Baseline Plasma GDF15 and Recovery of Physical Function Following Total Knee Replacement in The Study of Physical Resilience and Aging

Abstract

Growth-differentiation factor 15 (GDF15), a cytokine with the ability to regulate metabolic and inflammatory activity, is negatively associated with physical and cognitive function, and increases in circulation with age. Mechanistically, the expression of GDF15 is stimulated by mitochondrial stress across multiple tissues. We hypothesized elevations in basal circulating GDF15 were negatively associated with physical function following surgery in older adults. This was assessed in 112 Study of Physical Resilience and Aging (SPRING) participants (age 69.6 ± 6.9 years, 66% women) undergoing total knee replacement (RESTORE). The associations between pre-operative plasma GDF15 levels and longitudinal post-operative physical resilience measures including the Short Physical Performance Battery (SPPB) and Pittsburgh Fatigability Scale (PFS) were evaluated. At baseline, higher circulating GDF15 levels were associated with older age, higher BMI, diabetes, and physical frailty (P<0.05). Circulating GDF15 levels were not associated with SPPB or PFS scores prior to knee replacement (P>0.05). Higher baseline circulating GDF15 levels were negatively associated with the recovery of SPPB scores six months following knee replacement (P<0.05). However, there was no significant association between baseline circulating GDF15 levels and the recovery of PFS scores within the same timeframe (P>0.05). There were no significant associations between baseline circulating GDF15 and recovery of SPPB or PFS scores at 1-month or 12-months follow up (P>0.05). These findings suggest that pre-operative circulating GDF15 levels may provide some insight into the capacity to recover physical function following total knee replacement surgery. Further investigation is necessary to elucidate relationships between GDF15 and the biology of physical resilience.

INTRODUCTION

Growth-differentiation factor 15 (GDF15) is an age-associated regulatory cytokine which influences metabolic and inflammatory pathways across multiple tissues, including skeletal muscle [1–3]. GDF15 is classified as a member of the TGF-b superfamily and is typically viewed as an indicator of mitochondrial stress [4–6]. Elevations in circulating GDF15 have been associated with numerous comorbidities such as metabolic disease, cancer, cachexia, sepsis, cognitive impairment, mitochondrial dysfunction, and all-cause mortality [7–10]. In fact, the Framingham Heart Study (n=3428) found that GDF15 was associated with incident heart failure and major cardiovascular events in addition to all-cause mortality [11]. Longitudinal examination of 660 men and women aged 60-94 indicated that higher GDF15 levels were associated with a 45% higher risk of developing mobility disability [12]. Recent evidence indicates that the health and integrity of skeletal muscle tissue may drive circulating GDF15 levels [13]. Collectively, there is mounting evidence to suggest a relationship between the maintenance of physical function and circulating GDF15 levels. Although the associations between circulating GDF15 levels and numerous disease-related outcomes have been described, information is lacking regarding its relationship with physical resilience [14].

Physical resilience as defined by the Study of Physical Resilience and Aging (SPRING) is the capacity of a person to withstand physical stress and quickly recover or improve upon a baseline functional level [15]. The primary aim of SPRING is to identify the specific physiological and biological components of physical resilience in response to clinical stressors in older adults undergoing either knee replacement, allogeneic bone marrow transplant, or renal failure [15]. The biological underpinnings of physical resilience are inherently complex, with multiple integrated response mechanisms making the existence of a single causative molecule unlikely. Still, GDF15 has emerged as a potential indicator of the physiology underlying physical resilience because of its relationship with metabolic and inflammatory activity [2 16 17]. Recent proteomic analysis of 1,752 older adults revealed that age-associated elevations in circulating GDF15 levels are associated with prevalent and incident frailty [18]. Interestingly, patients with diabetes, primary mitochondrial myopathies, and lower limb mobility impairment all exhibit similar elevations in circulating GDF15 [13 18–22]. GDF15 is believed to be produced in direct response to mitochondrial perturbations [23 24], raising possibility that its elevation indicates deficits of cellular energy production. These findings indicate GDF15 could have prognostic potential in characterizing the hallmark decline of physical function later in life.

