Serum PTHrP Predicts Weight Loss in Cancer Patients Independent of Hypercalcemia, Inflammation, and Tumor Burden

Abstract

Context:

Recent animal studies showed that tumor-derived PTHrP induced cancer cachexia by fat browning with increased energy expenditure; however, clinical evidence from human data is insufficient.

Objective:

We investigated whether serum PTHrP levels independently predicts weight loss (WL) in cancer patients.

Design, Setting, and Patients:

From a longitudinal observational cohort, body mass index (BMI) of patients with measured serum PTHrP levels (n = 624) was assessed (median follow-up of 327 d).

Main Outcome Measures:

Cox hazard models were used to examine the predictive value of PTHrP for WL defined by consensus definition (WL [consensus], percentage WL < −5% or percentage WL < −2% plus BMI < 20 kg/m²) and by BMI-adjusted grades (WL [BMI adjusted]).

Results:

The overall risk of WL (consensus) was 34.4%. Compared with PTHrP-negative subjects, patients with higher PTHrP levels (PTHrP ≥ median 5.7 pmol/L) had more WL (percentage WL, −6.9% vs −1.1%, P = .010) at follow-up. A higher PTHrP level was associated with an increased loss of body weight (β = −2.73), muscle (β = −1.85), and fat (β = −2.52) after controlling for age, sex, and BMI. Kaplan-Meier analysis demonstrated that subjects with higher PTHrP had increased WL risk compared with lower PTHrP or PTHrP-negative groups (52.0% vs 38.9% vs 29.7%, P < .001). Serum PTHrP was independently associated with an increased WL risk (hazard ratio [HR]1.23, P = .005) adjusted for potent predictors including serum levels of calcium, C-reactive protein, albumin, cancer stage, and performance status of patients. Consistent results were observed when BMI-adjusted WL was applied.

Conclusions:

Serum PTHrP levels predicted cancer-associated WL independent of the presence of hypercalcemia, inflammation, tumor burden, and other comorbidities.

PTHrP is currently considered as a prevalent cause of malignancy-associated hypercalcemia (1, 2). PTHrP acts as an endocrine hormone to stimulate calcium resorption from bone and renal reabsorption, which leads to hypercalcemia. Furthermore, it has been postulated to play a certain role in cancer cachexia through production of orexigenic peptides (3).

Cancer cachexia is a multifactorial wasting disorder characterized by involuntary weight loss (WL), chronic inflammation, and abnormal metabolism (4–6). It affects more than 50% of patients with various cancers and leads to reduced food intake, low performance status, frailty, intolerance to cancer therapy, and decreased survival (7). Progressive WL is often refractory to intensive nutritional support alone (8). Few therapies have been proven to be effective in treating cancer cachexia, although other approaches including anabolic agents, antiinflammatory drugs, and anticytokines are under investigation (5).

Underlying mechanisms of cancer cachexia have not been clearly understood. Systemic inflammation, reduced food intake, and altered energy metabolism have been suggested as central to the pathophysiology of cancer cachexia (5, 8). Metabolic alteration toward negative nitrogen balance and increased apoptosis induced by inflammatory cytokines contributes to muscle wasting (5). One key feature of cancer cachexia is thermogenesis via activation of brown fat or uncoupling protein 1 (UCP1)-expressing multilocular cells (beige cells) in white fat depots, termed browning, which leads to negative energy balance (9). Brown fat is a metabolically highly active tissue, which can expend its energy as heat via thermogenesis in mammals (10). A recent study proposed that tumor-derived PTHrP, a well-known paraneoplastic marker of humoral hypercalcemia in malignancy, might play a crucial role as a mediator of adipose tissue browning and cancer cachexia (11). In this study, PTHrP treatment increased the UCP1 protein levels and uncoupled respiration, whereas the anti-PTHrP antibody blocked wasting of fat and muscle tissue in an animal model.

