A longitudinal study of radiographic spinal osteoarthritis in a macaque model

Abstract

Cross-sectional analyses of naturally-occurring spinal osteoarthritis (OA) in primates have shown that age and body mass are significant predictors, but whether or not these relationships hold true in longitudinal evaluations remains unclear. Because spinal OA manifests similarly in humans and monkeys and macaque monkeys age > 3 times the rate of humans, macaque models offer opportunities for longitudinal study that are difficult in humans. Our objective was to characterize the longitudinal development over 11 years of spinal OA in 68 Macaca mulatta (41 males, 27 females, aged 11-32 years). Average disc space narrowing (DSN) and osteophytosis (OST) scores were computed for the thoracolumbar spine (T8-L7). Our longitudinal analyses confirmed the cross-sectional results: age and body mass (p <0.001) significantly predicted 50% and 39% of the variability in OST and DSN, respectively. Rates of change in DSN, but not OST, were associated with age at first radiograph. This study represents the first long-term longitudinal assessment of OA in primates and establishes that the relationship among the covariates in the cross-sectional and longitudinal approaches is similar.

Introduction

Osteoarthritis (OA) is a serious and debilitating health problem with a growing incidence worldwide. In the US in 2005, the number of people with OA was estimated at 27 million (12.1% of the US adult population) with an annual health-care cost of approximately $89.1 billion (1), a pattern of morbidity that is appearing in developing countries as well (2). Osteoarthritis affects various joints throughout the human body, most commonly in the knee, hand, hip and spine. Spinal OA is associated with an assemblage of diagnostic radiographic features that include narrowing of the intervertebral disc space, formation of osteophytes along vertebral margins, and vertebral endplate sclerosis of the spinal column (3; 4). Evidence of spinal OA in human prehistoric and historic populations establishes the long history of the disease (5-8), and spinal OA has been found in all primate species that have been examined (9-16), yet the etiology and the pattern of progression over the lifetime of an individual is still not well understood.

Human studies produce cross-sectional, clinically-based or cadaveric evidence that precludes the study of the onset and early characteristics of the disease (17-20). Population-based studies are also either cross-sectional or have limited follow-up (e.g. 21; 22-25). Due to the health risks of repeated radiographic exposures as well as the prohibitive costs, longitudinal studies in humans are not feasible, but animal models allow for non-clinical, longitudinal research. Macaque monkeys are frequently used as models of human disease due to phylogenetic, physiological and behavioral similarities to humans (16; 26; 27). While macaques are habitually quadrupedal, they frequently sit upright with a vertically oriented spine, especially when caged. Macaques age at a rate > 3 times that of humans (9; 13; 14) but exhibit a naturally-occurring form of spinal OA that is radiographically similar to that seen in humans (12; 16; 26).

The goal for this study was to determine if and the extent to which age and body mass are associated with changes in spinal OA over time in individuals. Previous in vivo or ex vivo studies of OA in primate models (e.g. 9; 12; 13-15; 28-30) utilized cross-sectional approaches. Longitudinal data are, however, essential to study the features that develop over time because growth and development in human and nonhuman primates is typically nonlinear and often characterized by periods of stasis, followed by rapid change. The progression of OA appears to be age-dependent and may be non-linear, making longitudinal confirmation of cross-section findings very important.

Our longitudinal validation that a cross-sectional approach to the study of spinal OA is appropriate benefits similar studies in humans. Long-term longitudinal radiographic studies in humans are necessarily limited due to the risk of injury to the participants from X-ray exposure and the expense and uncertainty of long-term follow-up. Consequently, most studies in humans have been restricted to cross-sectional investigations. With our longitudinal corroboration of the known risk factors of age and body mass in monkeys, future human research can focus on targeted cross-sectional investigations of other factors that may contribute to spinal OA, such as genetic, hormonal, and lifestyle/activity-related influences, without as much concern that cross-sectional approaches might miss important variables that contribute OA etiology or progression.

The longitudinal data described herein are unique in that our data include repeated measures obtained over more than a decade, where external factors (e.g. environment, diet, activity levels) were held relatively constant throughout the observation period. This allowed us to assess whether or not longitudinal data support the correlation of age and body mass to the disc space narrowing (DSN) and osteophytosis (OST) seen in previous cross-sectional studies and to determine whether or not the rate of change of DSN and OST varies. Here, we examine the relationships using longitudinally-obtained radiographs of Macaca mulatta spanning 11 years. We hypothesize that when external factors are held constant, age at time of radiography and average body mass will be significant predictors of spinal OA.

