Standard growth and diarrhea-associated growth faltering in captive infant rhesus macaques (Macaca mulatta)

Abstract

Reference growth studies of captive rhesus macaque infants have not accounted for diarrhea and the potential for growth stunting or growth faltering. Healthy infants without diarrhea could be used to build a standard growth chart and a tool used to detect growth faltering associated with diarrhea. We hypothesized infants who develop diarrhea during the first year of life would experience decreased linear weight gain compared to healthy infants, and we used healthy infants to establish standard growth of male and female infants. We hypothesized the lower 3rd percentile of standard growth would be cut-off criteria used in screening for diarrhea-associated growth faltering. Using a retrospective cohort of 6510 infant weight records in a multiple linear regression, daily weight gain through the first year of life was determined by sex, housing type, and health status. Male standard growth was 4.1 g/day (95% CI: 4.0 – 4.2 g/day) in corrals and 4.7 g/day (95% CI: 4.5 – 4.8 g/day) in shelter housing. Female standard growth was 4.0 g/day (95% CI: 3.8 – 4.2 g/day) in corrals and 4.4 g/day (95% CI: 4.0 – 4.7 g/day) in shelter housing. Diarrhea was significantly associated with decreased linear weight gain by up to 34% during the first year of life. Odds of growth faltering of infants, defined as those falling below the 3rd percentile of standard growth, were at least 8.9 higher given a history of diarrhea compared to healthy. The growth faltering cut-off criteria had a sensitivity of at least 53% for males and females to screen for diarrhea in infants between 6 and 12 months in shelters housing. Interinstitutional collaborations of infant rhesus macaque weight records would refine the standard growth charts and cut-off criteria, and additional morphometric data would provide a more nuanced picture of growth stunting

Introduction

Reference growth studies of captive infant rhesus macaques have not evolved along with decades of changes in diet formulation, husbandry practices, and social housing strategies (Small & Smith, 1986). During the same decades, the World Health Organization (WHO) began highlighting inadequacies in the National Center for Health Statistics (NCHS)’s child reference growth charts (de Onis & Yip, 1996). Until 2006, WHO had promoted the use of the NCHS reference growth charts developed from anthropometric data of children from one village in Ohio who were fed primarily infant formula (WHO Expert Committee, 1995), but WHO reported clinicians using the reference charts inappropriately to assess growth of breastfed children around the world (de Onis, Garza & Habicht, 1997). In response to concerns over the limited scope of the NCHS growth reference charts, the WHO Multicentre Growth Reference Study group pioneered standard growth definitions and charts to replace reference growth data (The WHO Multicentre Growth Reference Study, 2006a). The WHO standard growth charts were developed using only children thought to live an environment supporting healthy development; this was intentionally done to target how children should grow, rather than how they may have grown at a certain place and time (The WHO Multicentre Growth Reference Study, 2006a). The standard growth charts allowed clinicians to use the growth standards as tools when making judgement of an individual infant’s growth (Weaver, 2010). Currently, the WHO, The United Nations Children’s Fund, and the Centers for Disease and Prevention all recommend the use of WHO standard growth charts as a tool to facilitate identification and early intervention of infants at risk for malnutrition and growth stunting (Garza & de Onis, 2004; The WHO Multicentre Growth Reference Study, 2006b; Grummer-Strawn, Reinold & Krebs, 2010; Kuczmarski et al., 2002).

Diarrheal disease is a well-recognized cause of morbidity and mortality in children globally. Populations of children exposed to a high prevalence of diarrhea were excluded in the WHO standard growth charts due to the association with growth stunting (Man et al., 1998; The WHO Multicentre Growth Reference Study, 2006a). Among children, the odds of stunting increases with each diarrheal episode and each day of diarrhea (Checkley et al., 2008), and prolonged episodes of acute diarrhea increase risk of persistent diarrhea and growth stunting (Keusch et al., 2014; Moore et al., 2010; Prendergast & Humphrey, 2014). Reasons for the development of diarrhea remain unclear, and diarrhea is likely a complex interplay of pathogen, environment, and host. Enteric pathogens associated with diarrheal disease induce physiologic changes which may blunt villi in the intestines and could plausibly lead to poor nutrient absorption, decreased weight gain, and growth stunting (Mondal et al., 2012).

The morphometric data of rhesus macaque infant studies in the 1950s and 1980s did not address infant health or how infants should grow without diarrhea. Infant rhesus typically experience significant daily weight gains in a linear pattern through the first year of life (Saxton & Lotz, 1990; Small & Smith, 1986; van Wagenen & Catchpole, 1956). Conceivably, diarrhea may disrupt this linear growth. Associations of growth stunting and diarrhea have been demonstrated in multiple species besides primates (Carpenter & Burlatschenko, 2005; Windeyer, Leslie, Godden, Hodgins, Lissemore & LeBlanc, 2014). Piglets exposed to enteric pathogens frequently develop concurrent bacterial, viral, and protozoal infections, and experience weight gains 50% lower than healthy peers. Dairy calves with diarrhea have lower weight gain and higher mortality than healthy peers. For captive macaques, diarrhea has been a significant source of morbidity and mortality (Habermann & Williams, 1957; Holmberg, Leininger, Wheeldon, Slater, Henrickson & Anderson, 1982; Schneider, Prather, Lewis, Scatterday & Hardy, 1960; Russell, Krugner, Tsai & Ekstrom, 1988). Infant rhesus macaques with diarrhea were negatively associated with survival and infants were particularly vulnerable to diarrhea (Prongay, Byung & Muprhy, 2013). An infant’s odds of diarrhea-associated mortality were almost 2 times that of diarrhea-associated mortality of adult animals (Prongay et al., 2013).