However, there are no clinical longitudinal data on PTHrP levels and WL, a hallmark of cancer cachexia, in cancer patients. Therefore, we examined the association between PTHrP levels and the risk of WL in patients with cancer.

Materials and Methods

Study population and data collection

From a SEverance CanceR cachExia neTwork cohort, we analyzed adult patients (age > 18 y) who had measured serum PTHrP levels from November 2005 to June 2014. A total of 624 patients who were referred to an outpatient or inpatient medical oncology department at Severance Hospital, Yonsei University College of Medicine (a tertiary level, university affiliated cancer treatment center) were included in this study. Baseline data included age, sex, body mass index (BMI), types and sites of cancer, stage of cancer, performance status (PS; recorded as Eastern Cooperative Oncology Group [ECOG] PS), time since cancer diagnosis, and comorbidities such as congestive heart failure (CHF), chronic obstructive pulmonary disease (COPD), liver cirrhosis, and diabetes mellitus. Cancer stage was defined based on the American Joint Committee on Cancer stage grouping (stages I, II, III, and IV). This study was approved by the Institutional Review Board of Severance Hospital, Seoul, Korea (institutional review board number 4-2015-0246).

Definition of WL and changes in body components

BMI or body weight was measured in patients at the time of PTHrP measurement. Follow-up body weights were routinely measured at the outpatient or inpatient clinic over a median follow-up period of 327 days (interquartile range 188–423 d). Percentage WL (%WL) was calculated as follows: ([follow-up weight − baseline weight]/baseline weight) × 100. BMI was reported as weight (kilograms)/height (meters)². Clinically significant WL was classified using two definitions. WL as binary outcome was defined as %WL less than −5% or baseline BMI less than 20 kg/m² plus %WL less than −2% at follow-up, modified from the international consensus definition (WL [consensus]) (6). BMI-adjusted WL (WL [BMI adjusted]) was graded from 0 to 4 based on the risk of reduced survival by 5 × 5 matrix analysis of baseline BMI categories (<20.0, 20.0–21.9, 22.0–24.9, 25.0–27.9, and ≥ 28.0 kg/m²) and %WL categories (±2.4%, −2.5% to −5.9%, −6.0% to −10.9%, −11.0% to −14.9%, and ≤ −15.0%) according to a recent study (Supplemental Table 1) (12). WL (BMI adjusted) as binary outcome was defined as a composite of grades 2–4. Lean body mass (LBM), body fat mass (BFM; presented as percentage body fat), and resting energy expenditure (REE) were estimated using the following formulas, respectively: 1) LBM (kilograms) = (9270 × weight [kilograms]/[6680 + 216 × BMI]) in males and (9270 × weight [kilograms]/[8780 + 244 × BMI]) in females (13); 2) BFM (percentage) = (1.20 × BMI + 0.23 × age − 16.2) in males and (1.20 × BMI + 0.23 × age − 5.4) in females (14); and 3) REE (kilocalories per day) = (10 × weight [kilograms] + 6.25 × height [centimeters] − 5 × age [years] + 5) in males and (10 × weight [kilograms] + 6.25 × height [centimeters] − 5 × age [years] − 161) in females (15).

Measurements of PTHrP and other laboratory parameters

Serum PTHrP was measured using an immunoradiometric assay (Allegro immunoradiometric assay; Nichols Institute) (16, 17). The detection limit of this assay in our laboratory was 0.5 pmol/L, as determined by replicate analysis of the zero-standard. For PTHrP measurements, precooled evacuated collection tubes (supplied by Nichols Institute) containing aprotinin (2500 kallikrein units/tube), leupeptin (25 pg/tube), pepstatin (25 pg/tube), and EDTA (4.5 mg/tube) were used. Serum concentrations of calcium, phosphate, hemoglobin, and albumin were measured by standard methods. Corrected calcium was calculated as follows: measured serum calcium + 0.8 × (4.0 − serum albumin [grams per deciliter] if serum albumin < 4.0). Serum C-reactive protein (CRP) and creatinine levels were determined with a Hitachi 7600–110 automated chemistry analyzer (Hitachi Co). The estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease (CKD) Epidemiology Collaboration equation based on serum creatinine (18). CKD was defined as eGFR less than 60 mL/min per 1.73 m².

