Access the full text.
Sign up today, get DeepDyve free for 14 days.
IntroductionDespite recent advancements, cytotoxic chemotherapy is the leading systemic therapy for cancer [1, 2]. Chemotherapy in older individuals is usually complicated by frailty, medical comorbidities, frailty, poor functional status. According to the 2020 report of the International Agency for Research on Cancer (IARC), older adults (65 years and older) constituted 51.5 % of the total new cancer cases worldwide . Among 131,191 new cancer cases in 2020 in Iran, 66,251 (50.5 %) were older than 65 years old, and it will increase more in coming decades due to the demographic transition into population aging , , . Similar efficacy but more toxicity is the reason for hope and fear of chemotherapy in older patients , , . A study from Spain reported that 1.3 % of short-term mortalities were due to chemotherapy toxicities. Another study from the United Kingdom noted that chemotherapy-induced mortality was more common in older adults .Distinguishing patients who can tolerate standard doses from those who require less intensive therapy is of crucial importance. Over the decades, conventional functional status measures, such as the Eastern Cooperative Oncology Group performance status (ECOG) and the Karnofsky performance status (KPS), have been applied without significant evidence for their efficiency in the older population . Besides, ECOG and KPS measure the patient’s performance subjectively . Objective evaluation of performance status can further guide practitioners to select an appropriate treatment for patients, reducing toxicity and improving their quality of life . To address this requirement, Comprehensive Geriatric Assessment (CGA) was developed in the 1930s with the efforts of Marjory Warren . CGA is a multidisciplinary diagnostic and treatment process to identify frail older patients’ medical, functional, and psychosocial limitations to ensure appropriate treatment is provided . However, CGA is not available in all settings due to the need for coordination of multidisciplinary specialties, the time required for evaluation, and the lack of access to some disciplines (e.g., outpatient social work, pharmacy, and nutrition) in some practices . To facilitate CGA in routine oncology practice, several screening tools have been developed to assess the risk of severe toxicities using information from CGA, including the Chemotherapy Risk Assessment Scale for High-Age Patients (CRASH) and the Cancer and Aging Research Group (CARG) chemotoxicity calculators. Per the 2018 ASCO guideline for geriatric oncology, either the CARG or CRASH tools are recommended to obtain estimates of chemotherapy toxicity risk . However, CARG is more practical, using more data achievable during regular office visits .The CARG model was introduced by Hurria et al. in 2011 to predict chemotherapy toxicity in patients with cancer ≥65 years, considering both objective and subjective criteria, and developed a practical tool . This prediction tool comprises eleven questions in four domains: baseline characteristics, treatment, laboratory values, and geriatric assessment (Table 1). Cavdar et al. recently demonstrated that the CARG score strongly predicts chemotherapy toxicity in non-hematologic malignancies. In addition, it found that the predictive value of CARG is more than Geriatric 8 (G8) and Vulnerable Elders Survey (VES-13) (AUC-ROC 0.82 vs. 0.74 and 0.72) .Table 1:The cancer and aging research group study (CARG) prediction model for predicting chemotherapy toxicity in older patients.DomainVariableValue/ResponseScoreBaseline characteristicsAge, y≥72265–720Cancer typeGI or GU2Other0Planned treatmentPlanned CTx doseaStandard dose2Dose reduced upfront0Planned no. of CTx drugsPoly CTx2Mono CTx0Laboratory valuesHemoglobin<11 g/dL (male), <10 g/dL (female)3≥11 g/dL (male), ≥10 g/dL (female)0Creatinine clearanceb<34 mL/min3≥34 mL/min0Geriatric assessment questionsHow is your hearing (with a hearing aid, if needed)?Fair, poor, or totally deaf2Excellent or good0Number of falls in the past 6 months≥13None0Can you take your own medicine?With some help/unable1Without help0Does your health limit you in walking one block?Somewhat limited/limited a lot2Not limited at all0During the past four weeks, how much of the time has your physical health or emotional problems interfered with your social activities (like visiting with friends, relatives, etc.)