跳到正文
原文
Nature 健康护理主题·· 2026-10-12AI 评分61

G-CODE评分在老年癌症患者中预测6、12和24个月死亡风险的开发与外部验证

Using the Geriatric Core Dataset (G-CODE) to predict mortality in older patients with cancer: development and external validation

AI 导读

研究基于法国ELCAPA和GOM两个前瞻性队列,开发并外部验证了基于老年核心数据集(G-CODE)的预后评分。最终模型包含9个项目(总分0-17分),主要预测因子为无照护者、体重下降、合并症、日常生活能力受损和行动时间试验异常。结合年龄、癌症类型、转移和治疗后,该模型预测6、12和24个月死亡风险的tAUC分别为82%、78%和80%,用于老年癌症患者风险分层。

正文

Abstract

Background

The Geriatric Core Dataset (G-CODE) was developed to standardise geriatric assessment (GA) in older adults with cancer. This study aimed to develop and externally validate a mortality prediction score based on the G-CODE at 6-month (main endpoint), 12- and 24-month (secondary endpoints), in older patients with cancer.

Methods

Data came from two prospective French cohorts of cancer patients aged ≥70 referred for GA: the Elderly Cancer Patients (ELCAPA) cohort (training set) and the Geriatric Oncology Marseille (GOM) cohort (validation set). G-CODE items were evaluated using multivariate Cox regression analyses, with Akaike information criterion selection. Discrimination was assessed using time-dependent area under the receiver operator curve (tAUC) for 6-, 12-, and 24-month mortality.

Results

We included 950 and 766 patients in the training and validation sets (mean ± sd age: 83 ± 6 years/81 ± 6 years; men: 55%/59%; 6-month mortality: 27%/22%). The final G-CODE model highlighted five strong predictors of 6-month mortality: not having a caregiver, weight loss, comorbidities, an impaired 4-instrumental activities of daily living score, and an impaired timed-up-and-go test. The tAUCs of the model, including the G-CODE, age, cancer type, metastases, and treatment, were 82%, 78% and 80%.

Conclusions

The final G-CODE score (range: 0–17) accurately predicts 6-, 12-, and 24-month mortality.

Clinical trial registration

NCT02884375 (ClinicalTrials.gov).

Introduction

In 2012, adults aged 65 or over accounted for 48% of cancer cases, and this proportion is forecast to rise to 58% by 2035 [1]. Even though there is evidence to show that cancer is a disease of ageing, we still lack solid data on the optimal cancer treatments in older adults – largely due to the under-representation of this specific population in oncology clinical trials [2,3,4,5].

Furthermore, the ageing process is heterogeneous [6]. Older adults of the same chronological age may or may not have various ageing-associated conditions that can influence the tolerance and response to cancer treatments and thus their overall survival [7,8,9,10]. International guidelines emphasise the need to improve the quality of care for older adults and to include a geriatric assessment (GA) in oncology clinical trials [11,12,13].

In a GA, a healthcare professional evaluates several key domains of relevance to the older adult’s health status: functional status, comorbidities, cognitive and psychological status, nutritional status, concurrent medications, and social support [14]. Some studies have shown that GA is strongly predictive of chemotherapy toxicity, treatment completion, overall survival, early death, and quality of life [13, 15,16,17,18,19,20]. The results of a GA also help to modify treatment decisions in terms of dosing, intensity, the treatment regimen, and supportive care [14]. Despite these benefits, a GA is rarely used in oncology clinical trials [21]. A recent systematic review found that a GA was primarily used to examine associations between subdomains and treatment outcomes or to recruit participants based on their level of frailty [22].

In order to facilitate the use of GA in oncology research, innovative strategies are therefore needed [23]; these include the standardised collection of geriatric data in clinical trials. A national and international Delphi consensus by a multidisciplinary expert board in geriatric oncology led to the development of the Geriatric Core Dataset (G-CODE). This tool standardises GA for research purposes by offering a brief but comprehensive assessment of the health of older cancer patients [24]. To validate and reinforce the G-CODE’s use in clinical trials (particularly as an aid for stratification or treatment allocation as a function of the level of frailty), it is necessary to develop and validate a prognostic score derived from the G-CODE.

The objectives of the present study were therefore to (i) develop a scoring system based on the G-CODE tool, (ii) use the score to estimate the risk of death at 6 months in older patients with cancer, and (iii) externally validate the score at 6, 12 and 24 months.

Methods

Study populations

This study used data from two prospective cohorts in France. The multicentre Elderly Cancer Patients (ELCAPA) cohort from the Greater Paris urban area served as the training set for the development of the G-CODE score. The Geriatric Oncology Marseille (GOM) cohort from Marseille University Hospital (Marseille, France (Geriatric Oncology assessment in assistance-publique des hôpitaux de Marseille)) served as the independent external validation set. Both cohorts included patients aged 70 or over with a solid cancer or haematological malignancy and having been referred for GA at a geriatric oncology clinic prior to first-line treatment or a change in treatment. All patients gave their written informed consent prior to inclusion.

The study protocols were approved by the appropriate independent ethics committees (ELCAPA: CPP Ile-de-France I, Paris, France; GOM: CE d’Aix-Marseille Université: Marseille, France; reference: 2024-06-18-02). ELCAPA is registered at ClinicalTrials.gov (identifier: NCT02884375). The study report complies with the STROBE guidelines for observational studies and the TRIPOD statement for the development and validation of multivariable prediction models.