Statistical analysis

Differences between groups were analyzed using an independent t test, an Wilcoxon rank sum test, an ANOVA with Bonferroni correction, or Kruskal-Wallis (continuous variables) and χ² (categorical variables) tests. Correlation between continuous variables was assessed with Pearson correlation coefficient. A paired t test was used to examine changes in weight, BMI, LBM, and BFM from baseline to follow-up. PTHrP was entered into the regression models in forms of natural log-transformed PTHrP (log-PTHrP) to achieve normal distribution. β-Coefficients by linear regression models were obtained to assess correlation between PTHrP, %WL, and percentage change (%change) in other anthropometric variables. To examine the effects of PTHrP and covariates on WL, the Kaplan-Meier method was used and WL free curves were compared using log-rank tests. Multivariate Cox proportional hazard models for WL-free survival were established, and concordance (c) statistics with 95% confidence intervals (CIs) were estimated for PTHrP and multivariate models to assess discrimination in predicting WL (19). A two-sided value of P < .05 was considered significant. All statistical analyses were performed using STATA 12.1 (Stata Corp).

Results

Baseline characteristics of patients by serum PTHrP levels

Baseline data for 624 patients with a diagnosis of cancer (mean age 60.9 ± 13.7 y, 60.9% male) are presented (Table 1). Cancers of the respiratory tract, liver, pancreatobiliary, gastroesophageal, and genitourinary systems were prevalent in the study subjects (Supplemental Table 2). Those with PTHrP was positive in 367 patients (58.8%), ranging from 1.1 to 107.0 pmol/L with a median value of 5.7 pmol/L (interquartile range 2.8–10.1 pmol/L). When stratified by serum PTHrP levels, those with PTHrP at the median level or higher were associated with advanced cancer stage, poor PS, higher corrected calcium, elevated CRP, and lower serum albumin levels compared with those with PTHrP that was undetectable or lower than the median level. Baseline weight, BMI, LBM, BFM, and REE did not differ significantly among PTHrP groups when stratified by sex. There was a significant but weak positive correlation between PTHrP and corrected calcium level (correlation coefficient r = 0.18, P < .001) and CRP (r = 0.22, P < .001).

Table 1.