Methods

Study Subjects

The study sample consisted of male (n=41) and female (n=27) rhesus macaques (M. mulatta), aged 11-20 yrs at the start of the study period and aged 22-32yrs by the end. All animals were housed individually under the same conditions at the Wisconsin National Primate Research Center (WNPRC) and have known birthdates, pedigrees, and complete medical histories. These animals were born in captivity at the WNPRC and before being enrolled in this study were either singly caged, or lived in small group cages. Animals were supplied with food, water ad libitum, and environmental enrichment objects according to IACUC standards for laboratory housing (13). Body mass measurements (in kg) were obtained at the time of radiography and an average adult body mass (mass) was calculated for each animal. Body mass index (BMI), calculated for macaques as the crown-to-rump measurement over mass, was also considered but not included as it was not more predictive than mass alone. These monkeys are part of an ongoing study of the affects of long-term dietary restriction on morbidity and mortality in primates (31; 32). For all radiographic markers of OA, the difference between control and restricted animals was non-significant and “diet” was not a predictor of OA in multilinear analyses controlling for age, body mass, and sex. The prevalence of OA in animals that died earlier in the study was not different from that of those which survived. Consequently, the two groups were combined and are discussed as a single group.

Radiographic Assessment

Biannual radiographs of the lateral and anteroposterior thoracolumbar spine were assessed for each animal spanning a period of 11 years. The dataset contained a set of repeated measures for each monkey - 2 to 6 radiographs per individual - with a total of 322 radiographs in the entire sample. A complete suite of radiographs for each animal found OA in the peripheral joints as well, but by far the highest prevalence and most severe incidence (trauma notwithstanding) was in the spine.

All images were obtained onsite at WNPRC using the same equipment, technicians and protocols. Spinal radiographs were read and scored by a single observer (AED), blinded to the animal's age, body mass and sex. Each vertebral level was scored for DSN, defined as a change in disc height relative to the adjacent intervertebral spaces, and OST, defined by the presence of osteophytes on the anterior vertebral margin. Due to vertebral size variability among individuals and between levels and variability in spinal positioning, an atlas scoring method was used: 0 for unaffected sites; 1, 2 or 3 for slight, moderate, or severe involvement, respectively, as has been done previously (15; 33).

Radiographic clarity along the thoracic vertebrae varied considerably above T8; therefore, evaluation was restricted to the intervertebral spaces from T8/T9 through L6/L7 for DSN and vertebral bodies T8 to L7 for OST. The lumbosacral joint (L7/S1) was not scored for DSN due to variability in both height and shape of the disc (17; 34). Similarly, the radiographs varied in quality and clarity across the spine throughout the sample, which prohibited the scoring of some intervertebral spaces. This resulted in variable sample sizes at individual sites. Sample sizes for DSN ranged from 162 to 322 and for OST from 123 to 322.

Analyses

Several scores were computed for each animal: average scores across the entire spine for DSN (T7/T8 – L6/L7) and OST (T8-L7); thoracic averages for DSN (T7/T8–T11/T12) and OST (T8–T12); lumbar averages for DSN (L1/L2–L6/L7) and OST (L1–L7); thoracolumbar joint averages for DSN (T10/T11-L2/L3) and OST (T11–L2). No combined score between OST and DSN was calculated. Individual rates of change in OA were defined as the slope obtained from a regression analysis with age as a predictor of DSN and OST for each animal. This slope for age represented the rate of change across an individual animal's radiographic timeline.

The relationship of OA with potential covariates, including age at the time of radiography (age), mass, sex (female=0; male=1) and age at first radiograph, was established using uni- and multi-linear regression analyses with correction for repeated measures within individual animals (StataCorp. College Station, TX). Statistical significance was established as an alpha of p ≤ 0.05.

A cross-sectional analysis was also performed to allow for direct comparison to a cross-sectional study of M. nemestrina (17), which had analogous methods. To create the pseudo-cross-sectional sample, a single radiograph per animal was randomly selected from our longitudinal dataset (N=68). Unilinear and multilinear regression analyses were performed using the same potential covariates as for the longitudinal data.

Results

Male monkeys were significantly older and heavier than females (p<0.000) (Fig.1) and had higher average scores for both DSN and OST (Table 1). Radiographic features of DSN and OST were evident at all ages within the sample. In both male and female monkeys, average scores for DSN and OST increased significantly with age (Fig. 2) and mass (Fig.3). Age at time of radiography, mass, and sex were significant predictors of OA in unilinear analysis of the overall spine as well as regional sections (thoracic, lumbar, and thoracolumbar joint) (Table 2). There were no significant differences between spinal regions in regard to OA scores. Multilinear analyses found that age and mass explained 39% of the variation in DSN and 50% of the variation in OST in the overall spine. No sex bias was apparent in the multivariate analysis.

