Abstract
Background
Obesity increases risk for clear-cell renal cell carcinoma (ccRCC), yet obese patients appear to experience longer survival than nonobese patients. We examined body mass index (BMI) in relation to stage, grade, and cancer-specific mortality (CSM) while considering detection bias, nutritional status, and molecular tumor features.
Methods
Data were available from 2119 ccRCC patients who underwent renal mass surgery at Memorial Sloan-Kettering Cancer Center between 1995 and 2012. Logistic regression models produced associations between BMI and advanced disease. Multivariable competing risks regression models estimated associations between BMI and CSM. Somatic mutation, copy number, methylation, and expression data were examined by BMI among a subset of 126 patients who participated in the Cancer Genome Atlas Project for ccRCC using the Kruskal–Wallis or Fisher exact tests. All statistical tests were two-sided.
Results
Obese and overweight patients were less likely to present with advanced-stage disease compared with normal-weight patients (odds ratio [OR] = 0.61, 95% confidence interval [CI] = 0.48 to 0.79 vs OR = 0.65, 95% CI = 0.51 to 0.83, respectively). Higher BMI was associated with reduced CSM in univariable analyses (P < .005). It remained statistically significant after adjustment for comorbidities and albumin level, but it became non-statistically significant after adjusting for stage and grade (P > .10). Genome-wide interrogation by BMI suggested differences in gene expression of metabolic and fatty acid genes, including fatty acid synthase (FASN), consistent with the obesity paradox.
Conclusions
Our findings suggest that although BMI is not an independent prognostic factor for CSM after controlling for stage and grade, tumors developing in an obesogenic environment may be more indolent.
Renal cell carcinoma (RCC) comprises 85% of kidney cancer and its incidence is increasing, potentially because of more widespread use of diagnostic imagining, with a trend toward smaller tumors being diagnosed (1). Mortality from RCC, however, has not decreased in parallel, suggesting that perhaps not all small tumors are indolent. More than 40% of RCC is attributable to obesity as measured by body mass index (BMI). Recent meta-analyses from prospective observational studies estimate that RCC risk increases by 24% for men and 34% for women for every 5kg/m2 rise in BMI (2). Putative mechanisms through which larger body size increases risk of kidney cancer include chronic tissue hypoxia, altered hormonal milieu, and increased inflammatory response (3). Because these processes can both induce and promote carcinogenesis (4,5), one might expect higher BMI to be an adverse prognostic factor for kidney cancer. Remarkably, a recent systematic review and meta-analysis found overweight and obese kidney cancer patients experienced statistically significantly longer survival than normal-weight patients (6). This phenomenon, known as the “obesity paradox,” has been described among patients on dialysis (7) and with other hemodynamic and metabolic disorders such as heart failure (8), hypertension (9), coronary heart disease (10), diabetes (11), and chronic kidney disease (12), but not among other obesity-related cancers including prostate, colorectal, and postmenopausal breast cancer (13).
The impact of obesity on RCC survival has not been adequately examined, and mechanisms underlying the associations have not been explored. It is not clear whether the obesity paradox is spurious and influenced by detection bias or whether tumors developing in heavier patients are less aggressive and therefore confer a survival advantage. Given that one in three American adults is obese (14) and the incidence of kidney cancer is increasing (1), a better understanding of the obesity paradox is critical to inform secondary prevention efforts and maximize weight management strategies for kidney cancer patients. The purpose of this analysis was to examine the influence of presurgical BMI on stage at presentation and cancer-specific mortality (CSM), while considering the impact of detection bias, nutritional status, and molecular tumor features.
Methods
Study Population
Data were obtained from a prospectively maintained, institutional review board–approved database on 3155 consecutive RCC patients who underwent renal mass surgery at Memorial Sloan-Kettering Cancer Center (MSKCC) between 1995 and 2012. The MSKCC Institutional Review Board reviewed and approved the request for exemption of this retrospective chart review of clinical data and waiver of HIPPA authorization and informed consent. We restricted this analysis to clear-cell RCC (ccRCC; n = 2147) because clinical and pathological features differ markedly by histology (15). After excluding patients lacking BMI data (n = 28), the final sample size for this analysis was 2119. Surgical specimens were processed by standard pathologic techniques and reviewed by genitourinary pathologists. Tumors were staged according to the 2002 American Joint Committee on Cancer (AJCC) Tumor Nodes Metastasis (TNM) classification (16). Presurgical BMI was measured 2 to 6 weeks before surgery for all patients, calculated as patient weight in kilograms divided by their height in meters squared, and categorized into normal weight (BMI < 25kg/m2), overweight (25kg/m2 ≤ BMI < 30kg/m2), and obese (BMI ≥ 30kg/m2) (17). Patients reported whether they had ever been told by a doctor that they had hypertension, hypercholesterolemia, and/or diabetes (yes/no). Chronic kidney disease (CKD) stage was based on glomerular filtration as determined by the CKD Epidemiology Collaboration (18). Patients self-reporting a lifetime history of hypertension, hypercholesterolemia, diabetes, or having a CKD stage of three or greater were considered as having “any comorbidity” (yes/no). Presurgical serum albumin level less than 4g/dL was considered a proxy for poor nutritional status and disease-related weight loss (19,20). Presentation status categorized patients into those diagnosed as a result of local or systemic symptoms vs those who were diagnosed incidentally. Local or systemic symptoms were captured at the presurgical consultation in a standardized fashion and included flank pain, palpable mass, and hematuria for local symptoms or systemic manifestations such as weight loss or anemia.