Baseline Characteristics of Study Subjects by Serum PTHrP Levels

Variables	Study Subjects (n = 624, 100%)
	Total	PTHrP Not Detected (n = 257, 41.2%)	PTHrP < Median (n = 181, 29.0%)	PTHrP ≥ Median (n = 186, 29.8%)
Age, y	60.9 ± 13.7	61.2 ± 14.8	60.6 ± 13.6	60.8 ± 12.1
Male sex	380 (60.9)	121 (47.1)	123 (67.9)a	136 (73.1)a
Weight, kg	58.7 ± 11.0	58.9 ± 10.9	58.6 ± 11.0	58.5 ± 11.1
BMI, kg/m2	22.2 ± 3.7	22.7 ± 3.7	21.9 ± 3.2	21.7 ± 3.9a
LBM, kg	43.9 ± 8.8	42.5 ± 8.7	44.6 ± 8.9a	45.2 ± 8.4a
BFM, %	28.7 ± 7.8	30.8 ± 8.2	27.5 ± 6.7a	26.7 ± 7.3a
REE, kcal/d	1238 ± 210	1207 ± 214	1252 ± 213	1267 ± 198a
REE/LBM, kcal/kg·d	27.1 ± 1.5	27.1 ± 1.6	27.0 ± 1.5	27.1 ± 1.4
Cancer stage IV	399 (63.9)	115 (44.7)	129 (71.3)a	155 (83.3)a,b
ECOG PS 0–2	341 (54.7)	181 (70.4)	99 (54.7)	61 (32.8)
Time since diagnosis, d	209 [28–638]	179 [25–676]	241 [32–669]	216 [39–582]
Comorbidities
CHF	29 (4.7)	17 (6.6)	5 (2.7)	7 (3.7)
COPD	24 (3.9)	5 (1.9)	5 (2.7)	14 (7.5)a,b
Diabetes mellitus	163 (26.1)	73 (28.4)	55 (30.4)	35 (18.8)a,b
Liver cirrhosis	49 (7.8)	17 (6.6)	19 (10.5)	13 (6.9)
PTHrP, pmol/L	5.7 [2.8–10.1]	–	2.8 [1.6–4.2]	10.1 [6.8–14.4]b
Corrected calcium, mg/dL	10.4 ± 1.8	9.7 ± 1.4	10.6 ± 1.6a	11.3 ± 2.0a,b
Phosphate, mg/dL	3.3 ± 1.5	3.6 ± 1.2	3.3 ± 1.6	2.9 ± 1.6a
Albumin, g/dL	3.1 ± 0.7	3.4 ± 0.8	3.0 ± 0.7a	2.7 ± 0.5a,b
Hemoglobin, g/dL	10.3 ± 1.9	10.4 ± 2.0	10.2 ± 1.9	10.0 ± 1.7
Creatinine, mg/dL	1.3 ± 1.4	1.4 ± 1.7	1.2 ± 1.2	1.2 ± 1.0
eGFR, mL/min per 1.73 m2	77.8 ± 34.6	75.9 ± 36.7	78.7 ± 34.2	79.5 ± 32.1
CRP, mg/L	62.8 ± 74.7	42.9 ± 69.5	63.5 ± 73.8a	87.6 ± 74.8a,b

Data are presented as either mean ± SD or n (percentage). PTHrP values were presented as median [interquartile ranges]. –, not available.

^a

P < .05 compared with PTHrP negative group.

^b

P < .05 compared with PTHrP < median group.

Comparison of changes in body weight and other anthropometrics at baseline and follow-up in patients with cancer according to serum PTHrP levels

Figure 1 depicts changes in body weight, BMI, LBM, and BFM during a follow-up period in the study population. Patients lost weight overall but the %WL varied significantly across patients with different PTHrP levels. A trend toward an increased difference between baseline and follow-up weight was observed, ranging from a 1.1% reduction in the PTHrP-negative group to a 6.9% reduction in those with higher PTHrP levels (P = .010). Weight difference between baseline and follow-up within each PTHrP strata was significant only in PTHrP-detected groups. Similar patterns were shown with regard to BMI, LBM, and BFM in relationship with serum PTHrP levels.

Figure 1.

Box and range bar plots representing the mean and 95% CI for weight (A), BMI (B), LBM (C), and BFM (D) at baseline and follow-up (median follow-up duration 327 d; interquartile range from 188 to 423 d) grouped by serum PTHrP level in cancer patients.

Black and gray plots represent the mean value with 95% CI at the time of baseline and follow-up, respectively. Mean %change was calculated as ([follow-up value − baseline value]/baseline value × 100) for each anthropometric variable is presented above each plot. The median PTHrP value is 5.7 pmol/L. *, P values by paired t test, P < .05; **, P values by paired t test, P < .01.

Linear association of PTHrP with %WL and %change in BMI, LBM, and BFM adjusted for age, sex, and baseline BMI

To examine the linear relationship of serum PTHrP levels with %WL and %change in other anthropometric variables independent of age, sex, and baseline BMI, multivariate regression models were established using %change in weight, BMI, LBM, and BFM as dependent variables (Table 2). When PTHrP was entered into multivariate models, it showed a robust and independent negative association with %WL (−2.73% per 1 log unit PTHrP increase, P = .017) and also with %change in BMI, LBM, and BFM.

