Construction of a comprehensive observer-based scale assessing aging-related health and functioning in captive rhesus macaques

Screening of the aging-related items to assess overall health and daily functioning

Initially, 15 items referring to perceived aging, physical aging and social aging were selected according to the consensus of the panel. Preliminary scaling results based on daily observations resulted in ten items that could be clearly discriminated with acceptable intraobserver agreement. The scale containing 12 items with each point description is shown in Table 1. The lower the score was, the milder the aging process and the better the health and functioning of the animal. All screened items were graded on a three-point scale with the exception of tooth loss. An optimization-based development on “Age” and “Sex” was performed based on the comparison of the respective scales’ Cronbach’s alpha coefficient. The scale scoring by “chronological age” or “categorized age” as well as scoring “male with lower score” or “female with lower score” underwent optimization. The age-dependent category was divided according to the percentile of chronological age.

The profile of aging-related health and functioning in the rhesus macaque group assessed by the AHFS

The fundamental profile of aging-related health and functioning scored by the AHFS in the present rhesus macaque group are shown in Table 2. From the table, we can see that the majority of animals were in the “Adult category”. Notably, all these scores were assessed based on daily life performance, which indicated the small time commitment and relative noninvasive merit. A very short period of anesthesia was required to score the tooth condition and iliac mass in the male animals.

The interrelationship of the aging-related items in the scale

Using Pearson correlations, we further analyzed the interrelationship of the scored items on aging-related health and functioning in the AHFS (Table 3).

On the present scale, the most interrelated domain was four items referring to dietary uptake. As the only “officially” social activity in the captive group, dietary uptake behaviors were supposed to parallel psychosocial behaviors during aging and were consistent with the findings in humans [14, 15]. More interestingly, the order of dietary intake displayed a considerably high interrelation with other items, further supporting our proposal of the great significance of social behaviors. The second most highly interrelated item was iliac muscle mass. The loss of muscles commonly occurred with variation in location and time of onset [16]. Therefore, we selected the iliac muscle, characterized as the largest and most easily palpated muscle, to grade the overall muscle condition. Hair fragility and hair condition also reached an approximately 0.7 correlation with overall scored health and functioning. We subjectively scored hair conditions and objectively measured hair fragility using an indirect method for increased accuracy. During the fragility measurement, we experienced a clear discrimination between fragile hair with easy breakage and tough hair with difficult breakage. Finally, facial redness displayed a mild and significant correlation with the scored health and functioning. Perceived age via facial appearance has been found to be an independent biomarker correlated with physical and cognitive function as well as survival in humans [17, 18]. As no method assessing perceptive age in nonhuman primates had been reported, we rated the darkness of facial redness, mirroring skin aging in the rhesus macaques, for greater sensitivity to perceived aging (Figure 1).

Reduced aging-related health and functioning was expected to be in parallel to the increased darkness of facial redness. The high interrelationship of items in the scale indicated both convergent and discriminant validity. The lowest correlation was shown for chronological age. Only dental condition was found to be correlated with chronological age. This finding suggested that chronological age was a poor determinant of the complexity of multidimensional nature of health and functioning.

The reliability of the AHFS in rhesus macaques

The reliability was 0.877 (Cronbach’s alpha coefficient), indicating that the items in the scale measure as a latent construct with 87.75% internal consistency. Notably, during the development of the scale, we compared Cronbach’s alpha coefficient by denoting the male or female animals as milder aging (score=1). We found that when the female sex was rated with a milder aging score, higher reliability scores were obtained. Because higher socioeconomic status has been closely correlated with better health during aging in humans [14], we supposed that the male animals with higher levels of social dominance might display better health. However, according to the reliability test, females performed better with slower aging based on the daily life assessment.

To further validate the reliability of the scale, we estimated half-split reliability by randomly splitting the 12 items into two halves, each comprising six items. Half one included four items on dietary uptake, age and sex, and the remaining six items were in half two. The reliability of half one was 0.831, whereas that of half two was 0.687. From these results, we suggested that dietary uptake performance played a substantially independent role in aging-related health and functioning.

The validity of the AHFS in rhesus macaques

To validate the construction of the scale, we conducted an exploratory factor analysis of the 12 items. The adequacy of the scale was calculated by KMO (0.815), which indicated that these items harbor a dimension-reducible construct. We then divided the 12 items into three distinguishable factors using varimax rotation (Figure 2A). According to the results, three clustered factors together explained 76.4% of the variance in aging-related health and daily functioning. We further estimated the contribution to each factor using the “loading” statistic (Figure 2B–2D). Loadings above 0.5 were considered to contribute to a relatively independent role, helping “denoise” the complexity of aging-related health and functioning. Factor 1 comprised six significantly interrelated items, among which volume of dietary uptake explained the highest level of variance of the 12 items in the AHFS.

Evaluating the applicability of the AHFS using hierarchical cluster analysis

Hierarchical cluster analysis was conducted to validate the actual applicability of the scale. Individuals’ scores for aging-related health and function were clustered in all subjects as well as within each sex. According to the comparison of the respective presented dendrograms, three clusters were taken, and the optimal number of clusters reflected the severe-aging clustered subgroup, moderate-aging clustered subgroup and mild-aging clustered subgroup. Notably, the average scores of three clustered subgroups displayed significant disparity from each other in all samples, all females and all males (Table 4).