For the present study, we included patients with available follow-up data recruited between January 2015 and December 2021 (for the ELCAPA training set) or between January 2018 and December 2019 (for the GOM validation set).

Data collection

The collected data included demographic and clinical characteristics, and the results of the GA performed by a geriatrician specialised in oncology. Cancer-related variables, including the type of cancer, metastatic status, and treatment decision modalities (surgery, chemotherapy, targeted therapy, radiotherapy, hormone therapy, immunotherapy, and supportive care) were also recorded.

The G-CODE

The G-CODE tool [24] assesses seven geriatric domains via the following 10 items: (i) living alone (yes/no) and (ii) the presence of support at home or a primary caregiver (yes/no) for the social assessment; (iii) activities of daily living (ADL, abnormal if <6/6) and (iv) short instrumental activities of daily living (4-IADL, abnormal if <4/4) for the functional assessment; (v) the “timed up and go” (TUG) test (abnormal if >20 s) for the mobility assessment; (vi) weight loss over the previous 6 months (>10%) and (vii) the body mass index (BMI, <21 kg/m2) for the nutritional assessment; (viii) the Mini-Cog test (abnormal if <4/5) for the cognition assessment; (ix) the mini-Geriatric Depression Scale (mini-GDS, abnormal if ≥1/4) for the mood assessment; and (x) the updated Charlson Comorbidity Index (UCCI) for the comorbidity assessment.

Since the development of the G-CODE (2015–2018), updated guidelines on the BMI have set a new threshold of <22 (rather than <21) as a criterion for malnutrition in patients aged 70 or older. The revised cut-off was applied in the present study.

Regarding the UCCI, not all comorbidities needed to calculate the score were available in the training set. Consequently, we used other appropriate variables as proxies (for details, see Supplementary Material S1).

Endpoints

The primary endpoint was 6-month mortality, and the secondary endpoints were 12- and 24-month mortality; these were defined as the time interval between the GA and death within 6, 12 and 24 months or (for censored patients) last follow-up. We chose 6 months as the main endpoint because predicting short-term mortality is necessary to avoid over- or under-treatment. Vital status was identified in medical charts, at the public records office, and via the French national vital records register.

Statistical analyses

Summary statistics were used to describe patient characteristics and G-CODE items in each cohort. In a univariate analysis, we estimated the crude hazard ratio (HR) and its 95% confidence interval for the associations between each G-CODE item (from the original G-CODE tool) and 6-month overall mortality in the training set. Since an established cut-off point is not currently available for the UCCI, we initially used the median. We next applied a recursive partitioning analysis (RPA) to determine optimal cut-points for predicting 6-month mortality and then categorised the score into classes of increasing risk (for the details, see the Supplementary Material S1).

Firstly, crude HRs for the 10 G-CODE items were estimated for 6-month overall mortality in the training set. These items were subsequently included in a multivariate analysis, and adjusted HRs were computed. To identify the optimal Cox proportional hazards regression model, we employed a stepwise feature selection process (both backward and forward) based on the Akaike information criterion (AIC). The regression coefficients were then used as weights to derive the final score. These weights were rescaled (multiplied) and rounded to the nearest integer in accordance with Cole’s algorithm to ensure both ease of clinical application and accuracy, and were used to construct a continuous score in the training and validation sets. Crude HRs [95% CI] for the continuous G-CODE score were estimated for 6-month mortality in both sets. The score was then adjusted in a baseline model that included age, cancer type, metastatic status and anti-cancer treatment. For the multivariable analyses, age was included as a continuous variable, metastatic status as a binary variable (yes/no), cancer sites were grouped into eight categories: colorectal; upper digestive tract, liver and pancreas; prostate; urinary tract and bladder; breast; lung; haematological malignancies; and other cancer sites, and each treatment modality was included as separate binary variables (yes/no). Adjusted HRs were estimated for both sets. In an RPA, we categorised the continuous score into distinct risk classes.

The final G-CODE score was validated in an independent cohort. The model’s discriminant ability was assessed in both the training and validation cohorts by considering the time-dependent area under the receiver operator curve (tAUC), which extends the standard AUC to survival analysis. Calibration was assessed graphically using calibration plots of observed versus predicted mortality and quantitatively by estimating the calibration-in-the-large and calibration slope.

The proportion of missing data was ≤10% for all variables with the exception of the UCCI in the training set (38.5%). For the other variables, the proportion ranged from 1% to 10% (TUG) in the training set and from 0% to 3% (Mini-Cog) in the validation set. We used a random forest imputation algorithm to impute missing values (for details, see the Supplementary Material S1). In a sensitivity analysis, we performed a complete-case analysis including only patients with complete data for all variables included in the prediction model. The model was evaluated using the same performance metrics as in the primary analysis, including discrimination (tAUC) and calibration.

All analyses were performed using Stata software (version 16, StataCorp LLC, College Station, TX, USA), except for missing data imputation, which was performed using R software (version 4.0.2, R Foundation for Statistical Computing, Vienna, Austria). All tests were two-sided, and the threshold for statistical significance was set to p < 0.05.