Growth of a healthy population of rhesus macaques without diarrhea could be developed into a standard growth chart similar to the WHO’s weight-for-age standard growth charts of children. The WHO standard growth charts use the lower 3^rd percentile of median height-for-age to define growth stunting, but both weight-for-age and height-for-age reflect growth stunting and long-term health of individuals (de Onis, 1997). Measuring height requires strict adherence to standard methods (de Onis, Onyango, Van den Broeck, Chumlea & Martorell, 2004), and no such standard training was developed for rhesus macaque colonies. Captive macaques are regularly weighed, and weight-for-age charts would be a more applicable tool than height-for-age for detecting growth stunting for non-human primate clinicians. For the purposes of our study, we will refer to low weight-for-age below the 3^rd percentile of standard growth as growth faltering to avoid confusion with the WHO’s defined term for growth stunting as cut-off with height-for-age.

We looked retrospectively at infant weights to describe linear growth of male and female infant rhesus in two different outdoor housing situations. We established standard growth of our healthy infant rhesus macaques on either open earth substrate housing or covered sealed concrete housing. We hypothesized that infants with at least one instance of diarrhea during the first year of life would experience diminished linear growth during the first year of life. We also hypothesized infants with weights-for-age below the 3^rd percentile cut-off would have a higher risk of diarrhea in their history. Lastly, we described the sensitivity and specificity of our 3^rd percentile weight-for-age cut-off in screening for diarrhea.

Using a large retrospective cohort of healthy infants raised in outdoor social housing with dams, we present weight gain standards that can be used prescriptively by captive-colony managers to predict weight of rhesus infants of a specific age. This information will allow early identification of infants not meeting weight gain standards and, in conjunction with other physical and developmental milestones, allow early identification and treatment of growth faltering and diarrhea. We envision multiple facilities could contribute morphometric data to further refine our growth standards.

Methods

Study Population

This research adhered to the American Society of Primatologists Principles for the Ethical Treatment of Primates. No living animals were used directly during the course of this study, and an Institutional Animal Care and Use Committee (IACUC) approval was not required. Data were collected from Primate Records and Information Management (PRIMe), an electronic medical records system which uses case classifiers, Systematic Nomenclature of Medicine (SNOMED) codes, and free text to capture routine husbandry practices and veterinary medical care. All animals on which the electronic health records are based were assigned to the Oregon National Primate Research Center (ONPRC) breeding colonies and managed under the IACUC-approved protocol of the Oregon Health and Science University West Campus. All animals were tested annually for simian viruses (Simian Immunodeficiency Virus, Simian Retrovirus 2, Macacine herpesvirus 1, and Simian T lymphotrophic virus) and received a mammalian old tuberculin test semi-annually. 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.

Rhesus macaques in our study were Indian origin. Captive reared macaques in our study were group housed outdoors in open-earth corrals or sealed-concrete shelter housing. Figures for corral and shelter group housing are previously published (Prongay et al., 2013). Corral housing was 1 acre enclosures containing social groups of 125 to 250 rhesus macaques. Corral housing was composed of open-earth space with sheltered feed areas with cement floors, radiant heat, and cooling fans. Corral housing also contained play structures and dome-shaped polyethylene shelters for enrichment. Permanently installed sprinklers provided water for vegetation and a source of cooling water during hot weather. Surface drain areas within each corral and floor drains within each feed area prevented standing water. The feed areas were cleaned daily using water wash-down. Animals were observed twice daily by husbandry staff from observation towers, and an area walk-through was conducted daily. From towers, staff offered produce to entice animals to approach the tower where husbandry staff made observations using the naked eye or binoculars to assess animals for signs of illness. Animals that did not approach the observation tower during observation periods were assessed by husbandry staff using binoculars for signs of illness. Each infant in a corral was brought to the veterinary staff for evaluation if husbandry staff found the following signs of illness: dehydrated infants as evidenced by lethargy, weak grip on dam, not holding on with all limbs, sunken dull eyes, slow blinking or dragging eyelids, lack of alertness or interest in surroundings, more interested in sleeping than nursing, or a rough hair coat. Instances of observation of liquid stool were not sufficient for staff to remove infants from their social group for veterinary evaluation.

Shelter housing was either 3-bay, 1300 sq. ft. triplexes or 2-bay, 1000 sq. ft. duplexes containing 22 to 54 rhesus macaques. Play structures were located in each unit. The floors were sealed concrete. Structures were cleaned daily with water wash-down. Animals were observed daily by husbandry technicians and brought to the veterinary staff for evaluation if husbandry staff observed signs of illness which are listed in the previous paragraph.

