No living animals were used directly during the course of this study. Data were collected in an electronic medical records system that uses case classifiers, Systematic Nomenclature of Medicine (SNOMED) codes, and free text to capture routine husbandry practices and veterinary medical care. Data included conceptions, births, pregnancy losses, housing, and demographic and health condition information. All animals in the electronic health records were housed at the ONPRC and managed under the IACUC-approved protocol of the Oregon Health and Science University West Campus. This research adhered to the American Society of Primatologists (ASP) Principles for the Ethical Treatment of Non-Human Primates and adhered to the legal requirements of the United States, where the research was conducted. As part of the preventative veterinary medicine program at ONPRC, all animals were tested annually for SARS-CoV-2 with a serological antibody assay (CSA: SARS COV-2, Intuitive Biosciences©, Madison, WI). All animals tested negative throughout the study period and remain negative at the time of this publication. Therefore, SARS-CoV-2 was not a confounder in our study.

Housing was in accordance with standards established by the U.S. Federal Animal Welfare Act and the Guide for Care and Use of Laboratory Animals. All shelter structures associated with outdoor housing areas are open-air without air filtration. Animals are housed in groups in continuous full contact. Figures for outdoor housing are previously published (Prongay et al., 2013). Permanently installed sprinklers and misters provided water for vegetation and a source of cooling during the wildfire smoke event. Standard colony practices involving records of indoor/outdoor transfers allowed complete identification and enumeration of individual monkeys who were outdoors on a day-by-day basis during the 9-day exposure period across all 7 years.

One to four times per year, on a rotating schedule, all animals were sedated with ketamine HCl (10 mg/kg IM; Ketathesia^™, Henry Schein Animal Health^®, Dublin, OH) for routine group management, including physical examinations, pregnancy evaluation, weighing, mammalian old tuberculin skin test, antihelminth administration (Ivomec^®, ivermectin 1%, Merial, Lyon, France), and blood draws for serology and genetic testing. Pregnancy detection and trimester status was done by bimanual palpation with rectal probe by professional primate veterinarians. All pregnancies were confirmed, and a trimester status was assigned by palpation only (Mahoney, 1975).

Rhesus macaques and Japanese macaques housed outdoors were exposed to poor air quality September 10–18, 2020 from five major fires burning during this period covering 340,353 hectares (3,404 km²) in Oregon (Abatzoglou et al., 2021; Reilly et al., 2022). The closest major fire, the Riverside fire, burned 55,868 hectares (559 km²) and was approximately 68 km southeast from the ONPRC. (Federal Emergency Management Agency, 2020; Forest Grove, 2020)

PM2.5 were measured by the Oregon Department of Environmental Quality’s (DEQ) PM2.5 monitoring site at the Beaverton Highland Park Monitoring Station (EPA #41-067-0111; Latitude 45.47019, longitude −122.8164), which is approximately 7.4 km from ONPRC. The site is within the Tualatin Valley, where the ONPRC is located, and is the most likely air quality monitoring site to represent the ambient air in which outdoor macaques were exposed. This site uses a Correlated Radiance Research M903 Nephelometer with a heated inlet to measure fine air particulates. Daily 24 hour average PM2.5 measurement raw data from this sensor were obtained from the EPA’s Air Quality System database from years 2014 through 2020 (United States Environmental Protection Agency, 2020).

Other pollutants present in wildfire smoke were measured in a lesser capacity than PM2.5 by DEQ’s Air Toxin monitoring system. In this program, stainless steel passivated Summa cannisters are used to collect whole air samples from across Oregon approximately every six days. The closest air toxin testing to the ONPRC were KairosPDX National Air Toxics Trends Station (EPA #41-051-2010; Latitude 45.5581, longitude −122.671) in Portland, Oregon (16.5 km from ONPRC) where samples are tested for Volatile Organic Compounds, and Hare Field air monitoring station (EPA #41-067-0004; Latitude 45.5285, longitude −122.972) in Hillsboro, Oregon (7.4 km from ONPRC) where carbonyl compounds are tested. Wildfire smoke constituents that are tested at these sites include acrolein, benzene, chromium, and formaldehyde. Only data collected during the exposure year and month were available in a limited amount and reported for informational use. Comparisons or associations of these other components of wildfire smoke with health effects in macaques at ONPRC could not be investigated in this study.

