considering that individuals aged <10 years were excluded. There are more males than females and the main occupation is fishing; the majority (1309/1898 [69%]) reported daily contact with the lake. Activity levels were high: 920/1898 (48.4%) reported vigorous physical activity at least weekly, with a mean number of days of 2.8 (SD, 2.7 days). Diet mainly consists of fish (eaten on average 4.8 days a week) and is low in fruit (on average 1 day a week). The majority of participants (1218/1898 [64.2%]) reported having lived in the village in which they were surveyed throughout the 4-year intervention period. Only 283 of 1898 (14.9%) had ever undergone BP measurement and 80 of 1898 (4.2%) had previously had blood sugar levels tested.
There was no evidence of a difference in HOMA-IR between the study arms (Table 3). However, there was some evidence of a difference in mean LDL-c, higher in the intensive arm (2.86 vs 2.60 mmol/L; adjusted mean difference, 0.26 [95% CI, −.03 to .56]; P = .08). No differences were seen between trial arms for the other metabolic parameters (Table 3). Further analysis of the metabolic outcomes categorized as binary (diabetes, impaired fasting glucose, hypertension, obesity, and metabolic syndrome) showed no differences between trial arms (Supplementary Table 1).
Four serious adverse events (SAEs) were reported in the first 3 years of the anthelminthic intervention [28]. In the fourth year of the intervention, no SAEs were reported.

DISCUSSION
This is the first trial to investigate effects of an anthelminthic intervention on metabolic outcomes in an S. mansoni-endemic, high-transmission setting. We found no effect of 4 years of intensive vs standard anthelminthic treatment on insulin resistance, but there was a modest increase in LDL-c in the intensive vs standard arm. Consistent with this, there was strong observational evidence of association between S. mansoni infection and lower levels of total cholesterol and LDL-c. The lowest levels of triglycerides and diastolic BP were observed among participants with heavy S. mansoni infection and the lowest levels of LDL-c among those with moderate S. mansoni intensity. Strongyloides infection was also inversely associated with LDL-c. A doseresponse relationship was observed between increasing numbers of helminth species detected and lower cholesterol levels.
The effect of the intervention on LDL-c and consistency of this with our cross-sectional observations strongly suggests a causal link between helminth infection and lipid profiles. Our results, in a setting dominated by schistosomiasis, accord with findings in experimental animals [14,29] and with associations observed between intestinal nematode infection and lipids in humans [30,31]. One possible mechanism is direct consumption of lipids by helminths. Schistosomes do not synthesize cholesterol and rely on the host for their supply [32]. However, the effect of schistosomes is likely not limited to direct consumption,   with immunological effects implicated. La Flamme et al showed that chronic exposure to S. mansoni eggs reduced serum cholesterol in mice with changes due to egg-induced increased uptake of LDL by macrophages [29]. It has been suggested that helminths might modulate production of naturally occurring antibodies to cholesterol [33]. Helminths resident in the lumen might hinder lipid absorption from the gut directly. In addition, there is evidence that helminth infection influences the intestinal microbiome [34], which also impacts blood lipids [35]. Studies investigating the influence of helminths on the microbiome and how that relates to lipid metabolism would provide a deeper understanding of the underlying mechanisms. An unanticipated finding was the association between heavy S. mansoni infection intensity and lower diastolic BP. Essential hypertension accounts for 90% of hypertension cases. The etiology is not well understood, but it is hypothesized that   inflammation may cause essential hypertension [36]. Two pathways are proposed. First, inflammation of the renal interstitium results in monocyte/macrophage infiltration of renal perivascular fat, oxidative stress, and loss of peritubular capillaries. This may lead to medullary hypoxia, impaired sodium excretion, and consequently, hypertension [37]. Second, inflammation results in monocyte/macrophage infiltration of arterial walls leading to arterial stiffness. Arterial stiffness triggers angiotensin II, which results in hypertension. Helminths may influence these pathways via their immunomodulatory properties. Helminths may also indirectly influence blood pressure through effects on the intestinal microbiome. Gut microbiota digest fiber and release short-chain fatty acids which, on absorption into the circulation stimulate the release of renin, a major component of the renin angiotensin aldosterone system [38]. We recommend studies to further explore the relationship between helminth infection and BP. We were surprised to find no impact of anthelminthic treatment or of helminths on insulin resistance, given the strong evidence of effects from animal models and evidence of increased insulin resistance following STH treatment among participants in the Indonesian trial. Perhaps the most probable explanation lies in the transmission dynamics of schistosomiasis in this transmission "hot-spot" setting; and in the biology of schistosomes, as distinct from STH. In this trial, we observed a reduction in helminth prevalence in the standard arm over time, as well as in the intensive arm [25,28], and this may have weakened any chance of seeing a differential effect of trial interventions on the metabolic outcomes. As well, single stool Kato-Katz analysis has low sensitivity in detecting S. mansoni infection. As previously reported, the more sensitive urine circulating cathodic antigen used after 3 years of intervention suggested persistent (albeit reduced intensity) infection in >80% of participants in both trial arms, presumably due to ongoing transmission and reinfection. Moreover, schistosomes differ from STH in that they reside in the vasculature and treatment does not result in immediate expulsion from the body. Immunological changes induced by helminths persist even after treatment of the worms. Schistosoma egg antigen elicits reduced insulin resistance in an animal model in the absence of live worms, and Schistosoma egg antigen is persistent-liver egg counts were almost unchanged 26 weeks after treatment in a mouse model [39]; in humans, antibody to egg antigen was detectable 8 years after curative treatment of schistosome infection [40]. Therefore, effects of schistosome infection may persist after treatment of adult worms with praziquantel. Furthermore, animal models investigating effects of helminths on insulin resistance have tended to use obese models. By contrast, fishing community populations are generally young, slim, and active and therefore at low risk of metabolic disorders.