One to four times per year, on a rotating schedule, all animals were sedated for a physical examination, pregnancy evaluation, weighing, mammalian old tuberculin skin test, antihelminth administration (Ivomec®, ivermectin 1%, Merial, Lyon, France), and blood draw for serology and genetic testing. Animals were fed LabDiet 5000 (Ralston Purina, St Louis, MO, USA) twice daily and supplemental produce or other enrichment once daily. Municipal water was available ad libitum.

Data collection

Weight records from the electronic health records of all Indian rhesus macaques less than 1 year of age in corral and shelter housing at the ONPRC between January 1, 2010 and December 31, 2014 were reviewed for this study as a cross-sectional data set. Weights records available for this study were collected from animals at any group sedation for semi-annual physical exams or at presentation to the hospital. Each weight record also had attached information including date, animal identification number, species, sex, birth date, housing at the time of weight record, and the full master problem list of clinical cases opened for the animal during their first year of life. The master problem list of clinical cases attached to each weight record could include “GI-Diarrhea.” Diagnoses and subsequent assignment of a “GI-Diarrhea” master problem were initially made by veterinary staff following an evaluation of an animal brought to the hospital by husbandry staff after illness observation in social group, and the master problem was entered by veterinary staff into the electronic health record system at the time of diagnosis. A manual review of 10% of all weight records was conducted.

Animal ages at time of weight recording were calculated from birth dates. Birth dates for shelter housed animals were the observed date of birth. Birth dates for groups in corral housing were estimated from dental eruption patterns (Hurme & van Wagenen, 1953). Although dental eruption patters have been demonstrated to be a strong predictor of age in rhesus macaques (Hurme & van Wagenen, 1953), it is important to point out that ages of animals estimated by dental eruption in corrals were inherently less reliable than that of animals born in shelter housing with exact birth dates (Wang, Turnquist & Kessler, 2016).

Weight records were separated by housing, sex, and occurrence of at least one diarrhea problem in the first year of life of infant rhesus macaque. These groups were healthy corral males, diarrhea corral males, healthy corral females, diarrhea corral females, healthy shelter males, diarrhea shelter males, healthy shelter females, and diarrhea shelter females. Healthy groups were composed of rhesus macaque infants that did not develop diarrhea during the first year of life that required veterinary intervention. Diarrhea groups included rhesus macaque infants who developed diarrhea requiring clinical treatment one or more times during the first year of life. Therefore, the diarrhea group of animals included animals with a single or multiple diarrhea episodes requiring hospitalization, so that animals with many diarrhea episodes did not carry more weight than an animal with one episode. Weight records designated in diarrhea groups may have been weighed at any point before or after a clinical problem of diarrhea was diagnosed in the first year of life.

Statistical Analysis

Multiple linear regression analysis was performed to obtain standard weight gain of rhesus macaques in the first year of life, examine an association between diarrhea and weight gain, and develop a lower 3^rd percentile cut-off to screen weight records for diarrhea in group housed infant rhesus macaques. Housing and sex were included in our analysis to account for an interactive effect of housing and sex with diarrhea-associated mortality in rhesus macaques (Hird, Anderson & Bielitzki, 1992, Prongay et al., 2013). The inclusion of housing as a dummy variable in the model also separated corral and shelters housing infant weights into separate growth estimates. The outdoor-housed infant rhesus macaque body weight model was defined by

$$E\left(Weight\right)={\beta }_0+{\beta }_{1}Age+{\beta }_2{I}_{sex}+{\beta }_3{I}_{diarrhea}+{\beta }_4{I}_{housing}+{\beta }_5{I}_{diarrhea}{I}_{housing}+{\beta }_6{I}_{diarrhea}{I}_{sex}+{\beta }_7{I}_{sex}{I}_{housing}+{\beta }_8{I}_{sex}{I}_{diarrhea}{I}_{housing}+{\beta }_{9}Age{I}_{sex}+{\beta }_{10}Age{I}_{diarrhea}+{\beta }_{11}Age{I}_{housing}+{\beta }_{12}Age{I}_{sex}{I}_{diarrhea}+{\beta }_{13}Age{I}_{sex}{I}_{housing}+{\beta }_{14}Age{I}_{diarrhea}{I}_{housing}+{\beta }_{15}Age{I}_{sex}{I}_{diarrhea}{I}_{housing}$$

Where $I_{\mathrm{s}}$ are indicator (dummy) variables.