?Limited some of the time, most of the time, or all of the time1Limited none of the time or a little of the time0CTx, chemotherapy; GI, gastrointestinal; GU, genitourinary; no., number; aBased on the national comprehensive cancer network (NCCN) guidelines. bUsing Cockcroft-Gault equation. Source: reference no. 11.Accumulating evidence has reflected the impact of racial differences on chemotherapy toxicity , , , . Exploring the ‘ethnic-specific genetic signatures’ can guide practitioners in selecting the appropriate chemotherapy regimen . Examining and validating the prediction models of chemotherapy toxicity in different races is required until then. The development and validation studies of the CARG model included patients of three primary races: White-American, Black-American, and Asian. Hitherto, the CARG model has been merely evaluated in the United States (main CARG study), India, and Japan [7, 25, 26]. The CARG and Indian cohorts noted that the CARG model could predict chemotherapy toxicity in older adults [7, 25]. However, the Japanese cohort found that the CARG model has low discrimination in predicting hematologic toxicity . Given the remarkable impact of race on chemotherapy toxicity, extrapolating the study results across different races is not an acceptable practice. Therefore, it requires external validation in different populations before general application. This prospective cohort was therefore designed to examine the CARG model in Iran as a representative of Middle Eastern and North African (MENA) countries  and compare its predictive value with the physician-reported KPS.Materials and methodsStudy design, setting, and participantsThis study aimed to examine the CARG prediction tool in Iranian patients with cancer referred to the Clinical Oncology Department of Imam Hossein Hospital (Tehran, Iran). To this end, we took steps similar to the methods of the development cohort conducted by Hurria et al. . Besides, we compared the CARG model with the physician-rated KPS to predict the chemotherapy toxicities. The study group was not limited to a single regimen or malignancy to challenge the model’s generalization. The eligibility criteria were: age ≥65 years, any solid tumor with any stage, and patients who were candidates for a new chemotherapy regimen.After meeting the inclusion criteria and consenting to participate, we asked the patients to answer the geriatric assessment questions as the development CARG study. Before the first chemotherapy cycle, baseline characteristics (age, sex, weight, height, education level, marital status, and cancer type and stage), laboratory data (leukocyte count, hemoglobin, platelet count, and creatinine), and patients’ KPS were recorded. Table 1 represents the CARG prediction model. It is composed of eleven criteria in four categories: (1) baseline characteristics (age and cancer type), (2) planned treatment (standard or adjusted chemotherapy dose and the number of chemotherapy agents), (3) laboratory values (hemoglobin and creatinine clearance), and (4) general geriatric assessment (hearing status, recent falls, self-sufficiency in taking medications, and health problem limiting walking or daily social activities). Based on the overall risk score, participants were grouped into three categories: low-risk (sum score 0–5), intermediate-risk (sum score 6–9), and high-risk (sum score 