The clusters of individual scores in all samples, females and males, were also demonstrated by tree plots (Figure 3). The magnitude of an individual’s score for aging-related health and functioning is represented by the average interval between clusters according to the sequential similarity. In general, the discriminatory capability of the scale validated the applicability of the AHFS as good.

Furthermore, the scores of each item in the three clustered subgroups were also estimated and compared (data not shown). The similar clustered patterns between these composite items and the summed score were revealed, which indicated that the scale could be well separated with respect to the similarity of its composite items.

Healthy animal screening

We observed all laboratory-maintained group-housed rhesus macaques in the North Primate Research Center affiliated with the Institute of Laboratory Animal Sciences, Chinese Academy of Medical Sciences & Peking Union Medical College from November 2018 to March 2019. Approximately two-thirds of the rhesus monkeys aged 4 to 20 were initially screened as candidates. A total of 57 rhesus monkeys identified as consistently healthy animals with normal functioning were maintained to construct a scale termed the Aging-related Health and Functioning Scale (AHFS). The animals with transient abnormalities, including decreased dietary uptake, diarrhea, fatigue and solitary habits during the observation, were excluded. All selected animals were research-naïve and pregnancy-free during the observation. The sex ratio was 1:1.04 (M:F=28:29). All subjects were of known age, parity, mass, and social rank according to colony records. All protocols were approved by the Institutional Animal Care and Use Committee of the Institute of Laboratory Animal Sciences, Chinese Academy of Medical Sciences & Peking Union Medical College.

A step-by-step selection of items under panel consensus

A panel including three researchers in the field of human aging, two researchers who study nonhuman primates and five experienced rhesus macaque caretakers was assembled. A step-by-step item selection was performed to screen candidate items for the AHFS. First, the three human aging researchers were consulted to present candidate domains and related items in reference to human aging. Then, items that could be easily differentiated and were relatively noninvasive and occurred in daily life were selected by the two nonhuman primate researchers. Finally, the five rhesus macaque caretakers carried out the preliminary assessment to determine which items could be practically rated. A general 3-point scoring system was used with minor exceptions. The higher the score was, the more severe the age-related aspect of health and functioning.

Aging-related health and functioning observations and scoring assessment

For the rating of each rhesus monkey, a label designed for the grading week was highlighted. Five experienced caretakers who had undergone training carried out the scale rating as a team. Operationally, two randomly assigned caretakers independently rated each of the candidate animals. If the score disagreed between the two raters, a third caretaker was asked for a rescore. The final score was determined by the agreed upon score among the three raters. If no agreeable score was reached, the score of the animal was determined by consensus.

Perceived aging was graded by facial appearance and hair condition. The darkness of facial redness was subjectively qualified. Hair condition was subjectively graded by glossiness and season-unrelated overall hair loss. Hair glossiness was scored under natural light at a relatively fixed position from a fixed angle. The perceived severity of abnormal hair loss increased the item by one point. Hair fragility was scored by an easy experiment. Three pieces of hair were collected from the spine near the neck. Then, each hair was stuck with glue to a prepared piece of tape with 2 mm extending from the end of the tape. The average force to tear the hair away from the piece of tape was recorded. The extent (easy/moderate/difficult) to which the hair was torn from the piece of tape was used to score hair fragility. Musculoskeletal condition was graded by teeth and representative iliac muscles. The tooth condition was scored by the presence of tooth loss and by the extent of tooth wear. The iliac muscle was assessed by palpating the muscle mass [13] in addition to observing muscle movement. Four items were rated during dietary uptake, including the volume of dietary uptake, speed of dietary uptake per meal, size of the cheek pouch and order of eating the meal. The raters observed three occasions that an animal was eating meals in the group. The dietary uptake performance of each animal was observed for at least three meals, with each observation lasting no less than 10 minutes. The item score was averaged from the three observations.

Statistical analysis

All observed results on an ordinal scale were analyzed as continuous values. The mean ± standard deviation (SD) was calculated between repetitive assessments. Using the 25^th and 75^th percentiles as cutoffs, three age groups were defined: junior (8–12 years old), adult (13–16 years old) and senior (17–20 years old). Pearson correlations were used to determine the relationship between the scored AHFS and each of the rated items as well as between pairs of items. In order to examine the reliability of the scale, Cronbach’s alpha value and Cronbach’s alpha half-split value was calculated. In order to validate the construct of the scale, an exploratory factor analysis was performed. The number of underlying exploratory factors in the scale was determined by eigenvalues. Varimax rotation was performed to achieve the simplest structure of these exploratory factors. The factor loadings of each item within its exploratory factor were extracted within the rotated structure. In order to validate the applicability of the scale, a hierarchical cluster analysis was carried out based on the individual’s scored aging-related health and function. The optimal number of clusters was determined based on a comparison of the respective presented dendrograms. Similarity was measured based on the interval of intraindividual scores within each group. All analyses were conducted in SPSS (17.0) and R (3.5.0). A P value less than 0.05 was considered statistically significant.