Association of Biological Age with Tumor Microenvironment in Patients with Esophageal Adenocarcinoma

Esophageal cancer is the seventh most common cancer worldwide and typically tends to manifest at an older age, with a median age of 68–70 years at the time of diagnosis [1]. Despite improvements in treatment regimens, the overall prognosis remains poor with an approximate 5-year survival rate of 20% [2]. Survival rates are worse in older patients compared to younger patients and this survival discrepancy has increased over the past recent years [3]. Marked heterogeneity in time-dependent functional decline in older adults results in varying grades of clinically manifest patient fitness or frailty [4]. As opposed to chronological age, the concept of biological age aims to account for this health heterogeneity by integrating personalized clinical and molecular patient characteristics [5]. Remarkably, despite 60% of the patients being over 65 years of age at the time of diagnosis, current treatment guidelines for esophageal cancer do not discriminate between fit and frail older adults [2]. Both chronological age and frailty are associated with higher postoperative complication rates and mortality, and chemotherapy toxicity [6, 7]. In addition, we recently reported a high prevalence of geriatric deficits in older patients with curable esophageal cancer that was subsequently associated with high chemoradiotherapy discontinuation rates [8].

The tumor microenvironment (TME) comprises a dynamic ecosystem of cellular (i.e., immune cells, stroma cells) and non-cellular (i.e., extracellular matrix) components that surround neoplastic cells. Our group and others have previously demonstrated the clinical significance of assessing histologic TME parameters, such as the tumor-stroma ratio (TSR) and tumor immune cell infiltration, in esophageal cancer [9‒11]. A lower TSR reflects high intratumoral stroma quantity and is associated with poor patient prognosis and pathologic response to neoadjuvant chemoradiotherapy in esophageal squamous cell and adenocarcinoma [9, 10]. In contrast, increased tumor infiltration of immune cell subsets is correlated to a favorable patient prognosis in esophageal cancer [11]. Coinciding with local histology, blood cell counts in the circulation were recently shown to be associated with distinct histologic TME characteristics [12‒14]. A report on breast cancer demonstrated increased blood thrombocyte counts in patients with stroma-rich tumors [12]. In addition, blood leukocyte counts were found to be correlated to tumor-infiltrating immune cells [13, 14]. Circulatory blood cell count profiles may therefore reflect local TME features in patients with esophageal cancer. While age-related adaptations have been shown to modulate the TME, it remains unclear whether age-related re-programming of the TME is a result of the biological aging process, or merely dependent on chronological age [15].

In the current work, we aimed to study the association between biological age and the TME in patients with esophageal adenocarcinoma. To this aim, we compared intratumoral stroma quantity, histologic tumor immune cell infiltrate, and blood leukocyte and thrombocyte count over strata of biological age (younger adults <70 years, fit older adults ≥70 years, and frail older adults ≥70 years), as defined by a geriatric assessment (GA).

The patient cohort consisted of patients with esophageal adenocarcinoma, clinical stage I–IV, who underwent treatment at the Leiden University Medical Center or affiliated hospitals (Haag- landen Medisch Centrum, Alrijne Zorggroep, Groene Hart Ziekenhuis and Tergooi Medisch Centrum) between 2010 and 2019. The cohort was categorized into two main groups based on the cut-off age of 70 years. The patient population aged <70 years was part of a study, previously published by our group [10]. The patient population aged ≥70 years is concurrently part of an ongoing prospective cohort study, the Triage of Elderly Needing Treatment (TENT) study. More details on this study were described elsewhere [16]. Patients were diagnosed via esophagogastroduodenoscopy with histologic confirmation. Due to the small number of available patients, no inclusion criteria were based on tumor stage (American Joint Committee on Cancer Staging, 8th edition) or treatment regimens. Clinical data, pathology reports, and blood leukocyte and thrombocyte counts at the time of diagnosis were retrospectively collected from the electronic patient files. Hematoxylin and eosin (H&E) stained pre-treatment primary tumor biopsies were collected. In the case of referred patients, the original biopsy slides were collected from regional hospitals using the Dutch National Pathology Registry (PALGA). All tissue samples were coded and handled according to ethical standards (“Code for Proper Secondary Use of Human Tissue,” Dutch Federation of Medical Scientific Societies). This study was approved by the Medical Ethics Committee of the LUMC (ID number NL53575.058.15). All participants or a proxy provided written informed consent.