Statistical Analyses
Associations of patient and disease characteristics with BMI category were tested using the Kruskal–Wallis test when continuous and the χ2 test when categorical. Age and sex-adjusted logistic regression models generated odds ratios (ORs) and 95% confidence intervals (CIs) for the associations between BMI and advanced disease, defined as AJCC stage 3 or greater and as grade 3 or greater. We conducted tests for interaction between BMI and 1) any comorbidity, 2) presentation, and 3) serum albumin to determine whether the associations between BMI and advanced disease differed by the levels of any of these variables.
Follow-up time was calculated from study entry (date of surgery) to death or date of last contact. Cause of death was determined by chart review and corroborated by death certificates. CSM was defined as death due to ccRCC and analyzed using competing events methods, with death from other causes treated as a competing event. Patients still alive or lost to follow-up were censored at last follow-up date. The cumulative incidence of CSM was estimated, and Gray’s test was used to examine differences between groups. Multivariable competing risks regression models were adjusted for demographics and variables determined to be associated with ccRCC mortality. Because we hypothesized that BMI may influence ccRCC mortality through associations with stage and grade, making them intermediates on the causal pathway, we present estimates before and after such adjustment. We also conducted analyses between BMI and overall survival and present these results in the Supplementary Materials (available online). Potential interactions were tested using the likelihood ratio test and a P value less than .10 was considered statistically significant for interaction. Statistical significance was defined as P less than .05. Statistical analyses were conducted using SAS version 9.2 software (SAS Institute, Cary, NC) and R version 2.11.0 (R Development Core Team, Vienna, Austria), including the “survival” and “cmprsk” packages.
To explore the possibility that patients with different BMIs have molecularly distinct tumors, we examined associations between BMI and several genomic features from tumors of 126 MSKCC patients represented in The Cancer Genome Atlas (TCGA) for ccRCC. Available genomic data included somatic mutation (n = 124), copy number (n = 126), global methylation (n = 126), and gene/mRNA expression (n = 122). All data were downloaded through the TCGA data portal (http://tcga-data.nci.nih.gov/tcga/findArchives.htm) and linked with patient characteristics. We examined proportions of overall somatic mutation burden and nonsilent mutations by BMI using the Kruskal–Wallis test. Next, we examined associations between mutations in the top 12 recurrently mutated cancer genes identified in the TCGA (Figure 2) by BMI using Fisher exact tests with Bonferroni correction. To assess differences in copy number events by BMI, we examined both global copy number changes, defined as gains or losses greater than or less than a log2 ratio of 0.2 using the Kruskal–Wallis test, as well as focal copy number events, as determined by the GISTIC 2.0 (21). Global DNA hypermethylation frequency was defined as the percentage of hypermethylated CpG sites out of a total of 15101 loci unmethylated sites in normal kidney tissue and normal white blood cells. The Kruskal–Wallis test was also used to examine differences in methylation frequency by BMI.
Figure 2.
Box plots of genomic alterations stratified by body mass index (BMI). A) Boxplots of nonsilent mutations stratified by BMI (two-sided P value from Kruskal–Wallis test) and recurrent cancer gene nonsilent mutations by BMI. *Indicates statistically significant higher mutation rate of KDM5C in the obese group (P = .03) but not when controlling for multiple testing (P = .34).
B) Boxplots of genome-wide amplifications, losses, and DNA hypermethylation stratified by BMI (two-sided P value from Kruskal–Wallis test).