$$I_{\mathrm{s}\mathrm{e}\mathrm{x}}=\{\begin{array}{l}0\mathrm{f}\mathrm{e}\mathrm{m}\mathrm{a}\mathrm{l}\mathrm{e}\\ 1\mathrm{m}\mathrm{a}\mathrm{l}\mathrm{e}\end{array},I_{\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{r}\mathrm{r}\mathrm{h}\mathrm{e}\mathrm{a}}=\{\begin{array}{l}0\mathrm{h}\mathrm{e}\mathrm{a}\mathrm{l}\mathrm{t}\mathrm{h}\mathrm{y}\\ 1\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{r}\mathrm{r}\mathrm{h}\mathrm{e}\mathrm{a}\end{array},I_{\mathrm{h}\mathrm{o}\mathrm{u}\mathrm{s}\mathrm{i}\mathrm{n}\mathrm{g}}=\{\begin{array}{l}0\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{r}\mathrm{a}\mathrm{l}\\ 1\mathrm{s}\mathrm{h}\mathrm{e}\mathrm{l}\mathrm{t}\mathrm{e}\mathrm{r}\end{array}$$

The three dummy variables included in our multiple regression model simultaneously categorize each infant record into groups differing by sex, health, and housing. To create these sets of dummy variables, we decided on a reference group, the healthy corral males. Dummy variables were created for the remaining groups, females, diarrhea, and shelter housing which are all coded 1 for infant records in those groups, all others are coded as 0. In the multiple regression model, the coefficients with each dummy variable are interpreted as the expected difference of the outcome variable (weight) from the reference group. The model is able to analyze the six groups simultaneously and yields coefficients which essentially equal the change necessary to switch predicted weight for age from our chosen base group where the dummy variables all are 0, healthy corral male infant group, to another group with another combination of dummy variables. The three binary dummy variables enable the model to account six different growth curves separately. Despite the appearance of just one multiple linear model used in our study, at no point in the regression analysis were shelter and corral housing data analyzed together.

Using the dummy variables indicated for healthy corral males, the model presents growth in the least amount of terms because we used zero to indicate male, healthy, and corral infants (healthy corral male group). This leads to most variable terms in the model canceling out due to the factor of 0 present in all but ${\beta }_0+{\beta }_{1}Age$ terms. Hence, healthy corral male growth becomes${\beta }_0+{\beta }_{1}Age$. Similarly, the growth of diarrhea corral males becomes simplified from our full model to ${\beta }_0+{\beta }_{1}Age$+ ${\beta }_3{I}_{diarrhea}$+ ${\beta }_{10}Age{I}_{diarrhea}$ because$I_{sex}=0$, $I_{diarrhea}$= 1, and $I_{housing}$ = 0. This enables the model to predict the growth of diarrhea corral males separately from healthy corral males because the model contains a dummy variable to discriminates the data and results of the growth of healthy corral males by the factors that differ between the healthy and diarrhea corral males, ${\beta }_3{I}_{diarrhea}$and ${\beta }_{10}Age{I}_{diarrhea}$. Therefore, we can test for the difference between growth of the two groups by testing the growth term that differs, ${\beta }_{10}Age{I}_{diarrhea}$and asking if this term’s coefficients ${\beta }_{10}$ is significantly not equal to zero. We used the F-statistic test to determine if coefficients of differing terms between groups were equal to 0.

After model fitting, model adequacy was assessed using residual analysis, normality checking, and influential observation detecting toll Cook’s distance. The model was considered significant if the F-statistic p-value was less than 0.01. A histogram of frequency of weight records per animal was generated to visually check if data could be considered cross sectional or longitudinal during the multi-year period that may have included more than one weight record per animal. Most of the rhesus macaques had only one to two weight measurements, and none of animals had more weight observations (less than 10 weight records per animal) than the number parameters in the infant weight model (15 model parameters). Each observation in this modeling was considered as independent even though some rhesus macaques were measured more than once.

Growth of healthy and diarrhea groups were compared using the F-statistic, and the deficiency of growth for each sex (the growth difference between healthy and diarrhea groups) were compared between shelters and corrals housing groups using the F-statistic. Two-tailed tests were used in all instances. P-values less than 0.01 were considered significant.

Standard growth was defined as the growth found in healthy groups in our body weight model. Cut-off criteria were made for growth faltering of infants using the lower 3^rd percentile of standard growth from our body weight model. Cut-offs were reported as equations with weight as a function of age for corral females, corral males, shelter females, and shelter males.

Model and cut-off performance with validating dataset

We evaluated the performance of our infant body weight model using a second separate set of infant weight records from January 1st, 2015 to December 31st, 2016 obtained from our electronic health record system in an identical method as the 2010 to 2014 infant weight records. This dataset was referred to as the validation dataset, as the data was not used to generate the model; the validating dataset was used to estimate the model performance with a new set of infant records. 2015 to 2016 weight records with infant ages, health, sex, and housing information were placed in the body weight model to obtain each record’s predicted weights.

Bland-Altman difference plot analysis was performed to quantify the agreement between predicted weights from our model and actual infant weights (Bland & Altman, 1999). We plotted a point for each infant weight record with y-axis corresponding to the difference between the predicted and actual weights and x-axis corresponding to the mean of predicted and actual weights. The mean difference of all points was obtained to detect bias of our model predicted weights. The 95% limits of agreement were defined as two standard deviations of the mean difference. Pearson’s correlation coefficient (r) measures the strength and direction of the association between our predicted weights and actual weights. Strong associations were expected between r values 0.9 and 1.0. Significance of r was tested with a two-tailed T-test, and P-values less than 0.01 were considered significant.