Patients included in the TENT study underwent a geriatric assessment (GA) based on the four geriatric domains: the social, somatic, psychological, and functional domain. Patients with a deviant score on ≤1 geriatric domain were classified as fit. Patients who scored abnormally on at least two domains were classified as frail. A domain was considered abnormal if at least one test of the corresponding domain was scored abnormally. The social domain was assessed by asking about the current living situation and was considered abnormal when patients were institutionalized. While it is recognized that home-dwelling older adults living alone may be more susceptible to frailty, pre-frailty falls outside the scope of the current study. Consequently, a living alone status was not classified as abnormal within our analysis of the social domain. The somatic domain contained comorbidity, polypharmacy, and malnutrition assessed by the Mini Nutritional Assessment Short Form (MNA-SF^®; range 0–14, cutoff score ≤11) [17]. Comorbidities were assessed by the Charlson Comorbidity Index (CCI) [18]. The somatic domain was abnormal in the case of CCI ≥1 point (on top of points for solid tumor presence), polypharmacy, and/or MNA-SF^® ≤11. Polypharmacy was defined as ≥5 medications used at the time of diagnosis, without further regard for medication type or dosage [19]. The psychological domain included a history of delirium, dementia, and cognitive impairment according to the Six-Item Cognitive Impairment Test (6CIT; range 0–28, cut-off score ≥8) [20]. The functional domain was abnormal in case of a fall incident in the past 6 months or functional dependency according to the Katz Activities of Daily Living (ADL) questionnaire (range 0–6; cut-off score ≥2) or Lawton Instrumental Activities of Daily Living (IADL) (range 0–5 for men, cut-off score ≤4; range 0–8 for women, cut-off score ≤7) [21].

H&E-stained biopsy samples were digitally scanned using the Pannoramic 250 scanner (3DHISTECH, Budapest, Hungary) at ×20 magnification (0.39 μm per pixel) and stored as.mrxs files.

The TSR was scored on digitalized routine histology hematoxylin and eosin (H&E)-stained slides of pre-treatment biopsies in CaseViewer software (3DHISTEC, Budapest, Hungary). The tumor area with the highest amount of stroma was selected, as described previously in esophageal adenocarcinoma [9]. A detailed description of the methodology and scoring eligibility criteria can be found in the protocol for standardized TSR scoring [22]. In short, the percentage of stroma was scored in increments per ten percent, and the tumor was subsequently categorized as a stroma-high (>50% stroma) or stroma-low (≤50% stroma) tumor. All slides were independently scored by two observers (C.R. and S.H.). When consensus could not be reached, the assessment of a third observer (A.C., board-certified pathologist) was decisive.

The tumor immune cell infiltrate (TICI) was defined as the infiltrating immune cells within the tumor stroma and scored in accordance with the International Immuno-Oncology Biomarker Working Group guidelines for tumor-infiltrating lymphocyte (TIL) assessment in solid tumors [23]. Full sections of the biopsy samples were used to obtain a full assessment of the average stromal immune cell infiltration. Some adaptations to the standardized working group protocol were made. First, in line with our previous report, no distinction was made between infiltrating immune cell subsets [24]. Therefore, tumor immune cell infiltrate, as defined in this study, comprised all inflammatory cells (i.e., mononuclear and polymorphonuclear cells) as opposed to mononuclear cells exclusively. In addition, given the challenging nature of immune infiltrate scoring, the percentage of immune cell infiltration was scored in increments per ten percent as opposed to a continuous scale of single-digit percentages. Tumor immune infiltrate scoring currently lacks a validated cut-off value for risk stratification. In conjunction with other publications on immune cell scoring in esophageal cancer, we applied the median immune cell infiltrate percentage of our cohort as a cut-off value, where an immune cell infiltrate greater than the median percentage was defined as immune-high and an immune cell infiltrate less than or equal to the median percentage was defined as immune-low [11].

The R programming language (version 4.0.5; https://www.r-project.org/) was used for statistical analyses and data visualization (packages tidyverse, corrplot, ggfortify, ggExtra, and viridis). Variable distribution was assessed with the Shapiro-Wilk test, followed by either parametric or non-parametric testing. The Fisher’s exact test or the χ² test was used for categorical variables. For continuous variables, the Mann-Whitney test was used and the median value and interquartile range were reported. Linear regression analysis was used to test the statistical significance of categorical and continuous predictor variables on continuous outcome variables. Interobserver variability was evaluated with Cohen’s Kappa coefficient (Κ). A two-tailed p value of <0.05 was considered statistically significant.