Figure 1.

Age (A) and body mass (B) distributions in female and male macaques.

Table 1.

Group Characteristics.

	Males	Females
N animals = 68	41	27
N radiographs = 322	183	139
Age in years (range)	20.5 (11.46-31.49)	18.7 (10.92-26.60)
Average body mass in kg (range)	12.27 (8.70-19.06)	8.50 (6.24-11.23)
Average DSN scores 0-3 (range)	1.15 (0.00-2.80)	0.57 (0.00-2.53)
Average OST scores 0-3 (range)	1.44 (0.00-2.91)	0.90 (0.00-2.32)
Average rate of change in DSN score/year (range)	0.10 (-0.12 – 0.71)	0.06 (-0.11 – 0.21)
Average rate of change in OST score/year (range)	0.07 (-0.04 – 0.60)	0.09 (-0.02 – 0.19)

Figure 2.

Linear relationship of age with disc space narrowing (A) and osteophytosis (B) in females and males over time. Each point represents one measure within an animals's set of radiographs, connected by light grey lines. The darker line represents the fitted line for overall model including age and average adult body mass.

Figure 3.

Linear relationship of average body mass with disc space narrowing (A) and osteophytosis (B) in females and males over time. The darker line represents the fitted line for overall model including age and average adult body mass.

Table 2.

Correlation results for longitudinal (a) and cross-sectional (b) analyses of M. mulatta.

A. Longitudinal Data		Unilinear Regressions						Multilinear Regressions
Dependent Variables		age		body mass		sex		age		body mass
	n	r2	p	r2	p	r2	p	r2	p	p
DSN
entire spine	322	0.29	<0.000	0.13	<0.000	0.14	<0.000	0.39	<0.000	<0.000
thoracic spine	310	0.12	<0.000	0.04	<0.000	0.03	<0.048	0.16	<0.000	<0.014
lumbar spine	322	0.29	<0.000	0.13	<0.000	0.16	<0.000	0.39	<0.000	<0.000
thoracolumbar joint	322	0.20	<0.000	0.08	<0.000	0.11	<0.000	0.27	<0.000	<0.003
OST
entire spine	322	0.34	<0.000	0.21	<0.000	0.18	<0.000	0.50	<0.000	<0.000
thoracic spine	317	0.16	<0.000	0.14	<0.000	0.08	<0.003	0.27	<0.000	<0.000
lumbar spine	322	0.36	<0.000	0.17	<0.000	0.18	<0.000	0.49	<0.000	<0.000
thoracolumbar joint	322	0.34	<0.000	0.12	<0.000	0.09	<0.001	0.43	<0.000	<0.000

B. Cross-sectional Data		Unilinear Regressions						Multilinear Regressions
Dependent Variables		age		body mass		sex		age		body mass
	n	r2	p	r2	p	r2	p	r2	p	p
DSN
entire spine	68	0.28	<0.000	0.16	<0.001	0.14	<0.001	0.49	<0.000	<0.000
thoracic spine	63	0.09	<0.020	0.13	<0.004	0.04	<0.134	0.24	<0.001	<0.000
lumbar spine	68	0.32	<0.000	0.11	<0.005	0.17	<0.001	0.49	<0.000	<0.000
thoracolumbar joint	68	0.14	<0.002	0.18	<0.000	0.13	<0.003	0.35	<0.000	<0.000
OST
entire spine	68	0.22	<0.000	0.19	<0.000	0.18	<0.000	0.48	<0.000	<0.000
thoracic spine	63	0.04	<0.111	0.19	<0.000	0.11	<0.006	0.25	<0.010	<0.000
lumbar spine	68	0.31	<0.000	0.13	<0.003	0.16	<0.001	0.49	<0.000	<0.000
thoracolumbar joint	68	0.20	<0.000	0.14	<0.002	0.09	<0.011	0.39	<0.000	<0.000

Age at first radiograph (10.92-21.69 yrs), sex, and mass were used to examine variability between individuals in the individual rate of change in DSN and OST. Average rates for DSN were higher (p<0.006) in males than females, while average rates for OST were higher (p<0.009) in females than males (Table 1). The results of the multilinear regression analyses found only age at first radiograph to be a significant predictor for rate of change in DSN (p<0.014).