Odds ratios, sensitivity, and specificity were used to assess the cut-offs of growth faltering as tools to screen for diarrhea. Odds ratios for each group were the odds of a diarrhea weight record falling below the 3^rd percentile cut-off compared to the odds of a healthy weight record falling below cut-off. Ninety-five percent confidence intervals were reported for each odds ratio. Each odds ratio was tested for significance using Fisher’s exact test. P-values less than 0.01 were considered significant.

For sensitivity and specificity analysis, we binned the validating dataset weight records into four age groups, 0 to 90 days, 91 to 180 days, 181 to 270 days, and 271 to 365 days for shelter females, shelter males, corral females, and corral males. Sensitivity and specificity at each age group were calculated with growth faltering cut-offs to sort diarrhea and healthy weight records. A graphic description of the cut-off criteria, sensitivity, and specificity used in this study can be found in figure 1. Sensitivity, the true positive rate, is a measure of the ability of our cut-offs to correctly identify diarrhea weight records below the cut-offs. Diarrhea weight records below the cut-offs were considered true positives. Specificity, the true negative rate, is a measure of the ability of our cut-offs to correctly identify healthy weight records above the cut-offs. Healthy weight records above the cut-offs were considered true negatives. We used the 3^rd percentile from the mean of healthy weights as our cut-offs for growth faltering, so we expected our specificity rates for cut-offs to be near 94% in all groups. Sensitivity above 50% was considered better than chance alone to identify diarrhea weight records using the cut-offs and would possibly be an aid veterinarians in diarrhea screening by weight records.

Figure 1.

The growth faltering cut-off criteria performance in screening for diarrhea. Sensitivity is the true positive rate, a measure of the ability of our cut-offs to correctly identify those with diarrhea, and corresponded with diarrhea weight records less than the cut-offs. Specificity is the true negative rate, a measure of the ability of our cut-offs to correctly identify those without diarrhea, and corresponded with healthy weight records greater than cut-offs

SAS University Edition (SAS Institute Inc. 2016. SAS University Edition. Cary, NC,USA) and R statistical program (version 3.3.10; R Core Team 2016 R: A language and environmentfor statistical computing. R Foundation for Statistical Computing, Vienna, Austria; https://www.r-project.org/foundation) were used for data analysis.

Results

6536 weight records were obtained for the data set of infant rhesus macaques between 2010 and 2014. A manual review of 10% of records revealed a total of 25 weight records as having a high probability of an incorrect weight entry and were removed from the data set. A total of 1 duplicate weight record was removed from the data set. We assumed there was a 4% error rate among remaining weight records in our dataset. The remaining data set of 6510 weight records was used for statistical analysis. Table 1 shows the number of weight records for each group by sex, housing, and health.

Table 1.

Number of infant rhesus macaque weight records used in our multiple linear model

	Number of weight records
Group	Healthy	Diarrhea	Total
Corral females	1285	293	1578
Corral males	1246	316	1562
Shelter females	1161	584	1745
Shelter males	979	646	1625

Infants with at least one case of diarrhea within the first year of life were placed in diarrhea groups. Infants without a case of diarrhea within the first year of life were used in healthy groups.

Our multiple linear regression body weight model showed that independent variables age, sex, diarrhea, housing, and interaction variables were predictors of weight. Multiple linear regression analysis of our body weight model was significant (F(15,6509)=2974.49, P<0.0001) for predicting weight from our predictor variables. The model had a coefficient of multiple determination, R², of 0.882, indicating that all 15 variables in the model explain 88.2% of the variation in weight. Some of the unexplained variation may be due to incorrect weight observations that were not subject manual review.

Standard growth of healthy infant rhesus macaques and growth faltering

Healthy male and female infant rhesus macaque weight records in shelter and corral housing were used to establish standard growth of healthy infant rhesus macaques for comparison with diarrhea groups (Figure 2). The variation in prediction interval distance at all ages was negligible at less than 0.0005 kg, and the 3^rd percentile of standard growth was 0.311 kg subtracted from the 50^th percentile or mean weight-for-age in all healthy groups. Table 2 shows standard growth as functions with weight and age. The standard growth charts with 3^rd, 15^th, 50^th, 85^th, and 97^th percentiles are found in figure 3 and depict growth found in our healthy groups. In figure 3, corral groups were labeled as open earth substrate charts, and shelter groups were labeled as covered sealed concrete charts.

Figure 2:

Healthy (blue) and diarrhea (red) 2010–2014 weight records from outdoor housed infant rhesus macaques in corral or shelter housing. The dashed lines in each plot represent the 3rd and 97th percentiles of standard growth for male and female growth in either shelter or corral housing, corresponding to mean standard growth ±0.311 kg

Table 2.