Days when PM2.5 exceeded the EPA’s standard for PM2.5 24h average include September 10^th, 2020, through September 18^th, 2020; these dates were used as the timeframe for determining each exposure group (Figure 1). Cohorts from 2014–2019 were restricted to the same dates each year (September 10^th – September 18^th) as the wildfire smoke event to rule out seasonality as a confounding factor. Due to the difference in exposure to PM2.5 between animals living outdoors in 2020 and animals living outdoors during years 2014–2019, two exposure groups were formed. Fetuses of animals carried by outdoor housed female macaques in utero from September 10th, 2020, to September 18^th, 2020, were included in the exposed group. Fetuses of animals that were pregnant and lived outdoors in the six previous years during the same dates each year (September 10th, 2014 – September 18^th, 2014, through 2019) were included in the less exposed group. Data from any animals that were outdoors for less than the entire 9-day exposure time frame for each year (September 10^th – September 18^th of 2014 through 2020) were excluded from all analyses. In all, 2 observations were removed from the 2014 cohort, 10 from 2015, 7 from 2016, 10 from 2017, 7 from 2018, 6 from 2019 and 3 from 2020.

Pregnancies that were detected between 2014 and 2020 were captured during routine screenings. From the date of the pregnancy detection and the trimester of each pregnancy, a conception date was estimated. First-trimester pregnancies were estimated to be at 55 days gestational age, second-trimester pregnancies at 110 days, and third-trimester pregnancies at 165 days. When dams had multiple pregnancy records for the same pregnancy, these were reduced by only keeping the latest pregnancy record. Pregnancies that resulted in an offspring were connected to the birth record via parentage data. Births that occurred from dams without a recorded pregnancy were included as additional instances of conception. Pregnancies that were not connected to an offspring record were assumed to be pregnancy losses. A conception date was estimated for births as 165 days before the birth date (the average gestational age for macaques) or by gestational age for pregnancies (Hartman, 1928). Pregnancies and births were included in the study population if the estimated conception date was between the first day of the exposure period and 165 days past the last day of the exposure period.

Births were marked as pregnancy losses when categorized as one of the following; type of death was recorded as a miscarriage (<120 days gestation), stillbirth (>120 days gestation, never breathed), or a neonatal death that was born and died on the same day. Pregnancies that ended in prenatal loss, pregnancies that had no record of a live birth, or births with no pregnancy record where the offspring was born and died on the same day were defined as pregnancy losses; all other observations signified live births.

The total number of conceptions, pregnancy losses, live births, and the proportion of pregnancies lost each year was calculated. A background rate of pregnancy loss in the less exposed group was calculated as the average percentage of pregnancies lost over the years when data was collected for the less exposed group (September 10^th – September 18^th, 2014, through 2019). A relative risk (RR) was calculated for each less exposed cohort (each year from 2014–2019) compared to the exposed cohort (the year 2020) and for the less exposed group as a whole (years 2014–2019 combined) compared to the exposed group. The RR was calculated as the proportion of pregnancies lost in the exposed group divided by the proportion of pregnancies lost in the less exposed group and represents the likeliness of pregnancy loss in the exposed group compared to the likeliness of pregnancy loss in the less exposed group.

Health outcomes of exposed versus less exposed animals included a comparison of mortality, respiratory disease, and hospitalization up to one-year post-exposure, and weight gain in the first year of life. Three exposure groups were defined to compare the difference in association between fetuses exposed in utero, exposed infants, and less exposed infants.

The in utero exposed group was defined as animals born between the first day (September 10^th, 2020) and 165 days past the end of the exposure period (March 2^nd, 2021), where the dam was outdoors during the exposure period where 24h average PM2.5 concentrations exceeded standards set by the EPA (September 10th, 2020 – September 18th, 2020) and the infant was born alive and survived at least one day. The exposed infant group includes animals born between June 9th, 2020, and September 9th, 2020 (three months old or less at the beginning of the exposure period), and the infant was outdoors during the exposure period. The less exposed infant group was also born between June 9th and September 9th but during years without exposure (2016 – 2019) and were all outdoors during the same period as the exposure group (September 10th – September 18th of each year) during years with considerably lower PM2.5 concentrations (Figure 1). In all exposure groups, data from animals that were not outdoors for the entire 9-day exposure period of September 10^th – September 18^th of 2016 through 2020 were excluded from all analyses. In all, 0 observations were deleted from the 2020 in utero exposed data, 5 from the 2020 infant exposed data, and a total of 81 from the 2016–2019 less exposed data (19 from 2016, 22 from 2017, 15 from 2018, and 25 from 2019). Infants 3 months or younger have been previously identified as a vulnerable age group of macaques to wildfire smoke (Black et al., 2017).

