Airborne particulate matter exposure in male sugarcane workers at risk for chronic kidney disease in Guatemala

Abstract

Background

There is an international epidemic of chronic kidney disease of unknown cause (CKDu) in agricultural working populations. Particulate air pollution is a likely contributing factor in populations at risk for CKDu, but there is little personal breathing zone data for these workers.

Methods

We collected 1 to 3 personal breathing zone particulate matter <5 microns (PM₅) gravimetric measurements in 143 male sugarcane harvesters over 2 seasons and concurrent ambient samples using personal sampling pumps and cyclone inlets as a sampling train. Due to very high concentrations observed during a pilot of these methods, personal breathing zone sampling duration was set to 4 h, beginning either at the start of a work shift (AM) or delayed for 4 h (PM). To obtain full-shift exposure concentrations we calculated 8-h time-weighted average (TWA, in µg/m³) estimates of each worker’s full-shift personal breathing zone PM₅ exposure concentration by averaging their individual monitored concentration with the median concentration of the unmonitored AM or PM segment from all workers that day to obtain an 8-h TWA.

Results

Median full-shift personal TWA PM₅ concentrations were 449 μg/m³ (range 20.5 to 1,930 μg/m³), which were much higher than ambient concentrations in these fields (median 136, range 22.5 to 2,360 μg/m³). These findings document very high personal breathing zone PM₅ exposure in workers at risk for CKDu: median concentrations for all workers were 3.5 (range <1 to 33.6) times as high as concurrent ambient concentrations.

Significance

These findings suggest that ambient measurements of particulate matter are insufficient to estimate personal exposure in this population and that personal breathing zone monitoring should be used to fully explore air pollution as a risk factor for CKDu. Given that particulate matter from this source likely has multiple hazardous constituents, future research should focus on characterizing all constituents and explore associations with biomarkers of kidney injury.

Introduction

Agricultural workers in Latin America and Asia experience a disproportionate burden of chronic kidney disease of unknown origin (CKDu) (Vervaet et al. 2020; Arroyo et al. 2023). The etiology is uncertain, but the disease does not appear to be related to traditional risk factors such as diabetes, hypertension, aging, or glomerular disease (Jayamurugan et al. 2013; Ramirez-Rubio et al. 2013; Reddy and Gunasekar 2013; Weiner et al. 2013; Wesseling et al. 2013, 2014; Bodin et al. 2016; Glaser et al. 2016; Wijkstrom et al. 2017). In Central America, CKDu has been predominantly identified among sugarcane workers and workers in other agricultural sectors, and is estimated to have caused over 20,000 deaths in the last 10 years (Johnson et al. 2019). It principally affects men in their 20s to 40s, and results in chronic tubulointerstitial nephritis. Studies suggest that recurrent episodes of acute kidney injury, recurrent dehydration, exposure to high temperatures, and physically demanding working conditions contribute to CKDu risk (Roncal-Jimenez et al. 2015; Butler-Dawson et al. 2018, 2019, 2021; Johnson et al. 2019; Dally et al. 2020,2021), but other risk factors that have not been thoroughly examined include exposure to particulate air pollution and its constituents (Johnson et al. 2019; Schaeffer et al. 2020).

In Central America, sugarcane fields are burned prior to harvest to reduce plant litter and exposure to vermin. There is limited data on particulate matter (PM) exposure in this workforce, though published studies, including our pilot work in Guatemala, found that the PM from burned sugarcane is a complex mixture that is enriched with known and putative nephrotoxicants, such as silica and heavy metals (Le Blond et al. 2008, 2010, 2017; Schaeffer et al. 2020). Recent research suggests that occupational exposure to respirable silica increases the risk for CKDu (Steenland 2005; Millerick-May et al. 2015; Sponholtz et al. 2016; Mascarenhas et al. 2017). Inhalational exposure of amorphous silica nanoparticles has been shown to induce chronic kidney disease characterized by tubular injury and inflammation in rats (Sasai et al. 2022), consistent with the observation that silica and sugarcane ash-derived silica nanoparticles target proximal tubular cells (Warheit 2001; Nakagawa et al. 2015; Mascarenhas et al. 2018; Stem et al. 2023) Investigating the potential link between exposure to airborne PM₅ and its constituents will help assess the hypothesis that exposure to these airborne contaminants is associated with acute kidney injury in agricultural workers, and is therefore a contributing factor, along with heat stress and dehydration, that produces synergistic adverse effects on kidney function.