Results

Study populations

In total, 950 of the 1061 patients included in the ELCAPA training set, and 766 of the 772 patients included in the GOM validation set, had complete follow-up data (Supplementary Fig. 1). Overall, the patients in the validation set were younger (median age: 81 vs. 83 in the training set) and were more likely to have localised cancer (61% vs. 46%, respectively), lung cancer (23% vs. 5.5%, respectively) or head and neck cancer (11% vs. 1.3%, respectively) (Table 1). Concerning G-CODE items, the validation set included higher proportions of patients with abnormal 4-IADL, TUG and mini-GDS scores, a lower UCCI (median: 3 vs. 7 in the training set) and lower proportions of patients living alone (12% vs. 43%, respectively) or having a caregiver (73% vs. 91%, respectively).

Table 1 Characteristics of the patients in the training set (the ELCAPA cohort) and the validation set (the GOM cohort).

Full size table

Survival analyses

The median (min-max) follow-up time after the GA was 62.8 months (0.1–101.9) and 58.03 months (0.1–73.8) in the training and validation sets, respectively. The 6-, 12-, and 24-month mortality risks were respectively 0.27, 0.43 and 0.56 in the training set and 0.22, 0.34 and 0.51 in the validation set.

Univariate and multivariate Cox proportional hazards analyses

The item living alone was not significantly associated with 6-month mortality in univariate and multivariate analyses (Table 2). The ADL and mini-GDS scores were significantly associated with 6-month mortality in univariate analyses but were not significant in the multivariable analysis (Table 2). An RPA of the continuous UCCI score yielded three distinct classes: <8, 8–9, and ≥10. The final weights were calculated by rescaling the regression coefficients by a factor of 4.5, which yielded the highest accuracy (Table 3). When considering the final G-CODE score, the variable living alone was not included in the final score (Supplementary Table 1), following the AIC selection process. Thus, 9 items of the original G-CODE compose the final score with a range between 0 and 17 points.

Table 2 Univariate and multivariate analyses of the G-CODE items in the training set (the ELCAPA cohort; n = 950).

Full size table

Table 3 The final multivariate Cox proportional hazards model and weights for predicting 6-month mortality (training set; n = 950).

Full size table

The G-CODE score was significantly associated with 6-month mortality in a model including age, cancer type, metastatic status and treatment decision (Fig. 1). Crude and adjusted HRs of the continuous G-CODE score were very similar in the training and validation sets (Fig. 1). An RPA of the continuous G-CODE score yielded three risk groups: low (<7 points), intermediate (7–8 points), and high (>8 points). We observed a clear, graded relationship between higher risk categories and worse survival at 6, 12 and 24 months (log-rank p < 0.0001 for all time points) in both the training and validation sets (Fig. 2).

Fig. 1: Crude and adjusted hazard ratios with the continuous G-CODE score for 6-month overall mortality in the training set (n = 950) and the validation set (n = 766).

The baseline model included age, cancer type, metastatic status, and treatment.

Fig. 2: Kaplan–Meier survival curves by risk category (according to the G-CODE score) in the training and validation sets.

Hazard ratios (HRs) with 95% confidence intervals are shown relative to the reference risk group. P-values correspond to Wald tests for the hazard ratios estimated from the Cox proportional hazards model.

Discriminant ability of the G-CODE score in the training and validation sets

The tAUC in the external validation set ranged from 68.1% (for 24-month mortality) to 71.2% (for 6-month mortality). When adjusted for the baseline model including age, cancer type, metastatic status and treatment, the tAUCs ranged from 78.3% (for 12-month mortality) to 81.7% (for 6-month mortality), whereas the tAUCs for the baseline model ranged from 74.4% (for 12-month mortality) to 78.3% (for 6-month mortality) (Table 4). Inclusion of the G-CODE score in the baseline model increased tAUC by approximately 4% in both sets (Table 4). The corresponding hazard ratios and 95% confidence intervals for the baseline clinical model are provided in Supplementary Table 2.

Table 4 Discriminant ability of the G-CODE score in the training and validation sets, according to univariate and multivariate analyses.

Full size table

Calibration curves at various time points (3-, 6-, and 12-month follow-up) are shown in Supplementary Fig. 2. In the training set, calibration was excellent. In the external validation cohort, calibration remained good, with a calibration-in-the-large of 0.74 (95% CI 0.63–0.87) at 6 months and a calibration slope of 0.99 (95% CI 0.80–1.18). Calibration metrics were similar at 12 and 24 months.

Sensitivity analysis

In the complete-case analysis (n = 404), model discrimination remained good, with a tAUC ranging from 66.5% (at 24 months) to 69.8% (at 6 months) (Supplementary Table 3). After adjustment for the baseline model, the tAUCs were 79.7%, 76.6% and 79.6% at 6, 12 and 24 months, respectively. Calibration was satisfactory (Supplementary Fig. 3). These findings were consistent with those observed in the primary analysis based on the imputed dataset.