Table 2.

Linear Relationships Between Log-Transformed PTHrP and %change of Anthropometric Variables Adjusted for Age, Sex, and BMI Grades Using Multivariate Linear Regression Analysis

Dependent Variable	Log-PTHrPa
	β	P Value
Weight change, %	−2.73	.017
BMI change, %	−2.54	.030
LBM change, %	−1.85	.015
BFM change, %	−2.52	.044

Abbreviation: β, β-coefficient.

^a

Adjusted for age, sex, and BMI at baseline.

Independent predictive value of PTHrP for the risk of WL in cancer patients

We used two different definitions for clinically significant WL: WL (consensus) and WL (BMI adjusted). During a median follow-up duration of 327 days, 34.4% and 45.5% of patients were categorized as experiencing clinically significant WL based on WL (consensus) or WL (BMI adjusted), respectively. When grouped by serum PTHrP levels, there was a trend toward increasing WL (consensus) risk from negative PTHrP to higher PTHrP groups (29.6% vs 38.9% vs 52.0%, P for trend = .022, Figure 2A). More detailed grading by WL (BMI adjusted) yielded similar results, with the risk of moderate to severe WL (grades 2–4) increased in a stepwise fashion according to PTHrP levels (40.0% vs 50.0% vs 68.0%, P for trend = .008, Figure 2B). Kaplan-Meier analysis revealed that patients with higher PTHrP levels had the poorest WL-free survival compared with those in the lower PTHrP or PTHrP-negative groups, regardless of the definition of WL (log rank P < .001 in both WL definitions, Figure 3).

Figure 2.

Increased risk for clinically significant WL by two different definitions was observed in the higher PTHrP group compared with the PTHrP-negative or lower-PTHrP group.

Consensus definition: binary outcome as %WL less than −5% or %WL less than −2% plus BMI less than 20 kg/m² (A); or BMI-adjusted WL grades: grade 0 (mild) to 4 (severe) by 5 × 5 combination matrix of BMI and %WL categories (B); grades 0 and 1 are merged into a single category in this analysis. The median PTHrP value is 5.7 pmol/L.

Figure 3.

Cumulative Kaplan-Meier curves for WL-free probability defined by consensus definition and BMI-adjusted WL grading system are presented.

A, Cumulative Kaplan-Meier curves for WL-free probability was defined by the consensus definition: binary outcome as %WL less than −5% or %WL less than −2% plus BMI less than 20 kg/m². B, Cumulative Kaplan-Meier curves for WL-free probability was defined by BMI-adjusted WL grades: grades 2–4 are defined as significant WL for binary outcome. The median PTHrP value is 5.7 pmol/L. *, Log-rank P < .05; ** log-rank P < .01.

In a univariate Cox regression analysis (Table 3), higher PTHrP, CRP, and stage IV cancer were significant predictors of WL by both definitions, whereas a high serum albumin level and good PS (ECOG PS 0–2) were protective factors against WL. In subgroup analyses by cancer stage, PS, and cancer sites, hazard ratios were within the range of 1.21–1.48 across all subgroups without evidence of any effect modifications (Supplemental Table 3). A multivariate Cox analysis revealed the independent association of PTHrP with the risk of WL. One log-unit increase of PTHrP predicted 23% increased risk of WL (consensus) and 17% increased risk of WL (BMI adjusted), independent from baseline clinical characteristics including potent predictors in a fully adjusted model. In a multivariate model, cancer stage, PS, and albumin independently predicted WL alongside PTHrP. Meanwhile, the predictive value of CRP was attenuated (hazard ratio 1.04 for WL [consensus[, 95% CI 0.97–1.10, P = .265; hazard ratio 1.05 for WL [BMI adjusted[, 95% CI 0.99–1.10, P = .093) when cancer stage and PS were entered into the model, whereas the significance of PTHrP remained robust. PTHrP alone gave modest WL discrimination (c statistic 0.63 [0.56–0.69] for WL [consensus]; c statistic 0.62 [0.56–0.67] for WL [BMI adjusted]; and c statistic for the full model increased to 0.76 [0.71–0.82] for WL [consensus] and to 0.74 [0.69–0.79] for WL [BMI adjusted]).