The initial total cohort of 76 esophageal adenocarcinoma patients consisted of three groups, a younger adult (<70 years, n = 38) group, a fit older (≥70 years, n = 19) group, and a frail older group (≥70 years, n = 19). Out of the 76 patients in the cohort, a total of 72 H&E-stained slides proved to be of sufficient quality for histopathologic analysis and were included for further analysis (online suppl. Fig. S1; for all online suppl. material, see https://doi.org/10.1159/000536471).

The baseline tumor and geriatric characteristics of the 72 eligible patients can be found in Table 1. The mean chronological age of the younger adults was 58.5 years, 75.7 years for the fit older adults, and 76.6 years for the frail older adults. The chronological age of the fit and frail older adults was not statistically significantly different (p = 0.393). In addition, younger adults had significantly fewer comorbidities (Charlson Comorbidity Index (CCI) of 0) than fit older adults (CCI of 2, p = 0.005) and frail older adults (CCI of 2, p < 0.001). However, the CCI was not significantly different between the fit and frail older adults (p = 0.373). Frail older adult patients had a higher BMI than fit older adult patients (mean 26.6 vs. 26.0 kg/m², p < 0.001). The number of medications used (p = 0.007) and BMI (p < 0.001) were significantly higher in the frail older adult group versus the younger adult group but were not significantly different between the fit older adult group and the younger adult group. Concerning clinical tumor characteristics, the clinical T stage was significantly different between the younger adult and the separate older adult groups (p < 0.001). However, we observed no significant differences in clinical T (p = 0.667) or N (p = 0.196) stage between the fit and frail older adult groups. Clinical N stage was significantly different between the younger adult and fit older adult group (p = 0.004), but not between the younger adult and frail older adult group (p = 0.208).

The pathology slides of pre-treatment esophageal biopsy tissues were assessed on intratumoral stroma quantity, as scored by the histologic tumor-stroma ratio (TSR), and quantification of the tumor immune cell infiltrate (TICI). Illustrative images of stroma-low and stroma-high tumors can be found in Figure 1a, b. A total of 8 (11.1%) slides required a third review by an independent observer to reach a complete agreement. We observed stroma-high tumors in 29.2% (n = 21) of the patients and stroma-low tumors in 70.8% (n = 51) of the patients in our cohort. Subsequently, the TICI within the stromal compartment was quantified. Illustrative images of TICI scoring can be found in Figure 1c, d. A total of 13 (18.1%) slides required a third review by an independent observer to reach a complete agreement. The Cohen’s kappa coefficient for interobserver variability was 0.86, indicating near-perfect agreement. Given the current lack of a standardized cut-off value, tumors were categorized into immune-high (>40% stromal infiltration) and immune-low (≤40% stromal infiltration) based on the median TICI percentage (40%) in our cohort. We observed immune-high and immune-low tumors in 43.1% (n = 31) and 56.9% (n = 41) of the cases, respectively.

Next, the association between the intratumoral stroma quantity and the biological age subpopulations was evaluated. We first studied the association between chronological age and intratumoral stroma quantity and found no significant differences in stroma percentage between the younger adult (median 50%, IQR: 30–60%) and the total older adult (median 40%, IQR: 30–50%) group (p = 0.080; Fig. 2a). Chronological age was not significantly correlated to intratumoral stroma quantity (Rho = −0.192, p = 0.105; online suppl. Fig. S2a). We then performed linear regression, adjusted for the covariates sex, chronological age, clinical T and N stage, to test if a progressing biological age group as an independent variable predicted intratumoral stroma quantity. Using the fit older adults as a reference group, we found that the frail older adult group significantly predicted decreased intratumoral stroma percentage (B = −14.66% stroma, p = 0.022), whereas the younger adult group did not (B = −4.981% stroma, p = 0.562). Regression statistics can be found in online supplementary Table S1. Further analysis demonstrated significant differences in stroma percentage between the younger adult (median 50%, IQR: 30–60%) and the frail older adult (median 30%, IQR 20–40%) group (p = 0.006) and between the fit (median 40%, IQR: 30–60) and the frail (median 30%, IQR: 20–40%) older adult group (p = 0.014; Fig. 2b, c). The stroma percentages of the fit older group (median 40%, IQR: 30–60) resembled those of the younger adult group (median 50%, IQR: 30–60%; p = 0.876). Applying the standardized cut-off value for the TSR (50%), the proportion of stroma-high (>50% stroma) and stroma-low (≤50% stroma) tumors was significantly different between the strata of biological age (χ² = 8.107, p = 0.017; online suppl. Fig. S2b). Tumors in the younger adult group more often demonstrated a histologic stroma-high phenotype (42.9%) in comparison to tumors in the fit older (26.3%) and frail older adult (5.6%) group.