Discussion

The G-CODE tool was developed to help standardise the assessment of health impairments in older patients with cancer participating in clinical research. In the present study, we developed and externally validated a prognostic scoring system (out of 17 points) based on the G-CODE tool. We then used the score to predict 6-month mortality (tAUC: 71.2%) and highlight a graded risk across three categories (<7; 7–8; >8). We also demonstrated the G-CODE’s ability to predict mortality at 12- and 24 months (tAUC: 68.4% and 68.1%, respectively) and its overall good calibration properties.

External validity

The Geriatric Cancer Scoring System (GCSS) is the only prognostic score for 3-, 6-, and 12-month mortality; it was better able to predict 6- month (tAUC = 89%) and 12-month mortality (tAUC = 87%) in the external validation set than the G-CODE when the latter’s model included the tumour site and metastatic status (tAUC = 82% at 6 months and 78% at 12 months) [25]. This difference probably stemmed from the larger number of items (n = 18) and the types included in the GCSS: demographic variables (age and sex), oncological variables (tumour site, metastasis, and the Eastern Cooperative Oncology Group - Performance Status), laboratory variables (the C-reactive protein/albumin ratio and creatinine clearance), oncological treatment data, and other geriatric data alongside the G-CODE items (the G8 score, the Cumulative Illness Rating Scale – Geriatrics (CIRS-G) score, the number of medications, and the Mini Mental State Examination score). The GRADE prognostic score for 6-month mortality includes clinical and geriatric (gait speed) items but has not been externally validated. The GRADE development study showed good discriminant ability (Harrell’s C index [95% CI] = 0.75 [0.69–0.81] [26]).

Other prognostic scores have focused on 12-month mortality [27,28,29,30]. Only Bourdel-Marchasson et al. developed and validated a 12-month prognostic score that displayed much the same modest discriminant ability (AUC = 70%) as the G-CODE (tAUC: 68.4%) [28].

To the best of our knowledge, there are no prognostic scores for 24-month mortality; 50% of our included patients died after 24 months – then 2 years mortality is relevant in the older cancer population.

G-CODE items

The G-CODE score items that contributed most to the prediction of 6-month mortality were comorbidities (particularly when UCCI > 8; 3–5 points), impaired mobility (abnormal TUG test results; 2 points), loss of autonomy (abnormal IADL; 2 points), the absence of a caregiver (2 points), and weight loss over the previous 6 months (2 points).

The association between comorbidities and mortality has been documented in the literature [31]. In the other prognostic scores (such as the GCSS [25], METAGENUA [30] and Onco-MPI [27]), the comorbidities were assessed with the CIRS. None of the predictive models of 6-month mortality incorporates the UCCI. Impaired mobility is known to be a key geriatric factor for frailty and is consistently associated with poorer survival [32] and (in the GCSS) 6-month mortality (p < 0.0001). Weight loss - a key indicator of nutritional status - is included in several models [25, 27] and was confirmed in the present study to be a strong, significant prognostic factor. Loss of autonomy (an abnormal 4-IADL score) is strongly associated with cognitive impairment and predicts mortality in older patients receiving cancer therapy [32,33,34,35]. To the best of our knowledge, the G-CODE score is the only prognostic tool that has shown the relevance of the 4-IADL score as a predictor of 6-month mortality, although the 8-IADL tool has been identified as a predictor of 12-month mortality in GCSS and METAGENUA [25, 30]. The social item “not having a caregiver” is related to isolation. Although few studies have explored the importance of social support in predicting mortality, this factor is increasingly acknowledged to be crucial in the management of older patients with cancer. One study found that poor social support (assessed as living alone, lacking a primary caregiver and not having children) directly influenced 6- and 12-month mortality [8].

Clinical implications

The G-CODE tool was originally developed as a standardised dataset for geriatric oncology clinical research. It is administered to assess frailty before older patients start cancer treatment. The derived, externally validated G-CODE prognostic score facilitates stratification in clinical trials (e.g., for different cancer treatment strategies). Beyond its research uses, the G-CODE score might help clinicians who are unable to readily access GA to identify and monitor key impairments in the most important geriatric domains, inform patients and families, guide oncology decision-making, and prompt a comprehensive GA (CGA) if necessary, with the aim of offering appropriate geriatric interventions.

Strengths and limitations

This study had several strengths. The main strength was the external validation of the score using a large, independent cohort that differed from the development cohort with regard to the geographical region (northern France vs. southern France), the tumour site (mainly colorectal vs. mainly lung) and geriatric characteristics (e.g., multimorbidity and social status); this enhanced the generalisability and robustness of our findings. Unlike clinical trials, these cohorts included “real-life” older patients (mean age >80 years), many of whom were frail or vulnerable and nearly half of whom exhibited cognitive impairment and loss of autonomy in IADLs. Lastly, the G-CODE was based on a large national and international consensus formed by experts in geriatric oncology, which augments the score’s potential for implementation.

Our study had some limitations. Firstly, the patients were referred for GA because they had been identified as potentially frail by the G8 score or in an oncology assessment, which may have induced selection bias [36]. Further studies are needed to confirm the G-CODE score's prognostic value in the general older population. Secondly, the absence of some UCCI items and the use of surrogates might have induced classification bias. Indeed, because the original variables required to calculate the UCCI were unavailable in the development cohort, the index was reconstructed using proxy variables whenever possible, and remained missing for a proportion of patients. Although the validity of this proxy-based reconstruction could not be formally assessed, sensitivity analyses yielded similar model performance in the complete-case and primary analyses. Moreover, the model showed good discrimination and calibration in the external validation cohort, where the UCCI was available according to its original definition. Finally, G-CODE includes the most important geriatric domains, but this tool is not exhaustive in proposing all possible geriatric interventions.