Due to the biological relationship between GDF15, skeletal muscle health and function, and the clinical importance of physical resilience following medical procedures in older adults, this investigation assessed pre-operative circulating GDF15 levels as a potential indicator of resilience following total knee replacement surgery. The primary aim was to evaluate the potential associations between pre-surgical circulating GDF15 levels and physical resilience following total knee replacement. For the purposes of this investigation, physical resilience is operationally defined as changes in physical function measured with the Short Physical Performance Battery (SPPB), an established series of physical tests evaluating lower extremity function and neuromuscular coordination [25], the Pittsburgh Fatigability Scale, a self-reported measure related to energy availability [26], and an omnibus measure of resilience in knee replacement (defined below).

METHODS

Participants

The current investigation studied participants in the RESilience in TOtal knee REplacement (RESTORE) cohort of the ongoing Study of Physical Resilience and Aging [15]. RESTORE was implemented in a pilot phase followed by a confirmatory phase. In both phases, participants were considered eligible for RESTORE if they had the ability to walk without human assistance, were English-speaking, and were willing and able to sign a written informed consent document. Adults 60 years of age and older who were scheduled to undergo elective total knee replacement surgery for degenerative joint disease at the Johns Hopkins Bayview Medical Center or at the University of Maryland Medical System in Baltimore, Maryland were screened for inclusion. Exclusion criteria included severe visual or hearing impairments, intubation pre-surgery, severe medical comorbidities such as acute congestive heart failure or cardiac ischemia, current glucocorticoid use, and infection requiring IV antibiotics. Participants without a valid GDF15 measurement were excluded from the current analysis. All study protocols were approved by Johns Hopkins Institutional Review Boards and conducted according to the Declaration of Helsinki.

Study Design

RESTORE participants were asked to complete five study visits and the standard of care visits. Participant evaluations included baseline visits with follow-up assessments occurring at 1 month, 6 months, and 12 months post-surgery (Figure 1). After the determination of surgical intervention, potential participants were called by a trained study coordinator to confirm the patient’s eligibility and interest prior to scheduling the consent session and baseline visits. For this arm of the investigation, participants were scheduled to execute the informed consent process and then collect baseline data. Patients were extensively characterized regarding sociodemographics, psychosocial factors, self-reported health behaviors, accelerometry, medical history, medications, cognitive function, frailty, fatigability, grip strength, and Short Physical Performance Battery (SPPB). Self-reported health and function were assessed, blood was collected, and a battery of stimulus-response tests assessing fitness of stress-response physiology was administered. Elements of each patient’s surgical procedure were documented in detail, and measures of physical function were repeated at one-month, six-months, and twelve-months following the procedure to assess the potential relationship between baseline (pre-operative) plasma GDF15 levels and physical resilience.

Figure 1.

Overview of RESTORE Phase II study design. Participants were initially screened for eligibility for total knee replacement and completed a series of baseline measures including health history, the Short Physical Performance Battery, and several questionnaires including the Pittsburgh Fatigability Scale. These measures were repeated longitudinally for up to one year after total knee replacement surgery. ICTR1, Institute for Clinical and Translational Research Visit 1. ICTR2, Institute for Clinical and Translational Research Visit 2.

Plasma GDF15 concentration

Peripheral blood was collected into heparinized tubes and centrifuged at 300g for 10 minutes before plasma was aliquoted and stored at −80° C until GDF15 quantification. All plasma samples used for determination of GDF15 were removed from −80° C and thawed at room temperature. Plasma samples were diluted 4-fold in calibrator diluent in accordance with manufacturer instructions (catalog no. DGD150, R&D Systems, Minneapolis, MN) and analyzed in duplicate on a standard plate reader. Sample concentrations were determined with a 4PLC curve based on GDF15 standards diluted in assay buffer.