Only natural deaths that were defined as neonatal, spontaneous, or euthanasias due to health reasons were included for mortality events. Respiratory problems were defined as an animal that was diagnosed with respiratory disease by a veterinarian or died of a respiratory-related illness. All deceased animals underwent a necropsy performed by an American College of Veterinary Pathologists-boarded veterinary pathologist which provided respiratory causes of death. The respiratory outcome variable was collapsed into a binary outcome so that animals with multiple respiratory problems were only included once. Hospitalizations were defined by the primary condition when an animal was removed from their outdoor breeding group to a hospital location for an illness. Preventive medicine cases, traumas or injuries, and infants hospitalized for their dam’s illnesses were excluded from the hospitalization dataset. When animals were moved several times within a short period for the same medical condition, the records were de-duplicated. The outcome variable was binary, and animals with single or multiple hospitalizations were only included once. All health outcomes (mortality, respiratory problems, and hospitalizations) were collected for animals in each exposure group from the first day of exposure to 365 days past the last day of exposure. The number of cases in each outcome by exposure group and the percentage of the exposure group affected was calculated.

Exploratory and descriptive analyses were performed to evaluate demographic characteristics of the three groups and explore the timing of health outcomes. Covariates collected for inclusion in regressions include the sex of the offspring (male or female), species (Rhesus macaque or Japanese macaque), if the dam was primiparous, dam age at the time of birth (categorized into quartiles), and the age of the offspring for analyses of weight. The frequency of these four demographic characteristics was determined in each exposure group. For categorical variables, the number of animals and the percent of that level in each exposure group were determined. A p-value was calculated by Pearson’s chi-squared test as the difference between exposure groups. Among continuous variables, the mean and standard error was calculated within the variable for each exposure group and a p-value was calculated by analysis of variance as the difference between exposure groups.

Logistic regression models were utilized to evaluate the association between exposure group and each outcome. Offspring sex, species, dam parity, and dam age were included as covariates individually and as two-way interaction terms in logistic regressions for each outcome. Backward elimination using the reference p-value of 0.05 was used to determine the predictive variables. A total of 15 backward elimination steps were completed per outcome. Each model was checked for collinearity using variance inflation factor (VIF), which was less than 5. Lastly, body weights of animals in exposure groups were included for up to one year of age for each animal in only rhesus macaques. There were insufficient Japanese macaque weight records to use for analysis. In cases where animals were weighed multiple times within 30 days, only the oldest observation was kept. There was more than one weight sample for 260 of the 405 animals in the dataset. This dataset has within-participant data, which lacks independence, and a linear mixed effect model was chosen to evaluate the longitudinal relationship of exposure group on weight gain. The linear mixed effect model included covariates offspring sex, species, dam parity, dam age, and weight per days old as potential predictors. A Kruskal-Wallis rank sum test was utilized to determine if there was a difference between groups in the weight per days old as individual data points and not as within-participant observations.

Data cleaning, transformation, and statistics were performed using the R statistical program, version 4.1.2 (R Core Team, 2021). RR was calculated using the R package “epitools”. The logistic regressions were fit using the “glm” function in R with the specification “family = binomial”, and the linear mixed effect was calculated using the R package “nlme” (R Core Team, 2021).

Macaques in our wildfire smoke groups were exposed to elevated PM2.5 that exceeded the EPA’s 24h National Ambient Air Quality Standard (35 μg/m³; 100 AQI) for 9 days during the period of September 10–18, 2020, including three days in the EPA’s AQI category Hazardous (>250 μg/m³), four days in Very Unhealthy (150–250 μg/m³), one day in Unhealthy category (55–150 μg/m³), and one day in the Unhealthy for Sensitive groups category (35.5–55 μg/m³; Figure 1). Daily 24h PM2.5 air reached high as 278 μg/m³. The less exposed groups were animals in the years 2014 to 2019. Wildfire smoke and poor air quality occur every year in the less exposed group but to a much lesser severity (Figure 1). PM2.5 air concentrations were higher than the 24h standard 35 μg/m³ for 2 days in year 2015, 7 days in year 2017, and 6 days in year 2018. 24h PM2.5 air concentrations in the less exposed group never were in Unhealthy, Very Unhealthy, or Hazardous categories, and the less exposed group was never exposed to 24h PM2.5 concentrations more than 65 μg/m³.