Previous studies have rarely measured exposure in this population, and those that have used a combination of area sampling or worker-proxy breathing zone samples (Boeniger et al. 1988; Sinks et al. 1994; Le Blond et al. 2008, 2017; Schaeffer et al. 2020). To explore our central hypothesis, we evaluated sugarcane worker exposure to airborne particulate matter by collecting personal breathing zone air samples in adult male participants at risk for CKDu. The main objective of this study was to characterize the magnitude and variability in personal breathing zone particulate matter concentrations in individuals for use in epidemiological research assessing exposure-response relationships for biomarkers of effect measured pre- and post-shift. A secondary goal was to explore how full-shift particulate matter exposure estimates compare to local ambient concentrations.

Methods

Study site

This work was conducted in the fields near a large sugarcane mill in the Escuintla Department of Guatemala during the 2021–2022 and 2022–2023 harvest seasons. These measurements were collected as part of an ongoing investigation of worker exposure in male workers at risk for CKDu. Ethics review and approval for the study were received from the Colorado Multiple Institutional Review Board (COMIRB #20-0509) and ZUGUEME Comité de Ética Independiente in Guatemala.

Study design

Male harvesters were sampled over full-day work shifts across 2 harvests: in February and March 2022 and in December, February, and March during the 2022–2023 harvest. Personal breathing zone sampling (using filter-based methods described below) and PM ambient measurements (using both filter-based and real-time samplers) were collected during normal harvesting activities on each monitored day. An unsuccessful pilot of these methods was conducted in December 2021 using a parallel particle impactor (PPI, SKC) as part of the sampling train. Due to the overloading of the collection media after monitoring for as little as 1 h, we changed the sampling inlets. We then piloted a size-selective sampling train using an aluminum cyclone (described below). This pilot demonstrated that the maximum duration of personal breathing zone particulate matter sampling for a valid sample (ie cyclone and filter not overloaded) was 4 h. Due to the voluntary nature of participation, the desire of both the workers and their employer for minimal interruption to their work across a shift, and the dispersion of groups of workers across large areas, we developed a minimally invasive sampling strategy that allowed us to set up workers with our monitoring equipment in the morning and then retrieve sampling trains at the end of a shift. To obtain data representative of a full work shift personal breathing zone sample we outfitted all workers with size-selective sampling trains with randomly assigned delayed pump start times in half the workers. Workers were monitored either at the beginning of the work shift (hereafter: AM) or after a 4-h delay (hereafter: PM). As a result, at the end of each day of monitoring, we had obtained 4-h personal breathing zone samples in either the AM or PM for all workers. Each workday we also collected (i) a stationary filter-based full-shift ambient sampling using the same cyclone sampling trains as for personal sampling, and (ii) full-shift ambient real-time aerosol measurements with a 1-min time resolution.

Work setting and work practices are described in Schaeffer et al. (2020) and briefly summarized here. Fields to be harvested are burned up to a day before the workers enter them, and the sugarcane is then manually harvested using a specialized machete (cuma). Job tasks included cutting, trimming, and stacking cane. Workers typically have 2 rest breaks and 1 lunch break of ~1 h, and their shift typically ranges from 9 to 10 h. Workers are provided personal protective equipment consisting of goggles, a sun hat, a long-sleeved shirt, gloves, wrist and shin guards, and boots. Respiratory protection is not required during harvest.

Particulate matter measurements

Personal breathing zone particulate matter samples were collected using an SKC Aluminum Cyclone sampler (Model 225-01-02, Eighty-Four, PA) and conductive polypropylene cassettes with pre-weighted 37- mm polyvinyl chloride filter (PVC, SKC Model 225-5-37), with a pore size of 5.0 μm. The cyclone constructed out of conductive aluminum with polypropylene cassettes were used to minimize electrostatic effects and attendant losses to the internal wall of the cassette (Barron 2003; Ashley and Harper 2013).

Ambient particulate matter samples in the field where workers were monitored were collected using the same sampling trains as for the workers, and these gravimetric area samples were compared to the personal particle measurements. The ambient gravimetric sampler was co-located with a DustTrak DRX Aerosol Monitor (TSI, Shoreview MN) used to measure the respirable fraction, which the manufacturer defines as those particles with a d50 of 4 microns in aerodynamic diameter (TSI 2023). Both full-shift ambient gravimetric PM₅ and Dustrak monitors were placed on a table at approximately 1.0 m above ground level; the table with the sampling equipment was periodically moved as work groups progressively cut their way through the sugarcane field(s).