10–19). The treating team was blinded to the calculated risk score and category of the patients during the study.The Chemotherapy protocol was based on the physician’s discretion. The participants were carefully monitored for possible toxicities during the chemotherapy course before each cycle and were followed for two weeks after the chemotherapy ended. An experienced clinical oncologist (A.A.) evaluated the patients in each treatment course and between courses in case of existing toxicity and recorded grade 3 (hospitalization possibly indicated), grade 4 (life-threatening), and grade 5 (fatal), if any, adverse effects. Chemotherapy-related toxicities were recorded per the Common Terminology Criteria for Adverse Events (CTCAE) v5.0 . A decline in hematologic parameters was recorded as an adverse effect only if it was associated with patients’ symptoms or in the case of chemotherapy delay or dose modification. G-CSF (granulocyte colony-stimulating factor) as primary prophylaxis was started with chemotherapy at the physician’s discretion based on several factors, including the patient’s age, performance status, polychemotherapy regimen, and primary disease status. Patients who required primary prophylaxis received G-CSF at the same dosage during the chemotherapy cycle. G-CSF support during chemotherapy (as a secondary prophylaxis) was considered for those who developed severe/febrile neutropenia. The CARG scores and CTCAE grades were recorded independently by F.T. and A.A., respectively. The Institutional Review Board of Shahid Beheshti University of Medical Sciences (SBMU) approved the study protocol. This study was performed in line with the principles of the Declaration of Helsinki, and the ethical committee of SBMU approved this study (approval number: IR.SBMU.RETECH.REC.1398.106). The reporting of this prospective study follows the STROBE checklist for cohort studies (available at: https://www.strobe-statement.org/checklists/).Statistical methodsCategorical variables were summarized as numbers and percentages and were compared using the Chi-square test. Continuous variables were summarized using mean and standard deviation, and intergroup values were compared using the Mann-Whitney U test. The association of chemotherapy toxicities with CARG risk groups and KPS was evaluated using the Chi-square test. To this end, KPS scores were categorized into 70 and lower, 80, and 90–100 groups. The validity of the CARG model was evaluated by composing Receiver Operating Characteristic (ROC) curves and calculating its area under the curve (AUC). A similar analysis was done for KPS for comparison. All tests were two-sided, and the statistical significance was set to 0.05. We used IBM SPSS Statistics® (ver.26) for statistical analysis.ResultsParticipants and treatment characteristicsBetween November 2019 and May 2021, 207 patients with cancer were assigned to a new chemotherapy regimen at our center. Among 84 patients who fulfilled the eligibility criteria, 8 patients were missed for analysis: 6 missed the follow-up, and 2 did not consent to participate (Figure 1).Figure 1:STROBE flow diagram.This study examined 76 patients across 456 chemotherapy cycles. The study population had a mean age of 71.1 ± 5.9 years, of which 31 patients (40.8 %) were female. The three most common malignancies were head and neck cancers (24 cases, 31.6 %), gastrointestinal cancers (17 cases, 22.3 %), and breast cancer (10 cases, 13.1 %). Most patients were in the advanced stage (80.2 %), with stage IV disease in 46 patients (60.5 %). The baseline characteristics are detailed in Table 2. Most patients received first-line chemotherapy (62 patients, 78.9 %) with a polychemotherapy regimen (59 patients, 77.6 %) and reduced doses at the first cycle (39 patients, 51.3 %), according to physician discretion. Sixteen patients (21.0 %) required growth factor support during the chemotherapy course, which