Potentially Confounding Variables

To refine our study of association between GDF15 concentrations and resilience phenotypes, we sought to control for potentially confounding variables measured at baseline which might impact both and hence account for any association or, conversely, mask an association. The size of the RESTORE sample limited the number of such variables that could be considered while avoiding overfitting. We prioritized a few as most important. Participant age in years and sex were self-reported and included in the fully adjusted models. BMI was calculated from measured height and weight. The Charlson comorbidity index was calculated based on the sum of self-report of diseases, weighted based on disease severity [27]. Myocardial, vascular, pulmonary, neurologic, endocrine, renal, liver, gastrointestinal, cancer, rheumatologic, and coagulopathic diseases were included in the Charlson calculation. For analyses, Charlson scores were categorized as 0, 1, or 2 and greater. Prevalent diabetes was self-reported.

Medication use was initially assessed by self-report, and confirmed by study staff after patients were asked to bring all medications to their baseline screening. In 1997, the gene encoding GDF15 was found to be induced by non-steroidal anti-inflammatory drugs (NSAIDs) and subsequently referred to as NSAID-activated gene-1 [28 29]. Mechanistically, NSAIDs inhibit the cyclooxygenase (COX) enzyme, an important regulator of tissue-specific and systemic inflammation with a strong influence on aging skeletal muscle [30–33]. Recent evidence suggests that metformin increases circulating GDF15 levels by stimulating the cellular integrated stress response system [34]. Therefore, COX-inhibiting drugs and metformin were of particular interest in characterizing the study sample and the development of sensitivity analyses (described below).

Physical frailty was characterized using the Fried frailty phenotype [35]. The phenotype assesses five criteria (unintentional weight loss, muscle weakness as assessed by hand-held dynamometer, fatigue as determined by self report of energy, slow gait as determined by a four-meter walk at usual pace, and low self-reported physical activity) whose values are then categorized as 0 criteria = Robust, 1-2 criteria = Prefrail, and 3+ criteria= Frail. The Women’s Health and Aging Study versions of the criteria were used [36]. Physical frailty was considered but ultimately not included as a confounder because we hypothesized it potentially to be in the causal pathway between GDF15 and resilience phenotypes.

Outcome Variables Assessing Resilience

These were chosen for their relevance to energy metabolism and inflammation, or to reflect overall physical resilience. The Short Physical Performance Battery (SPPB) is a lower extremity performance-based assessment comprising 3 tasks: 1) 5 repeated chair stands (arms folded across chest); 2) 3 progressively harder standing balance poses (feet positioned side-by-side, semi-tandem, tandem) held for up to 10 seconds each; and 3) a 4-meter usual paced walk (meters/second [m/s]). This metric was chosen as a measure of physical resilience not only for the insight provided into physical function [25] but also the established relationship between SPPB scores and all-cause mortality [37]. Participants were scored on a scale from 0-4 in each component of the SPPB to provide a summary score of their physical performance measures using the standard SPPB calculation [25 37]. Participants also completed a self-administered 10-item Pittsburgh Fatigability Scale (PFS) questionnaire assessing perceived physical and mental fatigability related to fixed-intensity and duration activities [26], each scored on a scale of 0 to 50 (reversed, with higher scores indicating less fatigue). Only the physical scores were used in this analysis.

As described in recent work by Xue and colleagues, the omnibus RESTORE Resilience Phenotype was created using pre-surgery baseline data and follow-ups at 1, 6, and 12 months post-surgery, through a hybrid latent variable model [38]. This model included a first-stage latent profile analysis (LPA) and a second-stage latent class analysis (LCA). The LPA used repeated measurements of four measures (SPPB, PFS, KOOS-QOL, and SF-36 PCS) as continuous indicators of categorical latent variables, capturing different temporal trajectories (latent classes). The Knee Injury and Osteoarthritis Outcome Score Quality of Life subscale (KOOS-QOL) measures the impact of knee injury or osteoarthritis on quality of life, transformed to a 0–100 scale with higher scores indicating better knee health [39]. The physical component summary (PCS) of the SF-36 questionnaire provides an overall view of perceived physical functioning on a weighted sum of eight subscale scores and standardized to have a mean of 50 and a standard deviation of 10, with higher scores indicating better physical health [40]. A categorical indicator for each measure was then created based on the LPA results. The second-stage LCA used these indicators to identify an overarching resilience phenotype, aggregating the trajectory patterns into a binary variable of resilience vs. non-resilience [38].