With respect to gene expression, BMI was treated as both a continuous and binary variable contrasting the extreme categories (obese vs normal weight). Wilcoxon rank sum test was used to compare genes up- and downregulated in the obese vs normal-weight groups, and Spearman rank correlation was used to characterize the correlation between gene expression and BMI. We further performed pathway analysis of gene expression data. High ranking genes both positively and negatively correlated with BMI were analyzed using DAVID Bioinformatics Resources 6.7 (22). Finally, we performed gene set enrichment analysis (23) contrasting the extreme BMI categories. All gene sets (MSigDB) of size 15 to 500 genes (n = 5332 gene sets) were used in the gene set enrichment analysis. Student t test was used to quantify differential gene expression between the two groups. To account for gene–gene correlations in the enrichment analysis, gene set enrichment P values were computed with respect to a null distribution obtained from 1000 randomizations of the patient phenotype labels. Limited sample size prohibited conducting analyses stratified by sex or race. All statistical tests were two-sided
Results
The characteristics of the 2119 ccRCC patients available for this analysis are summarized in Table 1. Our study population had a median age at surgery of 61 years (interquartile range [IQR] = 52–70) and was predominantly male (66.4%) and white (91.3%). Of the patients, 19.8% (n = 420) were classified as normal weight, 38.0% (n = 806) were classified as overweight, and 42.1% (n = 893) were classified as obese. Age at surgery, sex, race, hypertension, diabetes, hypercholesterolemia, CKD stage, AJCC stage, grade, tumor size, albumin, and type of surgical procedure each differed statistically significantly by BMI (Table 1) (all P < .05). The difference in median tumor size, however, was 0.5cm or less and therefore not regarded as clinically meaningful.
Table 1.
Characteristics of 2119 clear-cell renal cell carcinoma patients by body mass index* category
| Characteristic | Overall | BMI category | P† | ||
|---|---|---|---|---|---|
| Normal (n = 420; 19.8%) | Overweight (n = 806; 38.0%) | Obese (n = 893; 42.1%) | |||
| Age, y, median (IQR) | 60.8 (52.1–69.6) | 61.4 (51.9–71.2) | 61.5 (53.4–70.4) | 59.8 (51.5–67.7) | <.001 |
| Sex | <.001 | ||||
| Male | 1408 (66.4) | 245 (58.3) | 587 (72.8) | 576 (64.5) | |
| Female | 711 (33.6) | 175 (41.7) | 219 (27.2) | 317 (35.5) | |
| Race | .001 | ||||
| White | 1935 (91.3) | 368 (87.6) | 736 (91.3) | 831 (93.1) | |
| Other | 164 (7.7) | 50 (11.9) | 58 (7.2) | 56 (6.3) | |
| Missing | 20 (0.9) | 2 (0.5) | 12 (1.5) | 6 (0.7) | |
| Hypertension | <.001 | ||||
| Yes | 1145 (54.0) | 164 (39.0) | 423 (52.5) | 558 (62.5) | |
| No | 974 (46.0) | 256 (61.0) | 383 (47.5) | 335 (37.5) | |
| Diabetes | <.001 | ||||
| Yes | 323 (15.2) | 30 (7.1) | 119 (14.8) | 174 (19.5) | |
| No | 1796 (84.8) | 390 (92.9) | 687 (85.2) | 719 (80.5) | |
| Hypercholesterolemia | <.001 | ||||
| Yes | 616 (29.1) | 75 (17.9) | 251 (31.1) | 290 (32.5) | |
| No | 1503 (70.9) | 345 (82.1) | 555 (68.9) | 603 (67.5) | |
| CKD stage | .04 | ||||
| 1 | 300 (14.2) | 75 (17.9) | 96 (11.9) | 129 (14.4) | |
| 2 | 1149 (54.2) | 235 (56.0) | 457 (56.7) | 457 (51.2) | |
| 3 | 642 (30.3) | 105 (25.0) | 245 (30.4) | 292 (32.7) | |
| 4 | 19 (0.9) | 3 (0.7) | 6 (0.7) | 10 (1.1) | |
| 5 | 3 (0.1) | 1 (0.2) | 1 (0.1) | 1 (0.1) | |
| Missing | 6 (0.3) | 1 (0.2) | 1 (0.1) | 4 (0.4) | |
| AJCC stage | <.001 | ||||
| 1 | 1325 (62.5) | 228 (54.3) | 518 (64.3) | 579 (64.8) | |
| 2 | 98 (4.6) | 22 (5.2) | 35 (4.3) | 41 (4.6) | |
| 3 | 506 (23.9) | 114 (27.1) | 178 (22.1) | 214 (24.0) | |
| 4 | 188 (8.9) | 56 (13.3) | 75 (9.3) | 57 (6.4) | |
| Missing | 2 (0.1) | 0 (0.0) | 0 (0.0) | 2 (0.2) | |
| Grade | .008 | ||||
| 1 | 98 (4.6) | 18 (4.3) | 35 (4.3) | 45 (5.0) | |
| 2 | 1095 (51.7) | 202 (48.1) | 426 (52.9) | 467 (52.3) | |
| 3 | 738 (34.8) | 144 (34.3) | 268 (33.3) | 326 (36.5) | |