Table 2. Effect of Intensive Versus Standard Anthelminthic Treatment on Helminth Prevalence in the Metabolic Survey of the Lake Victoria Island Intervention Study on Worms and Allergy-Related Diseases Cluster-Randomized Trial: Cluster-Level Analysis (n = 1853)
Our results in this setting of high schistosomiasis transmission, namely strong observational associations and supportive suggestion of a treatment effect for lipid profiles, contrasting with lack of association or treatment effect for insulin resistance, could be explained if the former depends predominantly upon active schistosome infection and the latter upon exposure to schistosome antigen.
Strengths of our study include the robust trial design, large sample size, and intense helminth exposure. However, ongoing helminth transmission and cluster-level randomization limited our ability to determine effects of helminth elimination and more focused, individually randomized trials would be useful Abbreviations: aGMR, adjusted geometric mean ratio; BP, blood pressure; CI, confidence interval; GM, geometric mean; GMR, geometric mean ratio; HDL-c, high-density lipoprotein cholesterol; HOMA-IR, homeostatic model assessment of insulin resistance; LDL-c, low-density lipoprotein cholesterol. a Adjusted for age and sex. b Geometric means used because of skewed cluster-specific means. c GMRs used because of skewed cluster-specific means.  Adjusted for the following variables: HOMA-IR (age, sex, occupation, residence, lake contact, treatment for worms, treatment with coartem, body mass index [BMI], family history of diabetes mellitus, maternal and paternal tribe); triglycerides (age, sex, occupation, residence, diet, treatment for worms, family history of obesity, paternal and maternal tribe); total cholesterol (age, sex, occupation, residence, diet, exercise, family history of obesity, paternal and maternal tribe); LDL-c (age, sex, occupation, diet, exercise, treatment for worms, treatment with coartem, parental tribe); HDL-c (age, sex, occupation, residence, diet, treatment for worms, treatment with coartem, parental tribe); diastolic BP (age, sex, occupation, residence, lake contact, treatment for worms, paternal tribe, BMI). There was no evidence of association between helminth infection and fasting blood glucose, glycated hemoglobin, systolic BP, BMI, waist circumference, or waist-to-hip ratio (see Supplementary Table 2).
to assess this further. While multiple outcomes were assessed, risking false-positive findings, the coherence of results on lipid profiles within the study, and when compared to other studies, is persuasive. Finally, we cannot rule out a direct effect of praziquantel and albendazole on metabolic outcomes, although no prior evidence suggests this. Taken together, our findings have important implications. They provide an understanding on how anthelminthic treatment may impact metabolic health, thus increasing the burden of noncommunicable diseases and accelerating the epidemiological transition. Helminth infections may be beneficial to lipid metabolism and BP control in humans. Identification of potentially protective interactions and their underlying mechanisms is important for the development of future therapies.

Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.