Conclusion

We developed and externally validated a G-CODE score (out of 17) for predicting 6-, 12- and 24-month mortality in older patients with cancer. The key predictors for 6-month mortality included comorbidities, impaired mobility, loss of autonomy, weight loss and a poor social environment.

Data availability

Restrictions apply to the availability of these data. Data were obtained from the ELCAPA Study Group and are available from the corresponding author with the permission of the ELCAPA Study Group investigators.

References

  1. Pilleron S, Sarfati D, Janssen-Heijnen M, Vignat J, Ferlay J, Bray F, et al. Global cancer incidence in older adults, 2012 and 2035: a population-based study. Int J Cancer. 2019;144:49–58. https://doi.org/10.1002/ijc.31664.

    Article  PubMed  Google Scholar 

  2. Sedrak MS, Freedman RA, Cohen HJ, Muss HB, Jatoi A, Klepin HD, et al. Cancer and Aging Research Group (CARG). Older adult participation in cancer clinical trials: a systematic review of barriers and interventions. CA Cancer J Clin. 2021;71:78–92. https://doi.org/10.3322/caac.21638.

    Article  PubMed  Google Scholar 

  3. Tack L, Lefebvre T, Lycke M, Pottel L, Cool L, Ketelaars L, et al. Underrepresentation of vulnerable older patients with cancer in phase II and III oncology registration trials: a case-control study. J Geriatr Oncol. 2020;11:320–6. https://doi.org/10.1016/j.jgo.2019.09.003.

    Article  PubMed  Google Scholar 

  4. Le Saux O, Falandry C, Gan HK, You B, Freyer G, Péron J. Inclusion of elderly patients in oncology clinical trials. Ann Oncol. 2016;27:1799–804. https://doi.org/10.1093/annonc/mdw259.

    Article  PubMed  Google Scholar 

  5. Canouï-Poitrine F, Lièvre A, Dayde F, Lopez-Trabada-Ataz D, Baumgaertner I, Dubreuil O, et al. Inclusion of older patients with cancer in clinical trials: The SAGE Prospective Multicenter Cohort Survey. Oncologist. 2019;24:e1351–e1359. https://doi.org/10.1634/theoncologist.2019-0166.

    Article  PubMed  PubMed Central  Google Scholar 

  6. Paillaud E, Hamaker ME, Soubeyran P. Advances in geriatric oncology: exploring practical ways to optimize treatment in older patients with cancer. Cancers. 2022;14:4129 https://doi.org/10.3390/cancers14174129.

    Article  PubMed  PubMed Central  Google Scholar 

  7. Benderra MA, Serrano AG, Paillaud E, Tapia CM, Cudennec T, Chouaïd C, et al. Prognostic value of comorbidities in older patients with cancer: the ELCAPA cohort study. ESMO Open. 2023;8:101831. https://doi.org/10.1016/j.esmoop.2023.101831.

    Article  PubMed  PubMed Central  Google Scholar 

  8. Canouï-Poitrine F, Martinez-Tapia C, Paillaud E, Mathoulin-Pelissier S, Pamoukdjian F, Frasca M, et al. Geriatric impairments were directly and indirectly associated with mortality in older patients with cancer: a structural equation analysis. J Clin Epidemiol. 2022;148:17–26. https://doi.org/10.1016/j.jclinepi.2022.04.004.

    Article  PubMed  Google Scholar 

  9. Magnuson A, Sedrak MS, Gross CP, Tew WP, Klepin HD, Wildes TM, et al. Development and validation of a risk tool for predicting severe toxicity in older adults receiving chemotherapy for early-stage breast cancer. J Clin Oncol. 2021;39:608–18. https://doi.org/10.1200/JCO.20.02063.

    Article  PubMed  PubMed Central  Google Scholar 

  10. Gendarme S, Zebachi S, Corre R, Greillier L, Justeau G, Bylicki O, et al. Predictors of three-month mortality and severe chemotherapy-related adverse events in patients aged 70 years and older with metastatic non-small-cell lung cancer: a secondary analysis of ESOGIA-GFPC-GECP 08-02 study. J Geriatr Oncol. 2024;15:101506. https://doi.org/10.1016/j.jgo.2023.101506.

    Article  PubMed  Google Scholar 

  11. U.S. Food and Drug Administration. Inclusion of older adults in cancer clinical trials: guidance for industry. Draft guidance. www.fda.gov/media/135804/download. Accessed 1 Mar 2020.

  12. Pallis AG, Ring A, Fortpied C, Penninckx B, Van Nes MC, Wedding U, et al. European Organisation for Research and Treatment of Cancer Elderly Task Force. EORTC workshop on clinical trial methodology in older individuals with a diagnosis of solid tumors. Ann Oncol. 2011;22:1922–6. https://doi.org/10.1093/annonc/mdq687.