Standard growth of outdoor housed infant rhesus macaques and cut-off criteria for weight records

	Infant growtha
Group	Standard	Growth faltering cut-off
Corral females	Weight = 4.0 × age + 471	Weight = 4.0 × age + 160
Corral males	Weight = 4.1 × age + 502	Weight = 4.1 × age + 191
Shelter females	Weight = 4.4 × age + 432	Weight = 4.4 × age + 122
Shelter males	Weight = 4.7 × age + 461	Weight = 4.7 × age + 150

^aGrowth is defined by weight in grams as a function of age in days. Standard infant growth and the lower 3rd percentile cut-off criteria for growth faltering were produced from a multiple linear model using weight records from healthy groups.

Figure 3.

Standard growth charts for male and female captive infant rhesus macaques raised in covered sealed concrete shelters housing or open earth substrate corral housing. Percentiles of standard growth by weight-for-age are provided

Associations of diarrhea and housing with growth of infant rhesus macaques

Significant deficits in weight gain of all diarrhea groups were found when compared to healthy groups (Table 3). In corral females, diarrhea growth was 0.9 g/day(22%) less than healthy growth (F-statistic: F(1,6494)=56.37, P<0.0001). For corral males, diarrhea weight gain deficits were less than corral females (−0.4 g/day; F-statistic:F(1,6494)=26.24, P<0.0001). Interestingly, shelter male and female diarrhea growth deficits were both greater (30% males and 34% females) than both counterparts in corral groups(shelter males: −1.4 g/day; F-statistic: F(1,6494)=267.48, P<0.0001; shelter females: −1.4 g/day; F-statistic: F(1,6494)=324.31, P<0.0001), and there was a multiplicative negative effect of diarrhea weight deficits in shelter housing compared to corral housing (females −0.5 g/day: F-statistic: F(1,6494)=10.48, P=0.0012; males −1.0 g/day F-statistic: F(1,6494)=38.53, P<0.0001).

Table 3.

Growth differences of healthy and diarrhea groups over the first year of life

	Infant growtha
	Healthy g/day (95%CI)	Diarrhea g/day (95%CI)	Difference g/day	%	p-value (F-statistic)
Corral Females	4.0 (3.8–4.2)	3.1 (2.6–3.6)	−0.9b	22	<0.0001
Corral Males	4.1 (4.0–4.2)	3.7 (3.5–3.9)	−0.4b	11	<0.0001
Shelter Females	4.4 (4.0–4.7)	3.0 (1.8–4.3)	−1.4b,c	34	<0.0001
Shelter Males	4.7 (4.5–4.8)	3.3 (2.7–3.8)	−1.4b,c	30	<0.0001

^aGrowth is defined by weight in grams as a function age in days. Healthy growth was measured with infants without a case of diarrhea in the first year of life. Diarrhea growth was measured with infants having at least one case of diarrhea.

^bSignificant differences in growth between healthy and diarrhea groups.

^cSignificant differences in growth deficits between shelter and corral housing groups.

Body weight model and growth faltering cut-off performance with validating dataset

1839 weight records from years 2015 and 2016 were obtained for the validating dataset. A manual review of 10% of records revealed a total of 2 weight records as having a high probability of an incorrect weight entry and were removed from the data set. No duplicate weight records were found in the validating data set. The remaining data set was used for statistical analysis as the validating dataset.

Bland-Altman plot analysis revealed a mean difference of +0.031 kg, a finding indicating our body weight model’s predicted weights were on average 0.031 kg less than actual infant weights they were predicting from infant ages, sex, housing, and health status (Figure 4). The 95% limits of agreement were ±0.351 kg from the mean, indicating 95% of predicted infant weights were within 0.351 kg of the actual infant weights. Pearson’s correlation coefficient (r) was run to determine the relationship between our 1837 actual and predicted weights. There was a significant and strong positive correlation between predicted and actual weights (r=0.94, n=1837, p<0.001).

Figure 4.

Bland–Altman plot of predicted and observed weights of infants from 2015 to 2016 used in a multiple linear model of body weight. The Pearson’s correlation between predicted and observed infant weights showed strong agreement (r = 0.94). The 95% limits of agreement were ±0.351 kg from the mean difference 0.031 kg

The validating dataset was used to demonstrate the capability of our growth faltering cut-off as a diarrhea screening tool in growth stunted infant macaques (Table 4). The odds of corral females falling below the cut-off was 11.2 higher given diarrhea history compared to healthy (OR 11.2; 95%CI: 4.2 – 30.2; Fisher’s exact test, P<0.0001; Table 4). Interestingly, the odds of falling below the cut-off given diarrhea compared to healthy for shelter females was higher than corral females (OR 13.2; 95%CI: 6.6 – 25.9; Fisher’s exact test, P<0.0001; Table 4). Similarly, the odds ratios for shelter males (OR 12.9; 95%CI: 6.5 – 25.4; Fisher’s exact test, P<0.0001; Table 4) were higher than corral males (OR 8.6; 95%CI: 3.5 – 21.2; Fisher’s exact test, P<0.0001; Table 4). Specificities for cut-offs of all groups were greater than 90% in all age groups (Table 5). Sensitivities for cut-offs in age groups before 6 months performed below 50% for all groups (Table 5). In age groups corresponding to infants between 6 and 12 months, shelter males and shelter female cut-off sensitivities were at least 50% (Table 5). Corral male cut-off sensitivities only peaked above 50% in the specific instance of corral males age group 181–270 days (Table 5). Corral male and corral female cut-off sensitivities in the age group 271–365 were all below 50% (Table 5).