Table 3.

Independent Predictive Value of PTHrP for WL Using Multivariate Cox Proportional Models

	Hazard Ratio for WL by Consensus Definition				Hazard Ratio for WL by BMI-Adjusted Gradesa
	Univariate		Multivariate		Univariate		Multivariate
	HR (95% CI)	P Value	HR (95% CI)	P Value	HR (95% CI)	P Value	HR (95% CI)	P Value
Log-PTHrP, log unit	1.38 (1.21–1.56)	<.001	1.23 (1.06–1.42)	.005	1.34 (1.20–1.50)	<.001	1.17 (1.04–1.33)	.012
Baseline age, y	1.00 (0.98–1.01)	.981	1.01 (0.99–1.03)	.522	1.00 (0.98–1.01)	.865	1.01 (0.98–1.02)	.493
Men (vs women)	1.56 (0.99–2.46)	.050	1.08 (0.65–1.77)	.776	1.43 (0.97–2.12)	.067	1.02 (0.65–1.56)	.946
Baseline BMI, kg/m2	0.96 (0.90–1.03)	.240	0.96 (0.87–1.04)	.291	0.92 (0.87–0.97)	.004	0.91 (0.85–0.98)	.008
Cancer stage IV	3.11 (1.91–5.03)	<.001	2.16 (1.21–3.86)	.009	2.79 (1.84–4.21)	<.001	2.04 (1.23–3.39)	.006
ECOG PS 0–2	0.25 (0.14–0.45)	<.001	0.39 (0.21–0.74)	.004	0.32 (0.19–0.55)	<.001	0.57 (0.31–1.04)	.067
CHF	1.25 (0.53–2.91)	.599	2.35 (0.81–6.81)	.117	1.34 (0.65–2.76)	.426	2.69 (1.14–6.39)	.024
COPD	0.48 (0.07–3.47)	.469	0.19 (0.02–1.69)	.137	0.72 (0.17–2.95)	.658	0.29 (0.06–1.45)	.134
Liver cirrhosis	1.53 (0.70–3.35)	.282	1.70 (0.71–4.06)	.233	1.43 (0.71–2.84)	.307	1.71 (0.81–3.64)	.161
Diabetes mellitus	1.18 (0.74–1.88)	.474	1.26 (0.73–2.17)	.399	0.96 (0.63–1.45)	.833	1.12 (0.50–1.37)	.646
CKD	0.96 (0.58–1.57)	.857	0.66 (0.73–2.17)	.162	0.97 (0.63–1.48)	.877	0.83 (0.50–1.37)	.461
Corrected Ca, mg/dL	1.09 (0.92–1.30)	.294	0.99 (0.80–1.22)	.897	1.12 (0.97–1.30)	.117	0.96 (0.80–1.15)	.664
Albumin, g/dL	0.49 (0.36–0.67)	<.001	0.60 (0.42–0.87)	.007	0.54 (0.41–0.71)	<.001	0.67 (0.48–0.94)	.019
CRP, 10 mg/L	1.10 (1.05–1.15)	<.001	1.04 (0.97–1.10)	.265	1.09 (1.05–1.14)	<.001	1.05 (0.99–1.10)	.093
C statistics	0.63 (0.56–0.69)b	<.001	0.76 (0.71–0.82)c	<.001	0.62 (0.56–0.67)b	<.001	0.74 (0.69–0.79)c	<.001

Abbreviations: corrected Ca, corrected calcium; log-PTHrP, natural log-transformed PTHrP. Values with statistical significance are printed in bold.

^a

WL by BMI-adjusted grade is defined as BMI-adjusted WL grades 2–4 as binary outcome for Cox model.