There were temporary increases in other air toxins found in wildfire smoke during the wildfire event period. Acrolein, acetaldehyde, benzene, hexavalent chromium, and formaldehyde were all elevated in air samples within 20 km of the ONPRC taken on September 12, 2020. As the wildfire smoke event ended, these air toxins levels returned to more typical concentrations in local samples taken September 24, 2020, and September 30, 2020. State-wide monitoring found Benzene, acetaldehyde, 1,3-Butadiene, acrolein, and manganese were at concentrations 40 times their annual average, and benzene, benzo[a]pyrene, and formaldehyde were found above acute risk-based concentration standards set by DEQ (DEQ, 2022; Fellows et al., 2021). Oregon DEQ reported that phthalates were measured at a level above background in state-wide monitoring during the wildfire event, but these levels were below the acute risk thresholds set by the state risk assessments.

Pregnancy loss was more likely in the group of pregnant females exposed to wildfire smoke. Overall, pregnancies conceived during the 2020 cohort were 4.1 times as likely to have been lost compared to pregnancies conceived between 2014 and 2019 (RR = 4.1; 95% CI = 3.2, 5.4, p < 0.001) (Table 1). In 2020, 52.4% of pregnancies were lost compared to an average of 12.7% from 2014–2019.

Wildfire smoke exposure was investigated as a risk factor for infant mortality, respiratory problems, and hospitalizations. The demographics table (Table 2) shows no significant difference in the percent of male animals (X2 = 0.86, df = 2, p = 0.652) and mean dam age between exposure groups (F(2, 577) = 0.10, p = 0.905). The percent of primiparous first-time females trended slightly higher in our exposure group (X2 = 2.36, df = 2, p = 0.308). The percentage of animals that were rhesus macaques was borderline significantly different (X2 = 6.10, df = 2, p = 0.047) between exposure groups. Rhesus macaques were the most common species in our less exposed and exposed groups. Notably, the in utero exposure group was entirely rhesus macaques, and no Japanese macaques were born after being exposed in utero to the wildfire events to include in our dataset.

We evaluated the relationship between the covariates (offspring sex, species, dam age, and dam parity), exposure group, and the health outcomes by logistic regression. There were seven respiratory problems diagnosed in the infant exposure group (9.3%, n=75) and eleven respiratory problems in the less exposed group (2.3%, n=489; Table 3). No problems were found in the in utero exposure group (n=16; Table 3). The respiratory problems diagnosed in the infant exposure group (n=75) included five (6.7%) pneumonia cases, one (1.3%) upper respiratory infection, and one (1.3%) pulmonary edema case. Problems in the less exposed group (n=489) included nine (1.8%) pneumonia cases, one (0.2%) pharyngitis case, and one (0.2%) epistaxis case. No problems were found in the in utero exposure group (n=16; Table 3). Infants exposed to wildfire smoke between 0 and 3 months old had significantly greater odds of diagnosis with a respiratory condition within a year post-exposure compared to infants not exposed to wildfire smoke (OR = 4.47; 95% CI 1.60, 11.76; p = 0.003; Table 4). There were no covariates that contributed to the model, and only exposure group remained in the model after backward elimination (Table S1). An odds ratio with the independent variable for exposure group and no other covariates is presented in Table 4. The in utero exposure group had an unremarkable OR, which rendered our comparisons unviable (OR <0.0001; 95% CI -inf, inf). This was likely due to the absence of Japanese macaques in the in utero group and the lack of events that occurred in the in utero group to compare with our less exposed group.