    Article  PubMed  Google Scholar 

  13. Hurria A, Cirrincione CT, Muss HB, Kornblith AB, Barry W, Artz AS, et al. Implementing a geriatric assessment in cooperative group clinical cancer trials: CALGB 360401. J Clin Oncol. 2011;29:1290–6. https://doi.org/10.1200/JCO.2010.30.6985.

    Article  PubMed  PubMed Central  Google Scholar 

  14. Caillet P, Laurent M, Bastuji-Garin S, Liuu E, Culine S, Lagrange JL, et al. Optimal management of elderly cancer patients: usefulness of the Comprehensive Geriatric Assessment. Clin Interv Aging. 2014;9:1645–60. https://doi.org/10.2147/CIA.S57849.

    Article  PubMed  PubMed Central  Google Scholar 

  15. Laurent M, Paillaud E, Tournigand C, Caillet P, Le Thuaut A, Lagrange JL, et al. ELCAPA Study Group. Assessment of solid cancer treatment feasibility in older patients: a prospective cohort study. Oncologist. 2014;19:275–82.https://doi.org/10.1634/theoncologist.2013-0351.

  16. Extermann M, Chen H, Cantor AB, Corcoran MB, Meyer J, Grendys E, et al. Predictors of tolerance to chemotherapy in older cancer patients: a prospective pilot study. Eur J Cancer. 2002;38:1466–73. https://doi.org/10.1016/s0959-8049(02)00090-4.

    Article  PubMed  Google Scholar 

  17. Boulahssass R, Gonfrier S, Ferrero JM, Sanchez M, Mari V, Moranne O, et al. Predicting early death in older adults with cancer. Eur J Cancer. 2018;100:65–74. https://doi.org/10.1016/j.ejca.2018.04.013.

    Article  PubMed  Google Scholar 

  18. Biesma B, Wymenga ANM, Vincent A, Dalesio O, Smit HJM, Stigt JA, et al. Dutch Chest Physician Study Group. Quality of life, geriatric assessment and survival in elderly patients with non-small-cell lung cancer treated with carboplatin-gemcitabine or carboplatin-paclitaxel: NVALT-3 a phase III study. Ann Oncol. 2011;22:1520–7. https://doi.org/10.1093/annonc/mdq637.

    Article  PubMed  Google Scholar 

  19. Li D, Sun CL, Kim H, Soto-Perez-de-Celis E, Chung V, Koczywas M, et al. Geriatric Assessment-Driven Intervention (GAIN) on chemotherapy-related toxic effects in older adults with cancer: a randomized clinical trial. JAMA Oncol. 2021;7:e214158 https://doi.org/10.1001/jamaoncol.2021.4158.

    Article  PubMed  PubMed Central  Google Scholar 

  20. Mohile SG, Mohamed MR, Xu H, Culakova E, Loh KP, Magnuson A, et al. Evaluation of geriatric assessment and management on the toxic effects of cancer treatment (GAP70+): a cluster-randomised study. Lancet. 2021;398:1894–904. https://doi.org/10.1016/S0140-6736(21)01789-X.

    Article  PubMed  PubMed Central  Google Scholar 

  21. Le Saux O, Falandry C, Gan HK, You B, Freyer G, Péron J. Changes in the use of comprehensive geriatric assessment in clinical trials for older patients with cancer over time. Oncologist. 2019;24:1089–94. https://doi.org/10.1634/theoncologist.2018-0493.

    Article  PubMed  PubMed Central  Google Scholar 

  22. Lee W, Cheng SJ, Grant SJ, Marcum ZA, Devine B. Use of geriatric assessment in cancer clinical trials: a systematic review. J Geriatr Oncol. 2022;13:907–13. https://doi.org/10.1016/j.jgo.2022.04.014.

    Article  PubMed  PubMed Central  Google Scholar 

  23. Mir N, Testa GD, Lichtman S. Geriatricizing” drug development - a Young International Society of Geriatric Oncology letter to the editor. J Geriatr Oncol. 2023;14:101398. https://doi.org/10.1016/j.jgo.2022.10.013.

    Article  PubMed  Google Scholar 

  24. Paillaud E, Soubeyran P, Caillet P, Cudennec T, Brain E, Terret C, et al. G-CODE collaborators. Multidisciplinary development of the Geriatric Core Dataset for clinical research in older patients with cancer: a French initiative with international survey. Eur J Cancer. 2018;103:61–68. https://doi.org/10.1016/j.ejca.2018.07.137.

    Article  PubMed  Google Scholar 

  25. Audureau E, Soubeyran P, Martinez-Tapia C, Bellera C, Bastuji-Garin S, Boudou-Rouquette P, et al. Machine learning to predict mortality in older patients with cancer: development and external validation of the geriatric cancer scoring system using two Large French Cohorts. J Clin Oncol. https://doi.org/10.1200/JCO.24.00117.

  26. Angeli E, Chouahnia K, Canoui-Poitrine F, Duchemann B, Aparicio T, Paillaud E, et al. Development, validation and clinical impact of a prediction model for 6-month mortality in older cancer patients: the GRADE. Aging. 2020;12:4230–46. https://doi.org/10.18632/aging.102876.

    Article  PubMed  PubMed Central  Google Scholar 

  27. Brunello A, Fontana A, Zafferri V, Panza F, Fiduccia P, Basso U, et al. Development of an oncological-multidimensional prognostic index (Onco-MPI) for mortality prediction in older cancer patients. J Cancer Res Clin Oncol. 2016;142:1069–77. https://doi.org/10.1007/s00432-015-2088-x.