^b

c statistic for log-PTHrP.

^c

c statistic for multivariate Cox regression model.

Discussion

The principal findings of this study are as follows: 1) cancer patients with higher PTHrP levels were associated with increased WL, advanced cancer stage, poor PS, and elevated CRP and corrected calcium levels; 2) the serum PTHrP level was negatively correlated with %change of body weight, muscle, and fat components, regardless of age, sex, and baseline BMI; and 3) PTHrP independently predicted the risk of WL in patients with various cancers using a multivariate Cox hazard model adjusted for corrected calcium level, BMI, tumor burden, comorbidities, and inflammation status reflected by CRP level.

Previous studies using animal models suggested that PTHrP may cause paraneoplastic syndromes not only in terms of hypercalcemia of malignancy but also in cancer cachexia by progression of WL and anorexia (20, 21). When cancer tissue-transplanted mice were treated with an anti-PTHrP monoclonal antibody, the serum calcium level was restored and body weight and food intake improved. Treatment with bisphosphonate, calcitonin, or the receptor activator of nuclear factor-κB ligand antagonist osteoprotegerin was also effective in stabilizing serum calcium level, but the effect on progressive WL was limited compared with anti-PTHrP antibody treatment, indicating that PTHrP might play a role as a catabolic driver via mechanisms discrete from hypercalcemia. Human studies also revealed that elevated serum PTHrP level was associated with decreased LBM and higher REE in a cross-sectional analysis of patients with lung and colorectal cancers (11). Our study extended previous findings to a longitudinal setting, revealing a significant association of serum PTHrP level with increased loss of muscle, fat, and body weight and illustrating the independent predictive value of PTHrP for risk of WL in patients with various types of cancer.

Mechanisms of cancer cachexia are known to be multifactorial involving multiple organs (5). Systemic inflammation is regarded as a main driving force of metabolic alteration in patients with cancer (8). An elevated serum CRP level and decreased albumin as well-validated acute phase reactants and surrogate markers of inflammation in cancer patients with WL have been reported (22, 23). A low albumin level might also reflect poor nutritional status due to reduced food intake (23). Tumor burden represented by tumor size or stage may contribute to WL by increasing body metabolic rate and cumulative REE (23, 24). Markedly reduced physical activity and poor PS in weight-losing cancer patients might further promote cancer cachexia (23). In a cohort of patients with gastroesophageal cancer, PTHrP was positively correlated with CRP and negatively correlated with albumin (25). Elevated PTHrP was also associated with repressed physical activity and energy metabolism independent of hypercalcemia and inflammatory cytokines in rats bearing human tumor xenografts (26). In line with these findings, our results demonstrate that cancer patients with higher serum PTHrP levels presented with elevated CRP, decreased albumin, advanced cancer stage, and poor PS. PTHrP, however, independently predicted WL, even when controlling for inflammation and tumor burden. Meanwhile, the significance of CRP was attenuated when the cancer stage and PS were entered into a multivariate Cox model, suggesting that the effect of tumor burden on the progression of cachexia was largely mediated by an inflammatory process but that PTHrP might reflect a different pathway.

Another key feature of cancer cachexia is increased catabolism, leading to negative energy balance (5). Recent studies suggested that activation of UCP1-expressing beige cells in white adipose tissue depots, a process called browning, switches mitochondrial electron transport from ATP synthesis to thermogenesis, leading to increased energy expenditure and catabolic status (9). In a study using a Lewis lung carcinoma mouse model, PTHrP played a crucial role as mediator of adipose tissue browning in cancer cachexia, and neutralization of PTHrP also reduced muscle wasting (11). These novel findings may provide a plausible explanation for the independent association between PTHrP and WL observed in our study. Our results showed that the discrimination of WL conferred by PTHrP was modest, although multivariate models containing PTHrP gave relatively good discrimination (27). Interestingly, PTHrP injection into healthy mice failed to induce atrophy-associated genes, but PTHrP administration to tumor-bearing mice exacerbated muscle wasting and atrophy-associated gene expression, suggesting that the effect of PTHrP on cancer cachexia depends on other tumor-derived factors (11). Elucidating these factors may enhance discrimination of WL in combination with PTHrP, although these speculations need to be further assessed in prospective investigations.