| 4 | 170 (8.0) | 50 (11.9) | 69 (8.6) | 51 (5.7) | |
| Missing | 18 (0.8) | 6 (1.4) | 8 (1.0) | 4 (0.4) | |
| Tumor size, cm, median (IQR) | 4.0 (2.5–6.5) | 4.5 (2.6–7.6) | 4.0 (2.5–6.5) | 4.0 (2.6–6.1) | .02 |
| Albumin | .003 | ||||
| <4ng/dL | 517 (24.4) | 127 (30.2) | 190 (23.6) | 200 (22.4) | |
| ≥4ng/dL | 1481 (69.9) | 262 (62.4) | 575 (71.3) | 644 (72.1) | |
| Missing | 121 (5.7) | 31 (7.4) | 41 (5.1) | 49 (5.5) | |
| Surgical procedure | <.001 | ||||
| Radical | 1029 (48.6) | 245 (58.3) | 378 (46.9) | 406 (45.5) | |
| Partial | 1090 (51.4) | 175 (41.7) | 428 (53.1) | 487 (54.5) | |
| Any comorbidity‡ | <.001 | ||||
| Yes | 1505 (71.0) | 248 (59.0) | 567 (70.3) | 690 (77.3) | |
| No | 611 (28.8) | 171 (40.7) | 239 (29.7) | 201 (22.5) | |
| Missing | 3 (0.2) | 1 (0.3) | 0 (0.0) | 2 (0.2) | |
| Presentation | .33 | ||||
| Incidental | 1534 (72.4) | 293 (69.8) | 595 (73.8) | 646 (72.4) | |
| Local or systemic | 566 (26.7) | 124 (29.5) | 207 (25.7) | 235 (26.3) | |
| Missing | 19 (0.9) | 3 (0.7) | 4 (0.5) | 12 (1.3) | |
* Data are No. (%) unless otherwise noted. AJCC = American Joint Committee on Cancer; BMI = body mass index; CKD = chronic kidney disease; IQR = interquartile range.
† P value from χ2 test when categorical and Kruskal-Wallis test when continuous.
‡ Patients endorsing a lifetime history of hypertension, hypercholesterolemia, diabetes, or having a chronic kidney disease stage of 3 or greater were considered as having any comorbidity.
Associations between BMI and advanced disease are presented in Table 2. BMI was statistically significantly and inversely associated with AJCC stage (P < .001) and to a lesser extent with grade (P = .046). Overweight (OR = 0.61; 95% CI = 0.48 to 0.79) and obese (OR = 0.65; 95% CI = 0.51 to 0.83) patients had lower odds of presenting with advanced stage as compared with normal-weight patients. There were not statistically significant interactions between BMI and any comorbidity, presentation, or serum albumin on either stage or grade (all P > .22).
Table 2.
Logistic regression analysis for associations between body mass index and advanced disease*
| Body mass index | Advanced stage (AJCC stage 3 or greater) | Advanced grade (grade 3 or greater) |
|---|---|---|
| OR (95% CI) | OR (95% CI) | |
| Normal | 1.0 (referent) | 1.0 (referent) |
| Overweight | 0.61 (0.48 to 0.79) | 0.73 (0.57 to 0.94) |
| Obese | 0.65 (0.51 to 0.83) | 0.81 (0.64 to 1.03) |
| P† | <.001 | .046 |
* AJCC = American Joint Committee on Cancer; CI = confidence interval; OR = odds ratio.
† P value from logistic regression adjusted for age at surgery and sex.
Median follow-up time among survivors was 4 years (IQR = 2–7). During follow-up, 481 patients died; 51.1% (n = 246) of these deaths were due to RCC. BMI was inversely associated with higher CSM (P = .005) in univariable analysis (Figure1A), and there was no statistically significant interaction between BMI and albumin (P interaction = .69) (Figure 1B). Notably, the overweight/obese patients with normal albumin experienced the lowest cumulative incidence of cancer-specific death (P < .001). Table 3 presents results from multivariable models of BMI and CSM before and after the adjustment for stage and grade. BMI remained inversely associated with CSM (P = .01) in a model adjusted for demographics and relevant confounders. Compared with normal-weight patients, obese patients had reduced risk of CSM (hazard ratio [HR] = 0.59; 95% CI = 0.42 to 0.83). However, as expected, adjustment for stage and grade attenuated the inverse association, and the P value became non-statistically significant (HR = 0.75; 95% CI = 0.53 to 1.07; P = .12). Similar results were seen for overall survival (data not shown).
Figure 1.
Cancer-specific mortality curves. A) Cancer-specific mortality stratified by body mass index (BMI). B) Cancer-specific mortality stratified by by combinations of BMI and albumin (ALB) (P interaction between BMI and albumin = .69). Two-sided P values are from Gray’s test.
Table 3.