The relationship between age and DSN and OST in our cross-sectional sample was similar to that found by others (17) (Fig.4). Uni- and multilinear regression analyses showed effects of age and mass (Table 2) and the relationships were similar among the three samples: our longitudinal, our cross-sectional, and the previous cross-sectional analysis (Table 3).

Figure 4.

A comparison of cross-sectional data from two macaque species, M. mulatta and M. nemestrina, showing disc space narrowing (A) and osteophytosis (B) with age. The data from the present study of M. mulatta have been overlaid on that of the methodologically comparable study of M. nemestrina from Kramer et al. 2002(fig 3).

Table 3.

Comparison of results from our data with that of Kramer et al, 2002

	This study Longitudinal (N=68 animals/322 radiographs)	This study Cross-sectional (N=68 animals)	Kramer et al. 2002 (N=192 animals)
DSN
Variation due to age and body mass	39%	48%	57%
N	66 animals/268 radiographs	58 animals	109 animals
intercept	-1.87	-2.83	-1.11
age slope	0.10	0.12	0.08
body mass slope	0.08	0.12	0.11
OST
Variation due to age and body mass	50%	48%	61%
N	68 animals/315 radiographs	67 animals	73 animals
intercept	-1.45	-1.62	-0.96
age slope	0.09	0.09	0.09
body mass slope	0.09	0.11	0.07

Discussion

The nonhuman primate model offers an exceptional opportunity for a longitudinal study of spinal OA. Our 11 years of radiographs for the monkeys represent an equivalent of > 30 years in humans, because monkeys age at a rate > 3 times that of humans. The goal for this study was to determine if longitudinal data support the results obtained from cross-sectional data and our results do confirm this relationship. Age is a strong predictor of OA. Consequently, OA could be a condition of senescence and the product of the deterioration of tissues over time. Body mass and obesity have also been linked to OA in humans (35), so biomechanical loading could also increase tissue degradation (29). Some authors (15; 36) have argued, however, that if OA is solely a product of an individual aging, then a more uniform pattern of deterioration should exist across the spine and the lack of this uniformity implicates other causes. In our sample, despite the significant increase in prevalence with age, the range of OST and DSN scores vary widely at every age. Additionally, examination of the rates of change in OA did not add to our understanding of the role of any of these variables.

Heavy work-related jobs, sports, and repetitive stress movements have been shown to be linked to OA in humans (4; 36-38), which may support the loading hypothesis. Those OA studies that have included body mass, however, have produced mixed results. In humans, obesity and body mass index have been significantly associated with OA of the limb and hip joints (21; 35; 39) and with spinal OA (20; 25; 40), but not in all studies (37). In nonhuman primates, some studies (15; 28) have found body mass was associated with OA, but others (13; 14) found no association. Kramer et al. (15) found the association only with DSN but not OST and they postulated that because body mass and age are significantly correlated in their macaque sample, body mass might function as a partial proxy for age. Due to the lack of association between age and body mass in our longitudinal sample, that body mass is a predictor of both DSN and OST may be evidence that it is functioning as a truly independent variable.

Macaca is a sexually dimorphic genus where males are larger than females. While sex was significant as a univariate predictor, it was non-significant in our multivariate analysis, potentially due to the correlation of sex with mass in macaques. The relationship between OA, sex and mass suggests several additional lines of inquiry that we will address in the future.

The 48% of cross-sectional variation in DSN and OST explained by age and mass was similar to the 57% of DSN and 61% of OST found by others (15) in M. nemestrina. The longitudinal data also showed similar percentages explained by age and mass: 39% of variation in DSN and 50% in OST. Although less of the longitudinal variability in DSN was explained, we still feel that the longitudinal results are supportive of the cross-sectional approach.

Disc space narrowing assessment can be more subjective than that of OST (36). From the point of view of the observer, the bony stages of osteophytes are relatively similar, have been well described in the literature (17; 41; 42), and are relatively independent of the subject's idiosyncratic spinal anatomy. In contrast, the intervertebral space is evaluated relative to adjacent disc spaces, which are highly idiosyncratic and are naturally variable along the spine, being much narrower in the thoracic region compared to the lumbar. Positioning can also affect the shape of the space, unless there is osteophytic bridging. In those cases, bridging can maintain a space even in the absence of the disc. Intervertebral disc space is, therefore, more variable in individuals and OST is more indicative of overall correlation between the longitudinal and cross-sectional approaches.