Next, we studied the association between visually-scored histologic TICI and biological age. The histologic TICI displayed substantial variability among patients within the same group, with median values of 40% (IQR: 30–60%) for younger adults, 40% (IQR: 30–75%) for fit older adults, and 35% (IQR: 30–67.5%) for frail older adults (see Fig. 3a, b). Consequently, no statistically significant differences in TICI were observed between the younger adult and fit older adult group (p = 0.301), the younger adult and frail older adult group (p = 0.389), and the fit and frail older adult group (p = 0.969) (Fig. 3a, b). In addition, histologic TICI demonstrated no significant correlation with chronological age (Rho = 0.070, p = 0.558; online suppl. Fig. S3).

We then combined quantification of the histologic intratumoral stroma quantity and the TICI into a single categorical parameter for histologic TME phenotype. Tumors were subsequently categorized as stroma-low/immune-low (SLIL), stroma-low/immune-high (SLIH), stroma-high/immune-low (SHIL), and stroma-high/immune-high (SHIH). Interestingly, the proportion of tumors with a desert TME phenotype (SLIL) successively increased in the younger adult (34.3%), fit older adult (36.8%), and frail older adult (55.6%) group (Fig. 4a, b). In contrast, the proportion of SHIL tumors, a TME phenotype associated with poor treatment response, markedly decreased in the successive younger adults (25.7%), fit older (15.8%), and frail older (0%) groups [24, 25]. A likewise age-related decrease was observed in the proportion of SHIH tumors, an immune-excluded TME phenotype, in younger adults (17.1%), fit older (10.5%) and frail older (5.6%) adults [25]. Despite these trends, the histologic TME phenotype distribution was statistically significantly different between the younger adult and frail older adult group (χ² = 8.212, p = 0.042), but not between the fit older adult and frail older adult group (χ² = 3.84, p = 0.279), likely due to small group numbers. In addition, we observed no statistically significant difference in histologic TME phenotype between the younger adult and fit older adult group (χ² = 1.800, p = 0.615).

Since we and others recently demonstrated increased transcription of genes associated with thrombocyte activation and increased blood thrombocyte counts in patients with stroma-high tumors, we wondered whether the lower stroma percentages observed in the frail older group were associated with decreased blood thrombocyte counts at the time of diagnosis [12, 26].

Similar to intratumoral stroma quantity, we performed linear regression, adjusted for the covariates sex, chronological age, and clinical T and N stage, to test if a progressing biological age group significantly predicted blood thrombocyte counts. Using the fit older adults as a reference population, both the frail older adult group (B = −44.94 thrombocyte count [10⁹], p = 0.089) and the younger adult group (B = 15.77 thrombocyte count [10⁹], p = 0.654) did not significantly predict blood thrombocyte count, relative to the fit older adults (online suppl. Table S2). Although no statistically significant results were observed in linear regression, the frail older adult group demonstrated a lower median blood thrombocyte count in comparison to the younger adult (median 295 [10⁹], IQR: 256.5–316.5) group (p = 0.032; Fig. 5a). No statistically significant differences were observed in blood thrombocyte counts between the fit older adult (median 255 [10⁹], IQR: 204.5–322.5) and frail older adult (median 239 [10⁹], IQR: 209–285) group (p = 0.486). In addition, we observed no significant correlation between blood thrombocyte count at the time of diagnosis and histologic intratumoral stroma quantity in this study (Rho = 0.215, p = 0.072; online suppl. Fig. S4). Blood thrombocyte count was not correlated to chronological age (Rho = −0.184, p = 0.124).