There were ten mortality events found in the infant exposure group (13.3%, n=75) and 40 in the less exposed group (8.2%, n=489). No mortality events occurred for our in utero exposure group (n=16; Table 3). The primary findings for mortality events found in the infant exposure group (n=75) included six (8.0%) cases of gastrointestinal issues, one (1.3%) pneumonia infection, one (1.3%) pulmonary edema case, one (1.3%) left ventricular hypertrophy case, and one (1.3%) cerebral edema case. Primary findings for a total of 40 mortality events in the less exposed group (n=489) included 33 (6.7%) gastrointestinal issues, two (0.4%) cholangiohepatitis cases, two (0.4%) pneumonia cases, one (0.2%) cerebral edema case, one (0.2%) wound case, and (0.2%) one encephalomyelitis case. No mortality events occurred for our in utero exposure group (n=16; Table 3). There was no significant difference in the odds of mortality in the year after a wildfire smoke event between the exposed and less exposed infant groups (OR = 1.78; 95% CI 0.80, 3.64; p = 0.129; Table 4). After backward elimination, only species was predictive of mortality in this model (Table S2). An odds ratio with the independent variable for exposure group adjusted for species is presented in Table 4. Again, the in utero exposure group had an unremarkable OR which rendered this group’s comparisons unviable (OR <0.001; 95% CI -inf, inf; Table 4).

There were thirteen hospitalizations that occurred in the infant exposure group (17.3%, n=75) and 81 hospitalizations in the less exposed group (16.6%, n=489). No hospitalizations occurred for the in utero exposure group (n=16; Table 3). Animals in the exposed group (n=75) were hospitalized for ten (13.3%) dehydration and diarrhea cases, and three (4.0%) pneumonia or upper respiratory infections. Hospitalizations for the less exposed group (n=489) included 76 (15.5%) diarrhea and dehydration cases and five (1.0%) pneumonia or upper respiratory infections. No hospitalizations occurred for our in utero exposure group (n=16; Table 3). There was no significant difference between exposed and less exposed groups in the odds of hospitalization in the year after a wildfire smoke event (OR = 1.09; 95% CI 0.55, 2.04; p = 0.799; Table 4). After backward elimination, only species was predictive of hospitalization in this model (Table S3). An odds ratio with the independent variable for exposure group adjusted for species is presented in Table 4. Like the infant mortality and respiratory models, the in utero exposure group had an unremarkable OR which rendered this group’s comparisons unviable (OR <0.001; 95% CI -inf, inf; Table 4).

Due to limited data in Japanese macaques, this dataset was limited to only rhesus macaques. Exposure group did not predict weight changes or growth rate in animals under a year. The linear mixed effect model showed no difference in weight gain between exposure groups (exposed in utero group: β = −0.0187, t(401) = −0.401, p = 0.689; exposed infants group: β = −0.002, t(401) = −0.093, p = 0.926, less exposed infants group: reference). After backward elimination, weight per days old and infant sex were predictive of weight gain, estimates presented are comparing exposure groups after adjustment for weight per days old and infant sex. A Kruskal-Wallis rank sum test discovered no significant difference in weight by age between exposure groups (H = 0.688, df = 2, p = 0.709). Figure 2 shows little variability in weight change up to one year of post-exposure between exposure groups.

The in utero exposure group was limited in the number of animals (n = 16), and therefore results in this case were uninterpretable. This effect may have been demonstrated had the sample size been larger. We also acknowledge that individual pregnant NHPs may have experienced varying levels of wildfire smoke in the prior years in addition to the exposure year 2020 which could be measured as chronic cumulative exposure, but, due to the complexities of studying chronic cumulative exposure, we were not able to discern this. Dichotomization of the quantitative exposure data also involved exclusion of observations on NHPs who were outdoors less than the full 9 days in all years. This led to selection bias and loss of information. However, we support the dichotomization of our study cohorts into “exposed” and “less exposed” since this approach is consistent with other observational studies in the field (Willson et al., 2021).

In conclusion, our results suggest wildfire smoke exposure in utero and to infants can lead to health consequences for infant macaques. Our monkeys were exposed nearly continuously to unhealthy and hazardous air quality and likely represent exposure to wild populations that cannot escape to safer air to breathe. This observational study of this closely monitored macaque colony demonstrated the associations between excess prenatal loss and respiratory disease to exposure to wildfire smoke, which would have otherwise been difficult to discern in a large population of wild macaques. It is critically important that we use all available nonhuman primate models to comprehensively understand the consequences of wildfire smoke, and we act with these comparative studies to mitigate health effects for humans and in wild populations by working with stakeholders to minimize the damage of global climate change. Without changes to human carbon emissions, inevitably, events like those of the Labor Day fires will occur more frequently.