Statistical Analysis

Descriptive analyses were performed to create summary statistics, visualize distributions, identify associations of GDF15 concentrations with the prioritized potential confounders and other key concomitant variables, and identify data anomalies.

Primary analyses evaluated associations of GDF15 concentration with the selected resilience phenotypes in confirmatory stage RESTORE participants, using linear regression for SPPB and PFS outcomes and logistic regression for the omnibus resilience phenotype. For each analytic outcome, a sequence of models was fit including as covariates GDF15 alone, age and BMI alone, the trio of these, all the confounders, and GDF15 plus all the confounders to create a fully adjusted model. Sex was also included in the fully adjusted model. For each phenotypic change analysis, the baseline phenotype value also was adjusted in each model. For SPPB and PFS, separate analytic sequences were fit for outcomes of baseline values and baseline-to-follow up change over one month, six months, and 12 months. For each phenotype and baseline/change time point, the analysis was restricted to participants with complete data in the full model with GDF15 and all confounders. Findings were visualized using partial residual plots in which the adjusted association is depicted overlaid with the full model residuals. In all cases, excellent model fit was indicated.

Extensive sensitivity analyses were performed including (i) utilizing data from both RESTORE phases; (ii) utilizing all available data in each model fit; (iii) winsorizing large outliers; (iv) restricting to the range of GDF15 overlap between diabetics and non-diabetics; (v) adding metformin use and NSAID use (separately) as potential confounders in the primary analyses. Winsorizing replaces large outlier values with a value proximal to their nearest values in the remaining bulk of the data. For these analyses, one GDF15 value exceeding 4000 pg/mL was recoded to 3000 pg/mL; one BMI value exceeding 50 kg/m² was recoded to 44 kg/m²; and two six-month decreases in SPPB exceeding 3 points were recoded to 3 points. In models adding metformin, only one individual without diabetes reported Metformin use: the two variables were represented by a three-category variable coded as 0=neither diabetes nor Metformin use (n=76), 1=diabetes without Metformin use (n=12), and 2=Metformin use (n=21).

RESULTS

Participants

This investigation evaluated 112 men and women (69.6 ± 6.9 years) who received total knee replacement surgery (Table 1). This cohort was notably diverse in sociodemographic characteristics including race (41% non-white), sex (66% women) and age (range: 59 – 87 years). This cohort was also physiologically diverse with a moderate prevalence of diabetes (30%), pre-frailty and frailty (77%), and range of BMI (20.2 – 50.7 kg/m²). A summary of the most prevalent diseases is provided in Table S1. As expected, higher baseline circulating GDF15 levels were associated with older age and higher BMI (P<0.05). Unadjusted circulating GDF15 levels were also strongly associated with prevalent diabetes and physical frailty (Figure 2). In our study sample, 22 participants were consuming metformin. Aspirin (81mg) was the most prevalent medication (n=37) in this study sample. Other drugs sharing a similar mechanism of action (COX-inhibitors, Table 1) include acetaminophen, celecoxib, diclofenac, ibuprofen, naproxen, meloxicam, and mesalamine.

Table 1.

Participant Characteristics

Variable	n=112
Age, years	69.6 ± 6.9
Height, in	65.6 ± 3.9
Weight, lbs	195.5 ± 35.7
BMI, kg/m2	32.1 ± 5.4
Sex, n (%)
Men	38 (33.9%)
Women	74 (66.1%)
Race, n (%)
White	66 (58.9%)
Black	42 (37.5%)
Other	4 (3.6%)
Education, years	14.5 ± 2.8
Charlson Comorbidity Index, n (%)
None (0)	63 (56.3%)
Mild (1-2)	30 (26.8%)
Moderate/Severe (>2)	19 (17.0%)
Diabetes, n (%)	33 (30%)
Metformin, n (%)	22 (19.6%)
COX Inhibitors, n (%)	63 (56%)
Physical Frailty, n (%)
Frail	21 (18.9%)
Prefrail	65 (58.0%)
Robust	26 (23.2%)
Gait speed, m/s	0.77 ± 2.24

Note: Data are presented as mean ± SD, or n (% of study sample).