Each harvester was outfitted with a marathon vest (Ultimate Direction, Model V2, Broomfield, CO), an SKC cyclone sampler, tubing, and pump as the sampling train for their entire work shift (Fig. 1). To prevent inversion of the sampler and ensure sample integrity, cyclone inlets were positioned in the personal breathing zone of the worker using a cassette holder that was securely fixed to the marathon vest in an optimal position based on our pilot sampling and previous work. Each harvester wore a sampling pump (SKC Inc., Model AirChek XR5000, Eighty-Four, PA) to draw air into the sampler at 2 L/min. Cleaning procedures were followed as recommended by the manufacturer (SKC 2024). Furthermore, while care was taken to ensure the cyclone did not invert during the sample collection period, post-sampling procedures for separating the cassette from the cyclone were also implemented.

All cyclone sampling trains were calibrated to 2 L/min before and after sampling using a secondary flow standard via an SKC multi-purpose calibration jar (SKC 225-111) and SKC chek-mate calibrator. The SKC Aluminum cyclone conforms to the ISO 7708/CEN criteria (ie NIOSH Method 0600) at 2.5 L/min, but we inadvertently did not update the flow rate after changing to the cyclone inlets, so all samples were collected at 2 L/min, which resulted in a d₅₀ of 5 µm (PM₅).

The PVC filters were weighted before and after each sampling event at Colorado State University using a Mettler MT5 Balance (Mettler-Toledo, Inc., Columbus, OH) to determine a pre-/post-sample weight. All gravimetric analyses were performed in a temperature and humidity-controlled chamber (Martenies et al. 2020; Schaeffer et al. 2020). Filters were prepared for gravimetry by desiccation for a 24-h period and then static neutralized using a U-Electrode (Mettler-Toledo, Inc.) prior to weighing. Field blanks were collected every tenth sample (n = 28) and the median field blank mass (0.25 µg; range 0 to 1.7) was used for field blank correction of all personal breathing zones and ambient samples. Time-weighted average (TWA) concentration (μg/m³) was calculated by dividing the mass on each filter after field blank correction by the volume of air sampled to obtain final personal and ambient gravimetric concentrations. Personal breathing zone sample final concentrations were 4-h TWAs for either AM or PM sampling; these were used to develop the full-shift 8-h TWA estimates as described below. Full-shift ambient sample duration was 8.9 ± 0.6 h (range 7.92 to 10.0) because these measurements continued during both lunch and rest breaks.

Statistical analysis

Descriptive statistics were generated as the mean, median, and 10th, 25th, 75th, and 90th percentiles as well as geometric mean and geometric standard deviation (GM/GSD). The normality of the data was assessed through visual examination of the Quartile-Quartile plots. Because the underlying distributions were skewed, we used a non-parametric statistical test (ie Wilcoxon Rank Sum Test) to compare the concentration distributions between AM and PM measurements as well as between shifts. One-way non-parametric analysis of variance (ANOVA) (Kruskal Wallis) was used to determine within worker variability of personal breathing zone exposure. Mixed-effects linear regression with an unstructured covariance matrix was used to assess between and within variability in worker personal breathing zone exposure concentrations.

We estimated 8-h TWA exposure for each worker using a combination of each individual’s AM or PM 4-h PM₅ personal breathing zone measurement averaged with data from other workers on the same shift to obtain full-shift 8-h TWA. Thus the 8-h TWA for an individual with a 4-h AM (PM) breathing zone concentration was obtained by taking the weighted average of that AM (PM) sample and the median value of the corresponding PM₅ personal concentrations for all the PM (AM) measurements on that day. To assess the validity of these 8-h TWA estimates were compared to the individual 4-h measurements with a scatter plot and corresponding regression line (Supplementary Fig. S1).

To characterize worker exposure relative to ambient concentrations, the 8-h TWA estimate was divided by the full-shift ambient measurement for that day using the same measurement method as for the personal breathing zone samples. All personal breathing zone to ambient comparisons were conducted using the cyclone-obtained measurements; real-time DustTrak data was used to document the number and magnitude of short-term peaks in a field each day. All statistical analyses were done in SAS version 9.4 (Cary, NC).