In linear regression, using the fit older adults as a reference population, we found that both the frail older adult group (B = −0.838 leukocyte count [10⁹], p = 0.276) and the younger adult group (B = −0.865 leukocyte count [10⁹], p = 0.402) did not significantly predict blood leukocyte count (online suppl. Table S3). Blood leukocyte count was not statistically significantly different between the younger adult (median 9 [10⁹], IQR: 7.8–9.9) and fit older adult (median 8.1 [10⁹], IQR 7.0–9.8) group (p = 0.586), the younger adult and frail older adult (median 7.6 [10⁹], IQR: 6.4–10.1) group (p = 0.354), and the fit and frail older adult group (p = 0.476; Fig. 5b). In addition, no significant correlation was found between blood leukocyte count and histologic TICI (Rho = −0.146, p = 0.224) and between blood leukocyte count and chronological age (Rho = −0.132, p = 0.272; online suppl. Fig. S4).

In the current work, we addressed the association between geriatric assessment (GA) and characteristics of the tumor microenvironment (TME) in patients with esophageal adenocarcinoma with an indication for intensive treatment. Tumors in frail older adults demonstrated lower intratumoral stroma quantity in comparison to tumors in chronological-age-matched fit older adults, independent of tumor stage. In addition, the low intratumoral stroma quantity in frail older adults was accompanied by a decreasing trend in blood thrombocyte count. Lastly, we found an increased proportion of tumors with a histologic desert TME phenotype, comprising low stroma quantity and low immune cell infiltration, in frail older adults.

Biological age aims to account for interindividual heterogeneity in health and can markedly contrast chronological age. In the past recent decades, extensive research efforts have been made to formulate reliable predictors of biological age [5]. To this aim, advances in molecular biology have identified associations between patient fitness or frailty and omics data, such as DNA methylation and gene expression profiles [27, 28]. Despite these findings, there is no consensus biomarker for biological age at this time. As the rate of biological aging varies across different tissues within the same individual, a universally applicable biomarker to capture the overall biological age might not be feasible [29]. A robust predictor of biological age in clinical practice may therefore be a context-dependent biomarker that can be used in framed clinical settings only. At this time, there is no specific biomarker for assessing biological age in the clinical oncology setting.

The distinct differences in TME intrinsic properties between biological age groups observed here may ultimately serve as a biomarker for the identification of oncologic frail older adult patients. To this date, only a few studies have elaborated on the relation between the TME and biological age, whereas multiple studies considering the chronologically aging TME report conflicting results [9, 30‒34]. In a recent study by our group, we observed increased intratumoral stroma quantity in progressive chronological age groups (70–<75, 75–<80, 80–<85, 85–<90, and ≥90 years) in patients with breast cancer [30]. The observation of increased intratumoral stroma quantity in older adult patients with breast cancer was previously validated by an independent research group [31]. In contrast, multiple reports on breast cancer patients failed to demonstrate an association between intratumoral stroma quantity and chronological age [32, 33]. Likewise, intratumoral stroma quantity was not associated with chronological age in esophageal cancer [9, 34]. In the current work, we did not observe an association between intratumoral stroma quantity and chronological age but demonstrated significant differences when chronological age groups were further stratified by a clinical parameter for biological age. In light of these findings, the prior analyses of intratumoral stroma quantity and chronological age may therefore potentially be confounded by skewed biological age distributions within the respective study cohorts.

In addition to distinct histologic TME features, we observed a decline in blood thrombocyte counts in the successive biological age groups included in this study. Blood thrombocyte counts were previously shown to decrease with advancing chronological age [35]. Interestingly, the declining blood thrombocyte counts demonstrated a similar trend to the decrease in intratumoral stroma quantity across the biological age groups. Tumor-infiltrating thrombocytes and coagulation factors are key constituents of the TME and impact tumor progression and drug resistance [36]. We recently reported increased transcription of genes associated with thrombocyte degranulation in stroma-high colon tumors [26]. In addition, a recent report on breast cancer described increased thrombocyte-associated protein expression in stroma-high tumors. In subsequent blood analyses, the authors observed increased blood thrombocyte counts in patients with stroma-high tumors [12]. Likewise, we observed decreased intratumoral stroma quantity and blood thrombocyte counts in frail older adult patients. Although we did not observe a statistically significant correlation between blood thrombocyte count and stroma quantity in this study, it is plausible that blood thrombocyte count may reflect histologic stroma quantity and, hence, be informative as a blood biomarker for TME characterization.