Longitudinal data provided an opportunity to examine the rate of change over time in OA. In primates, growth and development rates typically follow a sigmoid curve. For example, body mass increases rapidly in youth, but then slows or stops accumulating in adulthood. OA follows a similar sigmoid pattern, although the rapid changes appear to occur in adulthood (unpublished data). Although animals in this study were adult (> 10 yrs) when enrolled, using age at the age at first radiography allowed us to locate an individual along the growth curve. In the comparison of rates between individual animals, the only predictor of significance was age and only for DSN. Rates were not predicted by sex or mass. Unfortunately, we lacked sufficient data to accurately describe the initial stages of the rate curve for each animal, allowing for only a linear rate estimate. Our estimates, however, are comparable to those of DeRousseau (9), who found the rate of change for degenerative joint disease in M. mulatta was 0.06 per year, similar to our values of 0.06-0.10 per year. Our linear estimates ranged broadly and are best considered as a preliminary examination of tempo in regard to OA progression. We chose to include them because we felt it important to acknowledge this key aspect of longitudinal datasets. Fortunately, the animals in this study will be monitored for the remainder of their lives, enabling future work to characterize the later stages of OA development.

As with other studies of OA (9; 13-15; 28), much variability exists after controlling for age, body mass, sex, and activity level. We can offer only speculation at this time regarding what might be associated with this unexplained variability. It is possible that even though the monkeys were single-caged throughout adulthood, some may have been more active than others within the cage, but falls and other types of high-energy events are not likely to occur in the cage. Because the animals were adult when enrolled in the study, the opportunity for different environmental conditions to be present when the monkeys were juveniles and young adults exists. All of these monkeys are captive-bred, however, so these early differences would have been subtle. Finally, genetic predisposition remains a likely possibility that we hope to address in future work.

The limitations of this study are several. The animals in this study spent much or all of their lives housed in single cages, where their activities are limited compared to those of free-ranging animals, leading to possible adaptations to disuse. DeRousseau (9), however, argued against the likelihood of “cage paralysis” or aberrant patterns of degeneration based on work with both caged and free-ranging animals. Despite having access to the same equipment and technicians for the length of the study, differences in animal positioning were unavoidable. Some animals developed kyphosis and scoliosis, which also limited their positioning for radiography. Lack of clarity of some radiographs rendered readings of some regions of the spine impossible. These concerns regarding consistency are also issues that arise in human studies and are often more significant due to machinery and technician variation between clinics and radiographic facilities. Lastly, while several OA studies have been carried out using a primate model, the lack of consistency in the reporting of results as well as the variation in each study's anatomical foci (knee, hip, hand, spine) provide little information for direct comparison.

The macaque provides a valuable and useful, though not flawless, biomedical model of spinal OA in humans. The condition appears the same and mechanically acts the same: manifesting bony changes and remodeling such as traction and claw osteophytes, disc tissue degeneration, and change in mechanical properties in an almost identical manner in both groups. Whether or not these gross similarities reflect cellular-level processes is currently unknown, but the likelihood of shared physiological processes, based on other biologically shared traits and responses, is high. Although nonhuman primate models present ethical, practical, and monetary concerns, they also present unique advantages. The more rapid rate of aging in macaques, almost triple that of humans, is still appreciably slower than smaller lab animals and, therefore, nonhuman primate models have the potential to more accurately assess the effectiveness of new medical interventions and treatments designed for humans. The macaque's lifespan and aging rate also provide opportunity for longitudinal study not feasible in humans, as previously discussed. Lastly, the extensive macaque records and pedigrees available at the National Primate Research Centers provide opportunities for controlled studies examining the roles of genetics, activities, biomarkers and environment which are also not possible in human studies.

In summary, our results show that longitudinal data of naturally-occurring spinal OA in macaques support previous cross-sectional findings in monkeys, confirming that age and mass are significant predictors of DSN and OST. These results support the human results as well, validating the macaque monkey as a valuable in vivo model of spinal OA. Additionally, this model offers the unique opportunity to study rates of change and the progression of OA across the lifetime of an individual and biological and morphological studies are ongoing in an effort to better characterize OA. The clinical relevance of this study and the primate model lies in our ability to control for known risk factors, such as age and mass, which according to cross-sectional and longitudinal study, can predict only ~50% of variability in spinal OA. This indicates that there are other factors – genetic, hormonal profiles, etc. – that we should be looking for with future research. Ultimately, these new directions should present opportunities for developing preventative and therapeutic interventions for humans.