Table 4.

Infant weight records from healthy and diarrhea groups of the validating data set (n = 1837) against the growth faltering cut-off criteria

		Infant weight records
	Below Cut-off		Above Cut-off
	Diarrhea	Healthy	Diarrhea	Healthy	OR	95% CI	P-value
Corral Females	14	6	72	346	11.2	4.2 – 30.2	<0.0001
Shelter Females	43	12	84	306	13.1	6.6 – 25.9	<0.0001
Corral Males	12	9	61	392	8.6	3.5 – 21.2	<0.0001
Shelter Males	46	11	101	322	12.9	6.5 – 25.4	<0.0001

The weight records below the cut-off criteria with a history of diarrhea were compared to those with a healthy history using odds ratios (OR). Fisher’s exact test calculated a p-value for significance of each odds ratio.

Table 5.

Sensitivity and specificity of the growth faltering cut-off criteria in dividing healthy and diarrhea infant rhesus macaque weight records by age group.

	Age (days)
Sensitivity	0–90 (%)	91–180 (%)	181–270 (%)	271–365 (%)
Corral females	0	6	30	32
Corral males	0	6	55	32
Shelter females	0	24	53	59
Shelter males	0	24	53	59
Specificity
Corral females	100	100	100	95
Corral males	100	98	100	95
Shelter females	100	98	94	92
Shelter males	100	100	95	91

Sensitivity is the true positive rate of the cut-off for growth faltering in detecting diarrhea weight records, and specificity is the true negative rate of the cutoff for growth faltering in identifying healthy weight records.

Discussion

We present weight gain standards through the first year of life for healthy captive rhesus macaques reared in outdoor natal social groups. Our results demonstrate linear weight gain through the first year of life for healthy male and female rhesus macaques in outdoor corral and shelter housing. While linear weight gain through the first year of life has been previously demonstrated for this species using small or single year cohorts (Gavan & Hutchinson, 1973, Small & Smith, 1986), our study is the largest and first to use a multi-year dataset with medical record information to demonstrate standard growth of captive outdoor housed infant rhesus macaques. Aided with medical record information from each weight record, we found diarrhea-associated growth deficits in infants who developed diarrhea in the first year of life. Standard growth charts were provided that can be used prescriptively to monitor infant weights in two outdoor housing situations: corral-like open earth substrate housing and shelters-like covered sealed concrete housing. We used the 3^rd percentile of standard growth to define growth faltering, and we demonstrated the utility of our growth faltering cut-off in diarrhea screening using an independent dataset of infant weight records. We found strong associations of growth faltering and diarrhea with OR of 11.2 and 13.1 for corral and shelter females respectively and OR of 8.6 and 12.9 for corral and shelter males. Finally, the cut-off for growth faltering performed well for diarrhea screening in shelter infants older than 6 months of age.

Current practices and housing limit technicians in their ability to identify diarrhea of individual animals that are group-housed in large enclosures like our corrals. Clinicians ideally could monitor infant rhesus macaque weights using the growth faltering cut-offs to screen for subclinical gastrointestinal problems and mild diarrhea affecting infant growth that is not easily detect by husbandry observation. Using indicators of diarrhea-associated growth faltering in conjunction with further confirmation of the presence of diarrhea in an individual infant animal could allow veterinarians and colony managers to identify infants with problematic diarrhea before these vulnerable animals have a life-threatening problem.

Cut-off sensitivities for diarrhea were lower in corral animals, perhaps due to the imperfect estimation of corral monkey ages based on dental eruption. While this method of age estimation is better than many other anthropometric metrics, it can be imprecise, especially for animals older than 180 days (Gavan & Hutchinson, 1973). We attempted to ensure measurement precision by establishing a formal training program and maintaining a written protocol for weight measurement and data entry. When the date of birth is unknown for a child, WHO guidelines have provided two methods for determining a child’s age (WHO, 1983); age of children could either be rounded to the month or estimated by the investigator. In either age estimation, it was prudent for researchers to report which estimation method was used. Indices such as cut-offs for growth stunting or growth faltering depend on precision of age, and in one study, age estimation lead to misclassification of subjects along WHO cut-off criteria due to systemic bias of the estimation methods that shifted distributions of weight-for-age data (Gorstein, 1989). Thus, our growth faltering cut-offs for animals in corral-like housing are then limited to use of those developing their own in house depending on each facilities age estimation method. Regardless of the systemic bias and the misclassification bias, our corral animal dataset age records by estimation were robust enough to identify diarrhea-associated growth faltering in the specific housing situation and to produce charts for standard growth, but refinement in age estimation is needed to yield a useful cut-off criterion for diarrhea screening. Future prospective research could improve upon our model by collecting accurate birth dates on animals in a corral-like setting.