We found decreased histologic intratumoral stroma quantity, a phenotype associated with low stromal cell abundance, in tumors of frail older adults compared to tumors of fit older adults. This observation may suggest an increased proliferative arrest of TME stromal cell populations in tumors of frail older adults. Paradoxically, although age-related changes typically lead to growth arrest (cellular senescence), cell loss (apoptosis), and functional decline in stromal and immune cell populations, senescence-associated alterations in TME cell populations have been identified as contributors to tumor progression and metastasis [15]. This paradoxical phenomenon is encapsulated by the term “antagonistic pleiotropy,” signifying a trait that is advantageous in early life but becomes detrimental in later years [37, 38]. One of the key features thought to be responsible for this phenomenon is the acquisition of a senescence-associated secretory phenotype (SASP) of TME cell populations, a distinct secretory profile of senescent cells that demonstrates both pro-tumorigenic and tumor-suppressive properties [39]. The aging stroma accompanies the onset of cellular senescence and the induction of the SASP [40]. As such, senescent cells within the TME can contribute to tumor progression, metastasis, and therapy resistance [41]. Interestingly, a study on breast cancer reported an increased expression of senescence-associated genes in the intratumoral stroma of patients aged ≥80 years old in comparison to a young reference population aged <45 years old [42]. In addition, a recent study on aging and the SASP secretome observed a positive association between circulating SASP protein abundance and clinical patient frailty [43]. In conjunction with these reports, we observed a large fraction of tumors with a desert histologic TME phenotype (i.e., low stroma quantity and low immune cell infiltration) in the frail older adult patient population, illustrative of the biological age-related degenerative adaptations. The age-related increase in the prevalence of desert TME phenotype in our study was accompanied by an age-related decrease in the prevalence of tumors with stroma-high/immune-low and stroma-high/immune-high TME phenotypes. Interestingly, these two TME phenotypes were recently found to be associated with low tumor cell proliferation rates [25]. Our observation may therefore imply an increased SASP within the local TME cell populations in tumors of frail older patients compared to fit older patients. Although the stable cell cycle arrest of senescent cells is presumed to contribute to tumor suppression in later life, the chronic inflammatory characteristics of the SASP secretome have paradoxically been shown to promote tumorigenesis [44]. This could ultimately contribute to the development of highly proliferative tumors in frail older patients. At this time, however, it remains unclear how our biological age-associated histologic findings relate to clinical tumor aggressiveness, which should be the topic of future studies.

Of note, the contrasts in intratumoral stroma quantity and blood thrombocyte counts between younger adult and older adult patients were more pronounced when the older population was further categorized into fit and frail older adult subpopulations. Comprehensive geriatric assessment is widely accepted as the golden standard for determining older adult patient frailty [45]. Our findings provide a biological underpinning for the clinical relevance of assessing frailty, further justifying the need for standardized geriatric assessment in geriatric cancer patients. Recent randomized-controlled trials have shown that comprehensive geriatric assessment may result in chemotherapy dose reduction and overall lower chemo-toxicity in geriatric patients [46]. Patient selection by geriatric assessment may therefore reflect clinically relevant biological differences between fit and frail older adults, amongst which the tumor microenvironment differences described here. Although we were not able to assess tumor prognosis and aggressiveness in our study, our observation of decreased intratumoral stroma quantity and decreased blood thrombocyte counts in frail older adults may suggest a decreased potential of metastasis and aggressiveness of esophageal tumors in this patient population. Interestingly, a significant portion of the tumors in frail older patients displayed a combination of low stroma quantity and high immune cell infiltrates, a TME phenotype linked to a favorable response to immunotherapy [24, 25]. This association was reinforced by a recent study involving resectable esophageal adenocarcinoma patients, which revealed that a high immune cell infiltrate in pre-treatment biopsies correlated with a positive response to neoadjuvant chemoradiation [47]. Furthermore, 55.6% of the tumors in frail older patients exhibited a desert TME phenotype, characterized by both low stroma quantity and low immune cell infiltrate. This desert TME phenotype is known to be linked with highly proliferative tumor cells [25]. These findings collectively suggest overall heightened responsiveness to cancer treatment regimens in the frail older adult patient population, supporting the idea that frail older patients with esophageal cancer may benefit from treatment plans tailored to their biological age, involving less intensive therapeutic agents [8, 16]. Importantly though, in this study, we were not able to discriminate pre-existent frailty from oncology-induced frailty. Since the latter is believed to be the result of increased disease activity, oncology-induced frailty would imply a more aggressive tumor phenotype and thus poorer survival and treatment outcomes [48]. Hence, our results warrant a larger cohort follow-up study to validate the histologic TME features in frail older adults and whether these features are indeed associated with poor patient survival and treatment response.