Multivariable competing risks regression for the association between body mass index and cancer-specific death*
| Variable | Before adjustment for stage and grade | After adjustment for stage and grade | ||
|---|---|---|---|---|
| HR (95% CI) | P | HR (95% CI) | P | |
| BMI | .01 | .13 | ||
| Normal | 1.0 (referent) | 1.0 (referent) | ||
| Overweight | 0.73 (0.53 to 1.02) | 1.02 (0.72 to 1.46) | ||
| Obese | 0.59 (0.42 to 0.83) | 0.75 (0.53 to 1.07) | ||
| Age at surgery | 1.01 (1.00 to 1.02) | .03 | 1.00 (0.99 to 1.01) | .93 |
| Sex | .002 | .94 | ||
| Male | 1.0 (referent) | 1.0 (referent) | ||
| Female | 0.61 (0.45 to 0.83) | 1.01 (0.74 to 1.39) | ||
| Race | .12 | .06 | ||
| White | 1.0 (referent) | 1.0 (referent) | ||
| Other | 0.61 (0.32 to 1.15) | 0.53 (0.28 to 1.04) | ||
| AJCC stage | <.001 | |||
| 1–2 | NA | NA | 1.0 (referent) | |
| 3–4 | NA | NA | 7.47 (5.2 to 10.73) | |
| Grade | <.001 | |||
| 1–2 | NA | NA | 1.0 (referent) | |
| 3–4 | NA | NA | 3.70 (2.62 to 5.23) | |
| Hypertension | .05 | .01 | ||
| Yes | 0.75 (0.56 to 1.00) | 0.69 (0.51 to 0.93) | ||
| No | 1.0 (referent) | 1.0 (referent) | ||
| Hypercholesterolemia | .12 | .06 | ||
| Yes | 0.76 (0.54 to 1.08) | 0.71 (0.50 to 1.01) | ||
| No | 1.0 (referent) | 1.0 (referent) | ||
| Albumin | <.001 | .01 | ||
| <4g/dL | 2.71 (2.07 to 3.54) | 1.45 (1.09 to 1.94) | ||
| ≥4g/dL | 1.0 (referent) | 1.0 (referent) | ||
* AJCC = American Joint Committee on Cancer; BMI = body mass index; CI = confidence interval; HR = hazard ratio; NA = not applicable.
Characteristics of the 126 patients included in genomic analysis are detailed in Supplementary Table 1 (available online). As in our larger cohort, obese patients had reduced risk of CSM (data not shown). In our genomic analyses, no statistically significant associations between BMI and somatic mutations, DNA hypermethylation status, or global copy number events were observed (Figure 2). KDM5C mutations were statistically significantly more frequent in the obese cohort (P = .03) but not after correction for multiple testing (P = .34). GISTIC copy number analysis of statistically significant recurrent events showed enrichment for a broad low-level copy number gain of chromosome 7q encompassing several hundred genes (P = .02; q = 0.03) in the normal-weight patients (Supplementary Table 2, available online). Figure 3A presents a heat map of genes differentially up- or downregulated in tumors from normal-weight vs obese patients at a nominal P value less than .001, including several metabolic genes. Notably, fatty acid synthase (FASN) was statistically significantly upregulated in the normal BMI group and downregulated in the obese group. Because increased FASN expression is associated with aggressive disease and poor prognosis in several cancer types, including RCC (24–26), we examined the association between FASN upregulation and CSM in our patients (Figure 3B). FASN upregulation was statistically significantly associated with increased incidence of cancer-specific death (P < .001). This pattern of results was confirmed in the remaining TCGA cohort (data not shown). Differential expression of CYP4A22 and CYP4A11 by BMI status was also observed (Figure 3A). Both genes are enzymes of the cytochrome P450 superfamily, which catalyze the omega- and (omega-1)-hydroxylation of various medium chain fatty acids and are regulated by peroxisome proliferator-activated receptors (PPARs) (27). Remarkably, HMGCS2, and PCK1, two additional genes involved in the PPAR signaling pathway, were also ranked in the top 30 upregulated genes when comparing the normal-weight vs obese groups (Supplementary Tables 3 and 4 and Supplementary Figure 1, available online). Finally, we performed gene set enrichment analysis comparing obese patients with normal-weight patients. Again, enrichment for several gene sets related to fatty acid metabolism (ranked 8 out of 5332 gene sets) and fatty acid oxidation (ranked 12) were observed in obese patients (Supplementary Table 5, available online).
Figure 3.
Gene expression analysis linking body mass index (BMI), fatty acid metabolism, and survival from The Cancer Genome Atlas (TCGA) dataset. A) Heat map of gene expression by BMI. B) Cancer-specific mortality stratified by fatty acid synthase (FASN) in Memorial Sloan-Kettering Cancer Center (MSKCC) TCGA cohort (n = 122) Two-sided P value is from Gray’s test. C) Diagram illustrates the relationship between BMI, survival, and FASN pathway expression. Red color indicates gene or protein upregulation. Blue color indicates downregulation. Rectangle indicates mRNA expression. Diamond indicates protein expression.
Discussion
In our large, single-institute, clinical cohort of ccRCC patients who underwent renal mass surgery, we found that BMI was inversely associated with advanced stage, which did not differ by earlier detection related to any comorbidity (defined as self-reported lifetime history of hypertension, hypercholesterolemia, diabetes, or having a CKD stage ≥3), symptoms at presentation, or serum albumin level. Our genomic analyses of tumors from patients with different BMIs suggest that the survival differences are not because of specific mutations, increased genomic instability, or DNA hypermethylation. Rather, statistically significant differences in gene expression were noted in metabolic genes in a way that supports the obesity paradox. Collectively our results suggest that the decreased mortality observed among obese ccRCC patients may not merely be explained by detection bias or weight loss but that tumors developing in obese patients may be more indolent than those in normal-weight patients.