    Article  PubMed  PubMed Central  Google Scholar 

  28. Bourdel-Marchasson I, Diallo A, Bellera C, Blanc-Bisson C, Durrieu J, Germain C, et al. One-year mortality in older patients with cancer: development and external validation of an MNA-based prognostic score. PLoS ONE. 2016;11:e0148523 https://doi.org/10.1371/journal.pone.0148523.

    Article  PubMed  PubMed Central  Google Scholar 

  29. Kanesvaran R, Li H, Koo KN, Poon D. Analysis of prognostic factors of comprehensive geriatric assessment and development of a clinical scoring system in elderly Asian patients with cancer. J Clin Oncol. 2011;29:3620–7. https://doi.org/10.1200/JCO.2010.32.0796.

    Article  PubMed  Google Scholar 

  30. Giannotti C, Zoppoli G, Ferrando L, Murialdo R, Caffa I, Laudisio A, et al. Development of a predictor of one-year mortality in older patients with cancer by geriatric and oncologic parameters. J Geriatr Oncol. 2020;11:610–6. https://doi.org/10.1016/j.jgo.2019.10.018.

    Article  PubMed  Google Scholar 

  31. Tougeron D, Di Fiore F, Thureau S, Berbera N, Iwanicki-Caron I, Hamidou H, et al. Safety and outcome of definitive chemoradiotherapy in elderly patients with oesophageal cancer. Br J Cancer. 2008;99:1586–92. https://doi.org/10.1038/sj.bjc.6604749.

    Article  PubMed  PubMed Central  Google Scholar 

  32. Soubeyran P, Fonck M, Blanc-Bisson C, Blanc JF, Ceccaldi J, Mertens C, et al. Predictors of early death risk in older patients treated with first-line chemotherapy for cancer. J Clin Oncol. 2012;30:1829–34. https://doi.org/10.1200/JCO.2011.35.7442.

    Article  PubMed  Google Scholar 

  33. Falandry C, Weber B, Savoye AM, Tinquaut F, Tredan O, Sevin E, et al. Development of a geriatric vulnerability score in elderly patients with advanced ovarian cancer treated with first-line carboplatin: a GINECO prospective trial. Ann Oncol. 2013;24:2808–13. https://doi.org/10.1093/annonc/mdt360.

    Article  PubMed  Google Scholar 

  34. Couderc AL, Suchon P, Saliba-Serre B, Rey D, Nouguerede E, Arcani R, et al. Functional status in older patients with cancer. J Geriatr Oncol. 2022;13:40–45. https://doi.org/10.1016/j.jgo.2021.07.004.

    Article  PubMed  Google Scholar 

  35. Conti C, Paillaud E, Laurent M, Poisson J, Boudou-Rouquette P, Frelaut M, et al. Prognostic impact of neurocognitive disorders in older patients with cancer: the ELCAPA prospective cohort study. J Nutr Health Aging. 2024;28:100215. https://doi.org/10.1016/j.jnha.2024.100215.

    Article  PubMed  PubMed Central  Google Scholar 

  36. Soubeyran P, Bellera C, Goyard J, Heitz D, Curé H, Rousselot H, et al. Screening for vulnerability in older cancer patients: the ONCODAGE Prospective Multicenter Cohort Study. PLoS ONE. 2014;9:e115060 https://doi.org/10.1371/journal.pone.0115060.

    Article  PubMed  PubMed Central  Google Scholar 

Download references

Acknowledgements

The authors thank David Fraser (Biotech Communication SARL, Ploudalmezeau, France) for editorial assistance; The ELCAPA Study Group comprises geriatricians (Amelie Aregui, Geoffroy Beraud-Chaulet, Stéphanie Benyahia, Mikaël Bringuier, Catherine Conti, Philippe Caillet, Pascale Codis, Lola Corsin, Tristan Cudennec, Solène Doublet, Virginie Fossey-Diaz, Beatrice Gonzalez, Charlotte Goldstein, Aissata Konate, Marie Laurent, Julie Lavolle, Emmanuelle Lorisson, Madeleine Lefevre, Julien Le Guen, Soraya Merbaki, Celine Lazarovici-Nagera, Elena Paillaud, Monica Pierro, Johanne Poisson, Frédéric Pamoukdjian, Anne-Laure Scain, Godelieve Rochette de Lempdes, Pauline Rabier-Lebrun, Raphaelle Roth, Florence Rollot-Trad, Zoe ap Thoma, Selim Turk, Hélène Vincent, Johanna Vouriot), oncologists (Pascaline Boudou-Rouquette, Marc-Antoine Benderra, Stéphane Culine, Etienne Brain, Maxime Frelaut, Alexandre de Moura, Aurelien Noret, Romain Geiss, Djamel Ghebriou, Christophe Tournigand), a digestive oncologist (Thomas Aparicio), a radiation oncologist (Jean-Léon Lagrange), epidemiologists (Etienne Audureau, Florence Canouï-Poitrine, Claudia Martinez-Tapia), a pharmacist (Pierre-André Natella), clinical nurses (Florence Devillard, Sadia Bonhomme, Anne Philippe), data manager (Clélia Chambraud, Khaled Bouaziz), and clinical research assistants (Aurélie Baudin, Johanna Canovas, Axelle Histe, Sandrine Lacour, Laure Morisset, Besma Saasaoui)

Funding

Supported by the French National Cancer Institute (Institut National du Cancer, INCa), Cancéropôle Ile-de-France, and Gérontopôle Ile-de-France (Gerond’If) for the ELCAPA study, none of which had any role in the design and conduct of the study, the collection, management, analysis, and interpretation of the data, the preparation, review and approval of the manuscript or the decision to submit the manuscript for publication.