Strengths of the current study include the use of a standardized GA, including the four geriatric domains, to address patient frailty. In addition, the current work comprises the first study, to the best of our knowledge, to associate histologic TME characteristics with clinical patient frailty in older adults with esophageal cancer. Nevertheless, we want to acknowledge the limitations of our study. First, we performed this pilot study on a small sample size of patients that were included in the TENT study. Subsequently, we were not able to create patient groups with equal clinical baseline characteristics. We, therefore, refrained from prognostic analyses, which currently limits the causal interpretability of our TME findings in frail older adults. Although we do believe our results are valid, our observations may still have been confounded by unaccounted differences in baseline characteristics. Therefore, validation of our results in a larger study that accounts for baseline differences is warranted. Lastly, we adopted a threshold age of 70 years to categorize patients into younger and older subgroups. This specific age was chosen for practical reasons, as in the Netherlands, the legal mandate for frailty assessment begins at the age of 70. Given the varied aging trajectories of individuals, this age threshold may not fully reflect the biological or physiological alterations linked to frailty. Consequently, by categorizing age into distinct groups, our study may have overlooked valuable information about the continuous nature of age-related changes and their correlation with frailty.

The current work demonstrates biological age-related differences in histologic TME phenotypes in esophageal adenocarcinoma. Frailty in older adult patients was associated with decreased histologic intratumoral stroma quantity in comparison to chronological-age-matched fit older adults. In addition, a large proportion of the tumors in frail older adult patients demonstrated a desert histologic TME phenotype, consisting of low stroma quantity and low immune cell infiltration. These frailty-associated histologic findings were accompanied by a coinciding observation of decreased blood thrombocyte counts in frail older adults, a feature that was previously associated with decreased TME activity [12, 36]. Despite the presence of methodological limitations in our study, our results illustrate the stromal-reprogramming effects of biological age and subsequently demonstrate how GA may select for older patient subpopulations with distinct TME features. Given that the TME significantly influences therapeutic response and clinical patient outcome in cancer, the application of GA in older adult patients and, hence, the selection of TME subtype, may guide treatment decisions in the older adult patient population in clinical practice.

We are greatly indebted to the patients participating in the study.

This research was conducted ethically in accordance with the World Medical Association Declaration of Helsinki. All tissue samples were coded and handled according to ethical standards (“Code for Proper Secondary Use of Human Tissue,” Dutch Federation of Medical Scientific Societies). This study was approved by the Biobank Review Committee of the Leiden University Medical Center (ID number NL53575.058.15). All participants or a proxy provided written informed consent.

The authors declare that this study was conducted in the absence of any commercial or financial relationships that may be construed as a potential conflict of interest.

This work was supported by grants from the Bollenstreekfonds, Lisse-Hillegom; the Institute for Evidence-Based Medicine in Old Age (IEMO), and the Vitality Oriented Innovations for the Lifecourse of the Aging Society (VOILA) project. VOILA is funded by ZonMw (project number 457001001). C.R. is funded by an MD/PhD grant from the Leiden University Medical Center (LUMC). None of the above parties has had a role in the design of the study, collection, analysis and interpretation of data, and writing of the manuscript.

C.R., W.M., S.M. and M.S. conceptualized the study. W.M., S.M. and M.S. supervised the project. C.R., S.H. and S.C. performed the formal histologic analysis. Y.H., F.B. and R.T. provided clinical interpretation of the data. W.M., S.M. and M.S. performed funding acquisition. C.R. and Y.H. wrote the manuscript. S.C., S.T., J.P., N.G., D.H., F.B., W.M., S.M. and M.S. discussed and edited the manuscript. All authors reviewed the manuscript and gave final approval for submission.

Cor Ravensbergen and Yara van Holstein contributed equally to this work.Simon Mooijaart and Marije Slingerland shared senior authorship.

Inquiries concerning data requests can be directed to the corresponding author. The data that support the findings of this study are not publicly available due to confidentiality concerns, but are available from the corresponding author upon reasonable request.