The obesity paradox is an established phenomenon in RCC and other metabolic disorders such as diabetes and chronic kidney disease (11,12). Most recently, Choi et al. (6) examined associations between BMI and RCC-specific survival among 1543 Korean patients treated by nephrectomy and incorporated their results into a systematic review and meta-analysis. In their initial study, they found a statistically significant inverse association between high BMI and cancer-specific death (HR = 0.47; 95% CI = 0.29 to 0.77) which persisted after multivariable adjustment for stage, tumor size, grade, symptom presence, and baseline weight loss. The statistically significant association also remained after they performed sensitivity analyses restricted to patients without any weight loss, with early-stage disease, with clear-cell histology, and when they excluded patients who died in the first 2 years of follow-up. Their meta-analysis, which included 14 additional studies, produced a pooled hazard ratio of 0.59 (95% CI = 0.48 to 0.74) for cancer-specific survival. The authors concluded that research is needed to elucidate biological mechanism responsible for the obesity paradox in RCC.
To our knowledge only one prior study examined a somatic tumor alteration and BMI among ccRCC patients. Similar to our results, Smits et al. (28) found no association between BMI and VHL mutation status (P = .96). We extend these findings by interrogating multiple genomic platforms by BMI categories in the context of survival. Although there is no association between BMI and genetic difference on a DNA level, our analysis of global gene expression changes suggest that the survival advantage conferred by high BMI may be related to differences in fatty acid metabolism (Figure 3). Tumors from obese patients differed statistically significantly from those of normal-weight patients in both FASN and the immediate upstream enzyme acetyl-CoA carboxylase (ACACA) gene and its encoded protein ACC (Figure 3C). Both genes encode for rate-limiting enzymes involved in fatty acid synthesis, a process essential for tumor growth (29). FASN is a metabolic oncogene that is overexpressed in many cancers, including ccRCC. Horiguchi et al. has shown that FASN overexpression assessed by immunohistochemistry is associated with aggressive RCC and shorter cancer-specific survival and that pharmacological inhibition of FASN can reduce RCC tumor growth in vitro (24,30). Similarly, our analysis suggests that FASN upregulation is associated with worse ccRCC survival (Figure 3B). Further, we found that FASN is downregulated in obese patients but upregulated in normal-weight patients, a pattern that may provide mechanistic insights into the obesity paradox. A lower expression of FASN among obese colorectal cancer patients from the Nurses’ Health Study was recently reported by Kuchiba et al. (31), who speculated that FASN may confer a selective growth advantage to cells upon nutritional deprivation. Other studies among colorectal and prostate cancer patients suggest that the adverse impact of FASN overexpression is limited to obese patients (25,26), a finding we did not see in ccRCC (data not shown).
The interpretation of these results in the context of ccRCC is difficult and complicated by the fact that the obesity paradox is not observed in these other malignancies. Whether BMI’s contribution to FASN regulation is a cancer-specific phenomenon certainly requires further investigation, and our findings should be viewed as hypothesis-generating given the current paucity of mechanistic evidence. However, one possible link between obesity and lipogenic gene downregulation could come through the effects of the apoptotic protein TRAIL (TNF-related apoptosis-inducing ligand), which is reported to be elevated in the serum of obese individuals (32). A recent report linked TRAIL to downregulation of FASN, ACC, and Glut-4 through the cleavage of caspase 3 and caspase 8 and the inactivation of PPARγ (33). Our pathway analysis also suggests that tumors in obese patients have enhanced PPAR and fatty acid signaling, which, in turn, may be associated with more differentiated tumors in ccRCC.
Strengths of our study include its large sample size of ccRCC cases with rich clinical and genomic annotation, centralized pathology review, and long-term follow-up. We acknowledge that BMI is an indirect measurement of adiposity that does not distinguish between adipose tissue and lean body mass, nor does it reflect metabolic and endocrine disruptions associated with obesity (34). Prognostic studies for ccRCC that incorporate metabolomics are needed. In addition, a cross-sectional measurement of BMI also does not take into account disease-related weight loss, which could skew the findings toward worse survival among lower-weight patients. To address this issue, serum albumin level (a marker of nutritional status) was included in the models, yet findings persisted. In addition, only eight patients were classified as underweight (BMI < 18.5kg/m2), and all patients were healthy enough to undergo nephrectomy. Excluding the underweight patients did not change the findings (data not shown).