Results

Ambient particulate matter concentrations

Table 1 summarizes all TWA ambient PM₅ ambient measurements across the sampled work shifts (work shift length: median 9.5 h; range 4.8 to 10.67) when concurrent personal breathing zone samples were collected. Geometric mean ambient PM₅ concentrations were 151 μg/m³ (range 22.6 to 2,360) across all shifts and GM concentrations varied significantly between shifts (P = 0.008): for example, the maximum observed in March 2022 was less than the 25th percentile of February 2022 measurements.

One-minute average DustTrak measurements were collected each day alongside the ambient cyclone inlet samplers. These real-time measurements had multiple episodic short-term peaks that ranged up to ~100 times higher than median concentrations, which ranged from 28 to 178 µg/m³ (Supplementary Fig. S2). The 90th percentile concentration for these real-time measurements (332 µg/m³) was chosen to characterize the magnitude and duration of peak concentrations observed using this measurement method. Across all 29 monitored days, there were peak concentrations on 20 days. The median peak duration was 32 min (range 1 to 164). Windspeed in these fields were relatively low, with an average of 1.2 m/s (range 0.5 to 2.0).

Personal breathing zone PM₅ concentrations

Four-hour AM or PM personal breathing zone PM₅ air samples were collected from 143 male workers (mean age 32 ± 9.6 years, range 18 to 57) (Table 2). We obtained 287 valid samples, with 3 measurements in 35 workers, 2 measurements in 74 workers, and a single measurement in 34 workers.

The 4-h personal breathing zone PM₅ concentrations for all participants and days were highly skewed and ranged from 20.5 to 1930 μg/m³, with a median (GM) of 449 (415) μg/m³ and substantial day-to-day variability (Table 2; Fig. 2). We observed a difference in the distribution of personal breathing zone PM₅ concentrations between shifts, with significantly higher concentrations observed during the February time point in both years (P < 0.001) (Table 2, Fig. 2). Overall, the ANOVA revealed there was no significant difference in the distribution of personal PM₅ concentration between study participants (P = 0.243). The regression analysis revealed there was no significant difference in the overall distribution of the 4-h measurements between AM and PM samples (P = 0.194) across shifts, but we observed a significant interaction effect between month/year timepoints and shift, largely driven by higher personal PM₅ concentrations in the AM and PM shifts during the March 2022 and December 2022 (P = 0.004).

Table 3 summarizes 8-h TWA personal breathing .zone exposure concentration estimates for each day overall shifts. Across shifts, group sizes ranged from 3 to 16 workers, and median personal breathing zone TWA estimates ranged from 241 to 880 μg/m³ with a broad range of TWA exposure estimates within days. Supplementary Figure S1 summarizes the relationship between the 8-h TWA estimates and the 4-h personal measurements for each individual, with an R² of 0.741 and an intercept value of 188 μg/m³_.

Table 4 summarizes the overall relationship between the TWA personal breathing zone estimates and ambient levels in the field for that day expressed as a ratio. The distribution of ratios was skewed and ranged from ~0.14 to 33.6, with an overall median of 3.5. When stratified by shift, differences in ratios between shifts were consistent with observed trends in ambient and personal breathing zone concentrations. Similarly, when stratified by AM/PM, differences in ratios between morning and afternoon were consistent with observed trends in ambient and personal concentrations. Lastly, we also compared the ratio of personal 4-h PM₅ measurements to concurrent full-shift ambient PM₅: these ratios were generally smaller than observed for the TWA/concurrent ambient ratios, with an overall median of 3.1 and range of ~0.1 to 49. Regardless of the metric used in the numerator, median estimated personal breathing zone exposure concentrations of PM₅ in these workers were ~3- to 5-fold as high as the concomitantly measured ambient concentrations in these fields, although ratios were highly skewed and at the high-end varied up to more than 48-fold as high as the concomitantly measured ambient concentrations.