Author information

Author notes

  1. These authors contributed equally: Anne-Laure Couderc, Claudia Martinez-Tapia.

  2. These authors contributed equally: Florence Canouï-Poitrine, Elena Paillaud.

Authors and Affiliations

  1. Department of Internal Medicine, Geriatrics and Therapeutic, Assistance-Publique – Hôpitaux de Marseille (AP-HM), Marseille, France

    Anne-Laure Couderc & Jeanne Bertona

  2. CNRS, EFS, ADES, Aix Marseille Univ, Marseille, France

    Anne-Laure Couderc

  3. Univ Paris Est Créteil, INSERM, IMRB, Créteil, France

    Claudia Martinez-Tapia, Marie Laurent, Philippe Caillet, Florence Canouï-Poitrine & Elena Paillaud

  4. Centre INSERM U1219 Bordeaux Population Health, Epicene Team, University of Bordeaux, Bordeaux, France

    Simone Mathoulin-Pelissier

  5. Clinical and Epidemiological Research Unit Institut Bergonié, Comprehensive Cancer Centre, Bordeaux, France

    Simone Mathoulin-Pelissier

  6. Assistance-Publique – Hôpitaux de Paris (AP-HP), Hôpital Henri-Mondor, Department of Geriatric Medicine, Creteil, France

    Marie Laurent

  7. Institut Curie, Department of Medical Oncology, Saint-Cloud, France

    Michael Bringuier

  8. IHU des Cancers des Femmes - ANR - 23-IAHU-0006, Paris, France

    Michael Bringuier

  9. Université Paris Cité, Department of Geriatrics, European Georges Pompidou Hospital, Paris-Cancer Institute CARPEM, AP-HP, Paris, France

    Pierre Gay & Johanna Canovas

  10. APHP, Hopital Europeen Georges Pompidou, Department of Geriatrics, Geriatric Oncology Unit, Paris, France

    Philippe Caillet & Elena Paillaud

  11. AP-HM, INSERM, CNRS, CRCM, Hôpital Nord, Multidisciplinary Oncology and Therapeutic Innovations, Aix Marseille Université, Marseille, France

    Laurent Greillier

  12. Institut Bergonié, Department of Oncology, Université de Bordeaux, Inserm, Bordeaux, France

    Pierre Soubeyran

  13. Oslo University Hospital, Department of Geriatric Medicine, Oslo, Norway

    Siri Rostoft

  14. AP-HP, Hôpital Henri-Mondor, Department of Public Health, Creteil, France

    Florence Canouï-Poitrine

Authors

  1. Anne-Laure Couderc
  2. Claudia Martinez-Tapia
  3. Simone Mathoulin-Pelissier
  4. Marie Laurent
  5. Michael Bringuier
  6. Pierre Gay
  7. Philippe Caillet
  8. Jeanne Bertona
  9. Laurent Greillier
  10. Johanna Canovas
  11. Pierre Soubeyran
  12. Siri Rostoft
  13. Florence Canouï-Poitrine
  14. Elena Paillaud

Contributions

A-LC, SM-P, PS, SR, FC-P and EP contributed to the conceptualisation of the study. CM-T and FC-P developed the methodology. A-LC, LG, PG, F-CP and EP performed the investigation. CM-T conducted the formal analysis. A-LC, ML, MB, and PC provided the study materials, participants, and data collection tools. JB and JC contributed to the collection and curation of the data. A-LC, CM-T, FC-P and EP wrote the original draft of the manuscript. All authors reviewed and edited the manuscript. FC-P and EP provided supervision. All authors read and approved the final version of the manuscript.

Corresponding author

Correspondence to Claudia Martinez-Tapia.

Ethics declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

The study was performed in accordance with the Declaration of Helsinki. All patients gave their written informed consent prior to inclusion. The study protocols were approved by the appropriate independent ethics committees (ELCAPA: CPP Ile-de-France I, Paris, France; GOM: CE d’Aix-Marseille Université: Marseille, France; reference: 2024-06-18-02)

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Couderc, AL., Martinez-Tapia, C., Mathoulin-Pelissier, S. et al. Using the Geriatric Core Dataset (G-CODE) to predict mortality in older patients with cancer: development and external validation. Br J Cancer (2026). https://doi.org/10.1038/s41416-026-03641-5

Download citation

  • Received: 16 October 2025

  • Revised: 08 September 2026

  • Accepted: 18 September 2026

  • Published: 12 October 2026

  • Version of record: 12 October 2026

  • DOI: https://doi.org/10.1038/s41416-026-03641-5

来源:Nature 健康护理主题 · nature.com