Our study is not without limitations. We note that our definition of comorbidity was based on only four diseases. Overweight and obese patients may have a higher prevalence of many other comorbidities that could lead to imaging, incidental detection, and potentially less-aggressive tumors. Finally, we acknowledge that selection bias may be a contributing factor to our observed associations because our analysis was conducted among RCC patients (35). Generalizability of our findings is limited primarily to white patients, reflecting the demographics of our tertiary referral center.
In conclusion, our findings suggest that although BMI is not an independent prognostic factor for CSM after controlling for stage and grade, tumors of obese patients may be more indolent than those of normal weight patients, a pattern that is supported by alterations in gene expression signatures.
Funding
This work was supported by the National Cancer Institute (T32 CA082088); the Stephen P. Hanson Family Fund Fellowship in Kidney Cancer (to AHH); the Paula Moss Trust for the research into the cure and treatment of kidney cancer (to JJH); and The Cancer Genome Atlast grant (NCI-U24CA143840 to AJ, GC, NS, CS).
Supplementary Material
AH Hakimi and H. Furberg contributed equally to this work. JJ Hsieh and P. Russo contributed equally to the study. AA Hakimi, H. Furberg, EC Zabor, JJ Hsieh, and P. Russo had roles in study concept and design, data acquisition and analysis, interpretation of data, and manuscript preparation. N. Mikklineni and B. Fiegoli assisted with chart review and data acquisition. AA Hakimi, A. Jacobsen, G. Ciriello, N. Schultz, C. Sander, H. Shen, and PW Laird had roles in analysis of the TCGA data set. PH Kim, MH Voss, VE Reuter, and RJ Motzer provided insights into manuscript preparation.
The study sponsors had no role in the design of the study; the collection, analysis, and interpretation of the data; the writing of the manuscript; and the decision to submit the manuscript for publication. RJ Motzer has served as a consultant for Pfizer and received research funding from Pfizer, Novartis, Glaxo Smith Kline, Bristol-Myers Squibb, Aveo, and Eisai. P. Russo has served as a consultant with Wilex AG. All other authors have no conflicts of interest to declare.
References
- 1. Hollingsworth JM, Miller DC, Daignault S, et al. Rising incidence of small renal masses: a need to reassess treatment effect. J Natl Cancer Inst. 2006;98(18):1331–1334 [DOI] [PubMed] [Google Scholar]
- 2. Renehan AG, Tyson M, Egger M, et al. Body-mass index and incidence of cancer: a systematic review and meta-analysis of prospective observational studies. Lancet. 2008;371(9612):569–578 [DOI] [PubMed] [Google Scholar]
- 3. Roberts DL, Dive C, Renehan AG. Biological mechanisms linking obesity and cancer risk: new perspectives. Annu Rev Med. 2010;61:301–316 [DOI] [PubMed] [Google Scholar]
- 4. Vucenik I, Stains JP. Obesity and cancer risk: evidence, mechanisms, and recommendations. Ann N Y Acad Sci. 2012;1271:37–43 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Klinghoffer Z, Yang B, Kapoor A, et al. Obesity and renal cell carcinoma: epidemiology, underlying mechanisms and management considerations. Expert Rev Anticancer Ther. 2009;9(7):975–987 [DOI] [PubMed] [Google Scholar]
- 6. Choi Y, Park B, Jeong BC, et al. Body mass index and survival in patients with renal cell carcinoma: a clinical-based cohort and meta-analysis. Int J Cancer. 2013;132(3):625–634 [DOI] [PubMed] [Google Scholar]
- 7. Kovesdy CP, Anderson JE. Reverse epidemiology in patients with chronic kidney disease who are not yet on dialysis. Semin Dial. 2007;20(6):566–569 [DOI] [PubMed] [Google Scholar]
- 8. Oreopoulos A, Padwal R, Kalantar-Zadeh K, et al. Body mass index and mortality in heart failure: a meta-analysis. Am Heart J. 2008;156(1):13–22 [DOI] [PubMed] [Google Scholar]
- 9. Uretsky S, Messerli FH, Bangalore S, et al. Obesity paradox in patients with hypertension and coronary artery disease. Am J Med. 2007;120(10):863–870 [DOI] [PubMed] [Google Scholar]
- 10. Romero-Corral A, Montori VM, Somers VK, et al. Association of bodyweight with total mortality and with cardiovascular events in coronary artery disease: a systematic review of cohort studies. Lancet. 2006;368(9536):666–678 [DOI] [PubMed] [Google Scholar]
- 11. Carnethon MR, De Chavez PJ, Biggs ML, et al. Association of weight status with mortality in adults with incident diabetes. JAMA. 2012;308(6):581–590 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Kalantar-Zadeh K, Streja E, Kovesdy CP, et al. The obesity paradox and mortality associated with surrogates of body size and muscle mass in patients receiving hemodialysis. Mayo Clin Proc. 2010;85(11):991–1001 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Parekh N, Chandran U, Bandera EV. Obesity in cancer survival. Annu Rev Nutr. 2012;32:311–342 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Flegal KM, Carroll MD, Kit BK, et al. Prevalence of obesity and trends in the distribution of body mass index among US adults, 1999–2010. JAMA. 2012;307(5):491–497 [DOI] [PubMed] [Google Scholar]