started in the initial cycle for 12 patients (75.0 %).Table 2:Baseline characteristics and treatment details of the study population.n, %aPatients with any toxicity n (%)a, (%)bp-ValuecPatients with hematologic toxicity n (%)a, (%)bp-ValuecPatients with nonhematologic toxicity n (%)a, (%)bp-ValuecCharacteristicsTotal76n=18 (23.6)n=14 (18.4)n=7 (9.2)Age at diagnosis, years0.750.400.41Mean ( ± SD)71.1 ( ± 5.9)71.3 ( ± 6.1)72.4 ( ± 7.1)70.8 ( ± 4.2)Range65–8865–8865–8865–88≤69d, n, %44 (57.9)10 (22.7), (55.5)6 (13.6), (42.8)5 (11.3), (71.4)>69, n, %32 (42.1)8 (25.0), (45.5)8 (25.0), (57.2)2 (6.2), (28.6)Sex, n, %0.850.660.90Female31 (40.8)7 (22.6), (38.9)5 (16.1), (35.7)3 (9.6), (42.9)Male45 (59.2)11 (24.4), (61.1)9 (20.0), (64.3)4 (8.9), (57.1)Educatione0.220.730.004Preliminary (ISCED 0–2)37 (75.5)7 (18.9), (58.3)6 (16.2), (66.7)1 (2.7), (20.0)Intermediate (ISCED 3–5)7 (14.3)3 (42.8), (25.0)2 (28.5), (22.2)3 (42.9), (60.0)Advanced (ISCED 6–8)5 (10.2)2 (40.0), (16.7)1 (20.0), (11.1)1 (20.0), (20.0)NR27652Marital status0.420.490.64Married74 (97.3)18 (24.3), (100)14 (18.9), (100)7 (9.4), (100)Single2 (2.7)000Living0.880.660.88With wife/husband43 (79.6)9 (20.9), (69.2)6 (13.9), (60.0)5 (11.6), (100)With children3 (5.5)1 (33.3), (7.7)1 (33.3), (10.0)0Alone8 (14.2)3 (37.5), (23.1)3 (37.5), (30.0)0NR22542Cancer type0.550.460.85Head and neck24 (31.6)6 (25.0), (33.3)5 (20.8), (35.7)3 (12.5), (42.9)GI17 (22.3)2 (11.7), (11.1)1 (5.9), (7.1)1 (5.9), (14.3)Breast10 (13.1)2 (20.0), (11.1)2 (20.0), (14.2)0Gyn7 (9.2)2 (28.5), (11.1)1 (14.3), (7.1)1 (14.3), (14.3)Lung6 (7.9)3 (50.0), (16.7)2 (33.3), (14.2)1 (16.7), (14.3)Other12 (15.8)3 (25.0), (16.7)2 (16.7), (14.2)1 (8.3), (14.3)Clinical stage, n, %0.370.210.48Stage I3 (3.9)1 (33.3), (5.5)1 (33.3), (7.1)0Stage II12 (15.8)3 (25.0), (16.7)3 (25.0), (21.4)0Stage III15 (19.7)1 (6.7), (5.5)01 (6.7), (14.3)Stage IV46 (60.5)13 (28.2), (72.2)10 (21.7), (71.5)6 (13.0), (85.7)TreatmentCTx regimen0.870.820.81TC32 (42.1)9 (28.1), (50.0)6 (18.7), (42.8)5 (15.6), (71.4)FOLFOX10 (13.1)2 (20.0), (11.1)1 (10.0), (7.1)1 (10.0), (14.3)AC6 (7.9)2 (33.3), (11.1)2 (33.3), (14.2)0GC4 (5.2)1 (25.0), (5.5)1 (25.0), (7.1)0FOLFIRI3 (3.9)000CF3 (3.9)1 (33.3), (5.5)1 (33.3), (7.1)0Capecitabine3 (3.9)000Paclitaxel2 (2.6)000Other15 (19.7)3 (20.0), (16.6)3 (20.0), (21.4)1 (6.7), (14.3)No. of CTx drugs0.500.920.13Polychemotherapy59 (77.6)15 (25.4), (83.3)11 (52.4), (78.6)7 (33.3), (100)Single agent17 (22.4)3 (17.6), (16.7)3 (5.4), (21.5)0CTx dose (1st cycle)0.130.090.63Reduced dose39 (51.3)12 (30.7), (66.7)10 (25.6), (71.4)3 (7.7), (42.9)Standard dose37 (48.7)6 (16.2), (33.3)4 (10.8), (28.6)4 (10.8), (57.1)CTx line0.360.650.18First line62 (78.9)16 (25.8), (88.9)12 (19.3), (85.8)7 (11.3), (100)>First line14 (21.1)2 (14.2), (11.1)2 (14.3), (14.2)0Growth factor use0.760.130.54No60 (79.0)13 (21.7), (72.2)9 (15.0), (64.3)6 (10.0), (85.7)Yes16 (21.0)5 (31.2), (27.8)5 (31,2), (35.7)1 (6.2), (14.3)AC, adriamycin plus cyclophosphamide; CF, cisplatin plus fluorouracil; CTx, chemotherapy; FOLFIRI, folinic acid, fluorouracil, plus irinotecan; FOLFOX, folinic acid, fluorouracil, plus oxaliplatin; GC, gemcitabine plus cisplatin; GI, gastrointestinal; Gyn, gynecologic; ISCED, international standard classification of education; NR, not reported; TC, weekly paclitaxel (taxol) plus carboplatin. a% of the relevant group (of reported data) (i.e. horizontal order). b% of the toxicity group (of reported data) (i.e. vertical order). cUsing Pearson Chi-Square test. dThe cutoff is the median value of total cohort. eAccording to international standard classification of education (ISCED) 2011. Note: toxicity data pertains to grades 3–4.In addition, Table 2 details the distribution of variables based on chemotherapy toxicities. Overall, 18 patients (23.6 %) experienced high-grade toxicity, which was hematologic toxicity in 11 patients (14.4 %), nonhematologic toxicity in 4 