To our knowledge, this is the first study investigating the association of PTHrP with measured WL as a dependent variable in a longitudinal setting with adjustment for clinically important covariates. Considering the current emphasis on the importance of early recognition of cachexia in cancer patients, the present findings imply that measuring PTHrP may help to identify patients prone to cancer-associated WL early to introduce appropriate treatment and nutritional support (28). Furthermore, our study has implications for novel therapeutic modalities that target PTHrP as a mediator of cancer cachexia; however, it has several limitations that should be addressed with further research. Although we performed statistical adjustment for confounders, the possibility of bias cannot be ruled out due to our study design. Whether serum PTHrP is a mediator or a marker of cancer cachexia could not be conclusively answered by our data; however, PTHrP acted as an independent predictive value for WL in both local and advanced stages of cancer, implying that PTHrP might play as a driver for cancer cachexia rather than merely as a marker. Data regarding dietary intake was not available for the current study, although the impact of caloric intake on WL remains inconsistent in previous reports (22). Data regarding actual energy expenditure (calorimetry) and measurements of brown/beige adipose tissues using positron emission tomography would be helpful to interpret the association between PTHrP and cancer cachexia in details, although it was not available in this study. WL was defined based on single time follow-up weight measurement. Measurements of feeding-regulating neuropeptides were also not available, although related studies demonstrated that action of PTHrP was not linked to modulation of anorexigenic neuropeptides (3).

In summary, our study demonstrated that serum PTHrP levels predicted WL associated with cachexia in a range of cancer patients. This effect was independent of the presence of hypercalcemia, inflammation, tumor burden, or other comorbidities. PTHrP as a potential mediator of cancer cachexia holds promise not only for better understanding the mechanisms of cachexia via fat browning but also as a novel therapeutic target against cachexia, meriting further investigation.

Acknowledgments

We thank Dong-Su Jang, MFA (medical illustrator, Medical Research Support Section, Yonsei University College of Medicine, Seoul, Republic of Korea) for his help with the illustrations.

Author contributions include the following: concept/design by N.H., H.Y., and Y.L.; data analysis/interpretation by N.H., Y.L., H.R.K., Y.R., B.-W.L., and H.C.L.; drafting the article by N.H., H.Y., and Y.L.; critical revision of the article by B.-W.L., Y.R., E.S.K., B.S.C., and H.C.L.; statistics by N.H. and Y.L.; and data collection by N.H., H.Y., and Y.L. All authors were involved in the final approval of manuscript.

This study has not been presented or published elsewhere.

This work was supported by a grant from the Korea Health Technology Research and Development Project through the Korea Health Industry Development Institute, supported by the Ministry of Health and Welfare, Republic of Korea (Grant HI14C2476).

Disclosure Summary: The authors have nothing to disclose.

N.H. and H.Y. contributed equally to this work.

Abbreviations

 
• BFM

  body fat mass

 
• BMI

  body mass index

 
• c

  concordance (statistic)

 
• %change

  percentage change

 
• CHF

  congestive heart failure

 
• CI

  confidence interval

 
• CKD

  chronic kidney disease

 
• COPD

  chronic obstructive pulmonary disease

 
• CRP

  C-reactive protein

 
• ECOG

  Eastern Cooperative Oncology Group

 
• eGFR

  estimated glomerular filtration rate

 
• HR

  hazard ratio

 
• LBM

  lean body mass

 
• PS

  performance status

 
• REE

  resting energy expenditure

 
• UCP1

  uncoupling protein 1

 
• WL

  weight loss

 
• %WL

  percentage WL.