- 15. Cheville JC, Lohse CM, Zincke H, et al. Comparisons of outcome and prognostic features among histologic subtypes of renal cell carcinoma. Am J Surg Pathol. 2003;27(5):612–624 [DOI] [PubMed] [Google Scholar]
- 16. American Joint Committee on Cancer AJCC Cancer Staging Manual. 6th ed New York: Springer; 2002 [Google Scholar]
- 17. Obesity: preventing and managing the global epidemic Report of a WHO consultation. World Health Organ Tech Rep Ser. 2000;894:i–xii, 1–253 [PubMed] [Google Scholar]
- 18. Levey AS, Stevens LA, Schmid CH, et al. A new equation to estimate glomerular filtration rate. Ann Intern Med. 2009;150(9):604–612 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Baker JP, Detsky AS, Wesson DE, et al. Nutritional assessment: a comparison of clinical judgement and objective measurements. N Engl J Med. 1982;306(16):969–972 [DOI] [PubMed] [Google Scholar]
- 20. Gibbs J, Cull W, Henderson W, et al. Preoperative serum albumin level as a predictor of operative mortality and morbidity: results from the National VA Surgical Risk Study. Arch Surg. 1999;134(1):36–42 [DOI] [PubMed] [Google Scholar]
- 21. Mermel CH, Schumacher SE, Hill B, et al. GISTIC2.0 facilitates sensitive and confident localization of the targets of focal somatic copy-number alteration in human cancers. Genome Biol. 2011;12(4):R41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Huang da W, Sherman BT, Lempicki RA. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc. 2009;4(1):44–57 [DOI] [PubMed] [Google Scholar]
- 23. Subramanian A, Tamayo P, Mootha VK, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A. 2005;102(43):15545–15550 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Horiguchi A, Asano T, Asano T, et al. Fatty acid synthase over expression is an indicator of tumor aggressiveness and poor prognosis in renal cell carcinoma. J Urol. 2008;180(3):1137–1140 [DOI] [PubMed] [Google Scholar]
- 25. Ogino S, Nosho K, Meyerhardt JA, et al. Cohort study of fatty acid synthase expression and patient survival in colon cancer. J Clin Oncol. 2008;26(35):5713–5720 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Nguyen PL, Ma J, Chavarro JE, et al. Fatty acid synthase polymorphisms, tumor expression, body mass index, prostate cancer risk, and survival. J Clin Oncol. 2010;28(25):3958–3964 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Johnson EF, Palmer CN, Griffin KJ, et al. Role of the peroxisome proliferator-activated receptor in cytochrome P450 4A gene regulation. FASEB J. 1996;10(11):1241–1248 [DOI] [PubMed] [Google Scholar]
- 28. Smits KM, Schouten LJ, Hudak E, et al. Body mass index and von Hippel-Lindau gene mutations in clear-cell renal cancer: results of the Netherlands Cohort Study on diet and cancer. Ann Epidemiol. 2010;20(5):401–404 [DOI] [PubMed] [Google Scholar]
- 29. Menendez JA, Lupu R. Fatty acid synthase and the lipogenic phenotype in cancer pathogenesis. Nat Rev Cancer. 2007;7(10):763–777 [DOI] [PubMed] [Google Scholar]
- 30. Horiguchi A, Asano T, Asano T, et al. Pharmacological inhibitor of fatty acid synthase suppresses growth and invasiveness of renal cancer cells. J Urol. 2008;180(2):729–736 [DOI] [PubMed] [Google Scholar]
- 31. Kuchiba A, Morikawa T, Yamauchi M, et al. Body mass index and risk of colorectal cancer according to fatty acid synthase expression in the nurses’ health study. J Natl Cancer Inst. 2012;104(5):415–420 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Choi JW, Song JS, Pai SH. Associations of serum TRAIL concentrations, anthropometric variables, and serum lipid parameters in healthy adults. Ann Clin Lab Sci. 2004;34(4):400–404 [PubMed] [Google Scholar]
- 33. Keuper M, Wernstedt Asterholm I, Scherer PE, et al. TRAIL (TNF-related apoptosis-inducing ligand) regulates adipocyte metabolism by caspase-mediated cleavage of PPARgamma. Cell Death Dis. 2013;4:e474. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Florez H, Castillo-Florez S. Beyond the obesity paradox in diabetes: fitness, fatness, and mortality. JAMA. 2012;308(6):619–620 [DOI] [PubMed] [Google Scholar]
- 35. Berrington de Gonzalez A, Morton LM. Converting epidemiologic studies of cancer etiology to survivorship studies: approaches and challenges. Cancer Epidemiol Biomarkers Prev. 2012;21(6):875–880 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.