Figure 2.

Baseline GDF15 and associated health condition prevalence. A) GDF15 was significantly (P<0.05) elevated in participants with diabetes compared to those without (0=non-diabetic; 1=diabetic). B) GDF15 was significantly elevated in physically frail participants compared to prefrail or robust participants (P<0.05).

GDF 15 and Physical Resilience Measures

On average, participants improved their SPPB scores following total knee replacement and reported higher PFS scores, indicating lower levels of fatigue on the reversed scale (Figure 3). The concentration of circulating GDF15 was not associated with SPPB scores prior to, or their changes 1 month after or 12 months after, total knee replacement in any model fit (P>0.05). Circulating GDF15 concentrations were also not associated with change in fatigability scores at any follow-up time point (P>0.05). Higher concentrations of GDF15 at baseline were associated with poorer recovery of objectively measured physical function (measured as change in total SPPB score from baseline) at six-months (P<0.05) after full confounder adjustment (Figure 4). There was also no association between basal circulating GDF15 levels and the RESTORE Resilience Phenotype (P>0.05).

Figure 3.

Trajectory of objective (SPPB) and subjective (PFS) measures of physical function following total knee replacement. On average, physical function improved following total knee replacement with prevalent heterogeneity.

Figure 4.

Baseline GDF15 and post-operative recovery of physical function. Prior to total knee replacement, higher circulating GDF15 levels were associated with poorer recovery of SPPB scores after 6-months of follow-up. However, this relationship did not persist after 12-months of follow-up. Physical function data reported as change from baseline.

DISCUSSION

The primary findings from this investigation indicate that elevated circulating GDF15 levels prior to total knee replacement are associated with poorer recovery of physical function measured with the Short Physical Performance Battery six-months after surgery. Specifically, this model predicts rebound of SPPB score that is worse by nearly one point for those at the upper versus the lower quartile of GDF15 values. Although this relationship did not persist through 12 months of follow-up, the current investigation provides novel insight into the potential utility of circulating GDF15 as an indicator of physical resilience. The lack of association between baseline circulating GDF15 levels and resilience measures one month after surgery is likely explained by the timing of the follow-up measures. The distribution of SPPB scores at baseline and 1 month are similar, suggesting that 1 month remains early in the recovery period, perhaps too early to detect the impact of the biological factors determining resilience. The distribution shifts higher at 6 months compared to 1 month. While the overall distributions at 6 and 12 months are similar, there is considerable individual variability, with notable improvements in some and declines in others over those six months. Collectively, the data suggest that 6 months provides the necessary time for recovery while remaining proximal enough to the surgery to detect the impact of the biological factors determining resilience (Figure 3), and that subsequent changes may reflect extra-biological factors. There were no significant associations between baseline GDF15 levels and either the RESTORE resilience phenotype [38] or the Pittsburgh Fatigability Scale. This finding was unexpected and suggests that elevations in circulating GDF15 may indicate the presence of biological stress which was not subjectively identifiable by the participants.