Discussion

In a population of male sugarcane harvesters at risk for CKDu, we observed median personal breathing zone PM₅ exposure concentrations of nearly 450 µg/m³. Previous studies have rarely monitored the breathing zone of individual sugarcane workers, and most studies have used area monitors or a few personal samples in individuals simulating work practices that likely underestimate personal breathing zone exposures in this population (Boeniger et al. 1988, 1991; Le Blond et al. 2017; Leite et al. 2018; Schaeffer et al. 2020). Our data indicate that hand cutting and stacking of burned sugarcane contributes more to personal breathing zone concentrations than local ambient particulate matter levels. This study is also one of the first to measure breathing zone particulate matter in sugarcane workers and to compare it to ambient concentrations in the fields. Importantly, we found that reliance on ambient particulate matter sampling would greatly underestimate personal exposure to dust and ash. Median personal breathing zone measurements were ~3 to 11 times as high as ambient concentrations in these fields.

These workers did not smoke tobacco while on the job, so the observed exposure concentrations are a combination of their ambient and work-related exposure to PM₅. It is notable that monitoring of personal exposure was only possible if the duration of sampling was truncated to 4 hours to reduce the chance of overloading samplers. Nonetheless, by using staggered pump start times each day it was possible to obtain robust data to develop estimates of full-shift breathing zone TWA PM₅ exposure concentrations for a relatively large number of workers.

This is the largest study to explore particulate matter exposure in a population at risk for a disease that affects thousands of workers in Central America and at other locations across the world. Our observations as well as our pilot study strongly suggest that the majority of the particulate in the breathing zone of the harvesters comes from the resuspension of dust and ash (Schaeffer et al. 2020). We observed episodic smoke plumes that came from nearby upwind fields that were burned while the harvesters monitored in this study were working. These plumes were relatively infrequent and short in duration, typically lasting a few minutes. Ultimately this means that the constant resuspension of burnt sugarcane ash and dust while cutting is a larger source of PM₅ exposure compared to the ambient background or the short-term near-field burning sources. Agricultural production practices that affect workloads, lengths of shifts, and proximity to field burning, are likely to vary widely. As a consequence use of other methods, such as remote sensing to estimate exposure to smoke from burning or ambient monitors that are not located in the fields where this type of work is occurring, would substantially underestimate particulate air pollution exposure for sugarcane harvesters.

Rates of both occupational and non-occupational illnesses are among the highest in the agricultural sector, including rates of respiratory, cardiovascular, and renal disease (Rushton 2017; Johnson et al. 2019; GBD 2020; Pega et al. 2022), all of which have been hypothesized or shown to be related to particulate matter exposure. There are no current occupational standards specifically designed for workers in this industry, although burning sugarcane fields and agricultural burning practices have long been recognized as a substantial occupational hazard (Boeniger et al. 1991; Le Blond et al. 2008, 2017; Udeigwe et al. 2015). The personal breathing zone concentrations observed in this study are lower than the ACGIH’s (American Conference of Governmental Industrial Hygienists) guideline for Particles Not Otherwise Specified (insoluble or poorly soluble, PNOS) of 3 mg/m³. We note, however, that the ACGIH PNOS documentation indicates that this guidance applies to particles with “low toxicity,” which, in our view, significantly reduces its applicability in this case given that amorphous silica, polycyclic aromatic hydrocarbons, and metals are known to be present (Schaeffer et al. 2020; ACGIH 2023). Our findings suggest that future research should focus on: (i) measuring personal, rather than ambient particulate matter concentrations, and (ii) addressing occupational exposure limits for workers in this industry given the magnitude of this exposure and the other hazards (eg metals, silica) known to be present in the complex mixture.

Relatively little is known about the effects of particulate matter exposure on the kidney. A few recent studies conducted in the United States in veteran and Medicare populations observed that elevated ambient PM_2.5 concentrations are associated with a higher risk of declining kidney function, incident CKD, and end-stage renal disease (Mehta et al. 2016, Bowe et al. 2017, Bragg-Gresham et al. 2018). One of these studies observed an increased risk of a ≥30% decline in estimated glomerular filtration rate, eGFR, and that end-stage renal disease was associated with an increase of 10 µg/m³ in the concentration of PM_2.5. These findings were observed even at PM_2.5 concentrations well below health-based guidelines and standards promulgated by the WHO and EPA, respectively (WHO 2021; USEPA 2023).

This manuscript is part of a multifaceted study that is collecting personal-level outcome metrics and environmental exposures. These full-shift TWA particulate matter exposure estimates will contribute to subsequent manuscripts characterizing hazardous constituents, such as metals and silica, and relate these exposures to kidney biomarkers and other health outcomes across work shifts.