Our study is the first published study to evaluate linear growth as a metric of enteric health in infant rhesus macaques. Many enteric pathogens associated with diarrhea are associated with villus blunting of the intestines. Villus blunting may result in poor nutrient absorption leading to decreased weight gain and growth stunting (Mondal et al., 2012). Diarrhea and secondary undernutrition is associated with static weight gain, growth stunting, and increased mortality in numerous species (Checkley et al., 2008; Keusch et al., 2014; Moore et al., 2010; Petri, Miller, Binder, Levine, Dillingham & Guerrant, 2008). Our results align with similar studies, and support our hypothesis that rhesus macaques who develop diarrhea during the first year of life experience poor weight gain and growth stunting.

Failure of linear growth is associated with poor cognitive and motor development in children in low and middle income countries. This subclinical condition, called environmental enteropathy or environmental enteric dysfunction is induced by low level chronic immune stimulation caused by exposure to viral, bacterial, and parasitic pathogens both in utero and during growth and development (Keusch et al., 2014; Ngure, Reid, Humphrey, Mbuya, Pelto & Stoltzfus, 2014). Infant macaques, like human infants, have high oral exploratory drive and naturally ingest some fecal material from the environment. This behavior may be essential for development of healthy enteric flora in both species. However, in vulnerable infant humans and macaques exposed to dangerous enteric pathogens in the environment, the behaviors may lead to chronic inflammation and environmental enteropathy. Future studies, focused on gut inflammation, pathogen exposure, and microbiomes found in stunted rhesus macaques, will be necessary to evaluate the species as a natural model of environmental enteropathy.

The WHO uses infant weight standards as a metric for measuring growth in optimal conditions, and to identify infants and children at risk for stunting and disease (The WHO Multicentre Growth Reference Study, 2006a). While similar standards were not available for rhesus macaques, multiple references demonstrate linear weight gain during the first year of life for both wild and captive macaque populations (Michejda & Watson, 1979; Saxton & Lotz, 1990; Schultz, 1933; Small & Smith, 1986; van Wagenen & Catchpole, 1956). Much of the early published primate literature quantified growth and development in research settings. This normative data provided initial morphometrics for first-year growth of the rhesus macaque and has been evaluated as a means to estimate age (Gavan & Hutchinson, 1973; Michejda & Watson, 1979; Saxton & Lotz, 1990; Schultz, 1933; van Wagenen & Catchpole, 1956). Infant weights among rhesus macaques at Cayo Santiago demonstrated sex differences in weight (Schwartz & Kemnitz, 1992). Nutrition, hybridization, climate, and geographic location have also already been shown to impact growth patterns (Clarke & O’Neil, 1999; DeRousseau & Reichs, 1987; Paterson, 1996; Smith, Lorey, Suzuki & Abe, 1987; Taylor & Schillaci, 2008). Small and Smith (1986)’s study established a reference range of growth for infant rhesus macaques at the California National Primate Research Center (CNPRC) using open-earth enclosures for housing which is similar to the corral housing at the ONPRC. In addition to differences in time of study, location, and population, the methods of defining the reference range differed from our study. Smith and Small removed CNPRC infant weight records from their study if the weight record fell outside 2SD of the average age estimates. In our study, ONPRC infant weight records were only removed from our standard weight growth curve when diarrhea, a potential confounder of outlying weight measurements, occurred in the first year of life.

Weight is an easily obtained metric and could allow early identification and intervention of infants affected by diarrhea. Little additional personnel time, equipment, or training is needed. It provides a rapid and easily translatable way to assess growth during the first year of life. As research focus on infant growth and development increases, it will become increasingly important to collect additional morphometric data to accurately assess changes birth and growth in both research and control populations.

A major goal of our study was to generate accurate growth charts based on sex and location in order to identify infants that fall out of the expected growth curve, as well as evaluate the impact of diarrhea. In order to achieve a powerful and translatable model with a large sample size we utilized a retrospective study design, and only evaluated variables that could be easily attainable and broadly applicable (i.e., sex and housing type). Additional variables that may influence weight and infant diarrhea that were not utilized in the model include parity (Broadhurst & Jinks, 1965; Bercovitch, Lebron, Martinez & Kessler, 1998; Wilson, Gordon & Bernstein, 1978), specific breeding group, pedigree (Kanthaswamy, Elfenbein, Ardeshir, Ng, Hyde, Smith & Lerche, 2014), dam age, dam weight (Bercovitch et al., 1998; Wilson et al., 1978), milk quality, maternal behavior, and environmental stress. In the current study we were only interested in basic, broadly applicable predictors of infant weight, however future research could use these variables to create a more precise predictive model of infant weight gain.

Additional morphometric measurements like body length, head circumference, biparietal diameter, and limb length would provide a more nuanced picture of infant growth. Growth velocity, a variable derived from weight measurement over time, would have accounted for individual velocity changes, which our linear model could not (Bozzola & Meazza, 2012). Electronic health records may facilitate inter-facility comparisons, which could lead to meta-data standards with primate facilities around the world. This will facilitate the development of multicenter growth references, similar to the WHO Child Growth Standards (The WHO Multicentre Growth Reference Study, 2006a). Multi-facility studies would allow more robust growth standard for healthy infant rhesus macaques with wide applicability.