The biology of GDF15 signaling is complex and not yet fully understood. GDF15 is a known product of the cellular integrated stress response system and is believed to be related to mitochondrial perturbations [1 6 8 23 24]. GDF15 has also been shown to induce its own transcription by activating ERK1/2 after an extended period in cancer cells [41]. The auto-induction mechanism is particularly relevant to the age-associated elevations in circulating GDF15 reported by the current investigation and others [1 7]. It is possible that a biological desensitization to circulating GDF15 levels may occur in older men and women, and sex-specific differences have been reported to this effect [42 43]. Quantification of the GDF15 receptor, GFRAL, remains necessary to thoroughly test this hypothesis. GFRAL is a membrane-bound receptor known to be expressed in the brain with limited evidence in mouse models indicating its expression in peripheral tissues including skeletal muscle [4 44 45]. The relative lack of known GDF15 receptors suggest that autocrine/paracrine signaling is likely the primary mechanism of downstream GDF15 activity. Mechanistic investigations remain necessary to fully understand the biological regulation of GDF15 production and downstream pathway activation, especially in older adults. This information may prove useful in understanding the potential role of GDF15 in determining the physical resilience of older adults faced with a clinical physical stressor such as knee replacement surgery.

The associations between diabetes, physical frailty, and circulating GDF15 levels in this cohort are noteworthy. The current investigation supports previous reports of elevated circulating GDF15 in diabetic and physically frail older adults [18 21 22 46]. Transient elevations in circulating GDF15 have been observed after strenuous aerobic exercise and is believed to enhance skeletal muscle glucose uptake [20 47 48]. Conversely, chronically elevated GDF15 levels are associated with numerous disease states [7–10]. The disparate effects of transient and chronic elevations in GDF15 mirrors observations made with other cytokines. Specifically, the acute regulatory influence of GDF15 on metabolic and inflammatory pathways, and chronic elevations associating with age and numerous disease states, parallels previous observations in IL-6, TNFa, TNFR1, and CRP [17 49–52]. Therefore, it seems plausible that interventions mitigating the age-associated increase of GDF15 may be beneficial to older adults.

Recent investigations have provided preliminary insight into the relationship between circulating GDF15 and physical activity levels [12 20 53]. In a cross-sectional study of 50-year-old Swedish men and women circulating GDF15 levels were inversely associated with cardiorespiratory fitness (VO₂max), indicating a relationship between this mitochondrial stress cytokine and maximal oxygen consumption (i.e., exercise capacity) in healthy adults [54]. A separate longitudinal analysis of 1,083 healthy adults (63.8% women) aged 70 and over demonstrated that participants with higher physical activity levels at baseline had lower body weight and GDF15 levels after one year [55]. It is likely that increasing physical activity levels could reduce GDF15 expression due to the reduction in age-associated mitochondrial degeneration [56–61]. These findings are particularly relevant to the current investigation of physical resilience, as physical activity levels were limited by pain in this cohort of knee replacement patients, potentially contributing to elevations in baseline circulating GDF15.

This study has limitations. First, the relatively small sample size limits generalizability of these findings to the broader population of older adults undergoing total knee replacement surgeries. The sample size also limited the power to perform sex-specific analyses, a potentially important factor related to GDF15 biology in older adults [42 43]. There are counterbalancing strengths. Although the study sample is relatively small it is extremely well characterized. Extensive sensitivity analyses were performed, and appropriate confounding variables were considered in the models. This preliminary analysis of the associations between GDF15 and the recovery of physical function following a surgical procedure is, to our knowledge, the first of its kind and can serve as a template for future investigations to expand upon. Longitudinal analyses in SPRING investigating the potential relationship between objectively measured increases in physical activity following total knee replacement and changes in circulating GDF15 are ongoing.

This investigation provides novel preliminary insights toward the association between GDF15 and physical resilience following total knee replacement. These and similar findings could influence clinical practice by being integrated into pre-operative screening or risk stratification protocols for older adults undergoing surgery. This line of research has the potential to improve our ability to identify non-resilient patients prior to their surgery, ultimately reducing patient burden and providing more precise, individualized treatment plans. Furthermore, preserving the plasticity of metabolic and inflammatory pathways is critically important in older adults and GDF15 has recently emerged as a potential contributor to intrinsic capacity [17]. Longitudinal investigations focused on the dynamic relationships between metabolic and inflammatory activity, potentially mediated by GDF15, remain necessary. The authors suggest that GDF15 be considered in future mechanistic, translational, and epidemiologic studies of physical resilience in older adults.