Strengths and limitations

This study has several strengths that facilitated the completion of this sampling campaign. The cooperation of the employer-provided unique access to this working population to recruit, consent, and collect personal PM₅ exposure data from these workers. Therefore, we were able to document concurrent 4-h average AM or PM personal PM₅ concentrations in groups of workers as well as paired, full-shift ambient PM₅ concentrations. This sampling strategy allowed us to obtain valid personal breathing zone samples in many workers from this industry and estimate full-shift TWA exposure concentrations. Similarly, we were able to document the number and duration of short-term peaks in ambient concentrations in these fields.

These findings have limitations, most importantly that while we attempted to assess exposures to the respirable size fraction, after finding that PPI samplers would not work as designed due to the high particulate matter concentrations, we shifted to the SKC Aluminum cyclone but did not change the flow rate of 2 L/min. This shifted the d₅₀ of these samples to 5 µm. Despite this shift, our findings demonstrate that sugarcane workers experience repeated exposures to high concentrations of fine and coarse particulate. The PM₅ size fraction plausibly collects more particles than the respirable fraction. Because the exposure was within allowable occupational exposure limits (ie for PNOS), the workers did not exceed those limits associated with the respirable fraction. Nonetheless, these TWA exposures are undoubtedly in the range that can cause adverse health effects. Other potential limitations include unanticipated variability in both ambient and personal concentrations between the mornings and afternoons and between monitored months. The lower concentrations observed in the afternoons likely reflect several factors that were unavoidable, including a generally slowing work pace due to higher temperatures as the workday progressed, and groups of workers being moved on buses to new locations during the afternoon work shift on one of the monitored days. There were also short rain events in March 2022 on 2 of the days that temporarily stopped work and likely reduced ambient and personal PM₅ breathing zone concentrations as well. Real-time and average ambient field measurements were collected at only one location in each field: while ambient samplers were moved to keep them near the workers, this approach provides limited spatial information on the variability in PM₅ in these fields. In our judgment, this is a limitation unlikely to affect our overall findings given the relatively large observed differences in personal breathing zone and ambient concentrations. This indicates that work practices are a much larger source of personal particulate matter exposure than ambient concentrations while these harvesters are at work. Recognition of such real-world sources of variation raises additional important questions about disease prevention in the agricultural setting. Agribusinesses face significant challenges if they seek to establish best practices for an effective occupational health management system to monitor and mitigate work-related air pollution exposures. A sampling schedule would need to include frequent monitoring under a range of work and environmental conditions and include both ambient and breathing zone measurements in the sampling strategy.

From a public health perspective, these findings highlight the need to address other health outcomes in addition to CKDu. While our study was designed to understand particulate matter exposure in workers known to be at high risk for CKDu, the observed high particulate matter exposure establishes the need for researchers and public health officials to investigate agricultural workers’ risks for respiratory, cardiovascular, cancer, and other health outcomes related to air pollution (WHO 2021; USEPA 2023). Solutions for mitigating exposure and conducting surveillance for multiple health outcomes of concern need to be addressed if we are to reduce the burden of occupational illness in agriculture.

Supplementary material

Supplementary material is available at Annals of Work Exposures and Health online.

Acknowledgments

We wish to thank all the workers who have made this work possible. We would also like to thank Daniel Pilloni, MD, Alex Cruz, MD, and Inés Amenabar of Groupo Pantaleon.

Funding

Research reported in this publication was supported by the National Institute of Environmental Health Sciences of the National Institutes of Health under Award Number R01ES031585. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. It was also supported by data and resources from Grupo Pantaleon under a memorandum of understanding with the University of Colorado. Funders had no role in data analysis, interpretation of data, writing the manuscript, or the decision to submit the findings for publication.

Conflict of interest

The University of Colorado has a memorandum of agreement with Pantaleon, a Guatemala-based agribusiness. Pantaleon provides partial financial support for research through a contract with the University and has provided access to the employees who volunteered to participate in this research project. The University of Colorado employed appropriate research methods in keeping with academic freedom, based conclusions on critical analysis of the evidence and reported findings fully and objectively. The terms of this arrangement have been reviewed and approved by the University of Colorado in accordance with its conflict-of-interest policies. The authors declare they have no actual or potential competing financial interests.

Data availability

The data generated during this study can be found either within the published article or supplemental files. Additional data are available from the corresponding author on reasonable request.

References