patients (5.2 %), and both toxicities in 3 patients (3.9 %). All patients with nonhematologic toxicity received first-line chemotherapy with a multidrug regimen, and 3 patients who had both hematologic and nonhematologic toxicities had stage IV disease. Hematologic toxicities were nonsignificantly more common in patients who had received AC and CF (cisplatin plus fluorouracil) regimens (33.3 %, p=0.82). For other variables, the association analysis demonstrated that neither of the evaluated variables was significantly associated with chemotherapy toxicity (p>0.05); except for patients with intermediate education grades (according to International Standard Classification of Education, level 3–5) who had a higher chance for nonhematologic toxicities (p=0.004).Chemotherapy toxicityTable 3 demonstrates the type and number of toxicities in the overall cohort. The most common hematologic toxicities were leukopenia (11 patients, 14.4 %) in the form of neutropenia (10 cases, 13.1 %) and anemia (6 patients, 7.9 %), respectively. Of 10 patients with neutropenia, 8 (80 %) were respondents to growth factor support. In this cohort, severe chemotherapy-induced thrombocytopenia (grade 3–5) was not detected. Severe nonhematologic toxicities were limited to neuropathy, oral mucositis, and diarrhea, and no grade 3–5 toxicities were detected for fatigue, nausea, dehydration, thrombosis, syncope, and electrolyte imbalance. Like hematologic toxicity, most nonhematologic toxicities were in grade 3 (85.7 %). All 4 patients who experienced severe (grade 3, 4) peripheral neuropathy have been treated with a TC regimen. Grade 3–4 peripheral neuropathy was detected in 4 out of 35 patients who had received the TC regimen (11.4 %, p=0.01). No peripheral neuropathy was detected in 3 patients who received docetaxel or nab-paclitaxel and 2 patients who received paclitaxel as monotherapy. Three patients with oral mucositis had rectal cancer receiving the FOLFOX regimen, metastatic laryngeal cancer receiving the TC regimen, and brain primitive neuroectodermal tumor (PNET) receiving Vincristine, Adriamycin, plus Cyclophosphamide, followed by Ifosfamide plus Etoposide (VAC-IE) regimen. Both patients who reported diarrhea during chemotherapy were treated with a TC regimen (p=0.09).Table 3:Toxicity distribution in the overall cohort.Grades 3–4 n, %aGrade 3 n (%)a, n (%)bGrade 4c n (%)a, n (%)bHematologic toxicityn=14n=9n=5TypeNeutropeniaLymphopenia10 (13.1)9 (11.8), (90.0)1 (1.3), (10.0)Anemia1 (1.3)1 (1.3), (100)0Thrombocytopenia6 (7.9)6 (7.9), (100)0000Number14 (5.2)4 (5.2), (100)027 (9.2)4 (5.2), (57.1)3 (42.9)33 (3.9)1 (1.3), (33.3)2 (66.7)Nonhematologic toxicityn = 7n = 6n = 1TypePeripheral neuropathy4 (5.2)3 (3.9), (75.0)1 (1.3), (25.0)Oral mucositis d3 (3.9)3 (3.9), (100)0Diarrhea e2 (2.6)1 (1.3), (50.0)1 (1.3), (50.0)Number13 (3.9)3 (3.9), (100)022 (2.6)1 (1.3), (50.0)1 (1.3), (50.0)3000a% of the total cohort, b% of the toxicity group (i.e., vertical order), cAt least one grade 4 toxicity, dNo patient received concurrent radiotherapy to head and neck region. eNo patient received concurrent radiotherapy to abdominopelvic region.The association of chemotherapy toxicities with the CARG score and KPSTable 4 presents the association of chemotherapy toxicities with CARG score and physician-rated KPS. Concerning the CARG model, there was no significant difference in the incidence of chemotherapy toxicities across the risk groups (p>0.05)._ Concerning the KPS score, the single patient with KPS 70 was excluded from the analysis. Collectively, patients with KPS 80 were more prone to chemotherapy toxicities compared to KPS 90–100 (75 % vs. 21 % respectively, p=0.04). There was no significant difference in the incidence of hematologic toxicity across the KPS risk groups (p=0.55); however, patients with KPS 80 had a significantly higher chance for nonhematologic toxicities compared to KPS 90–100 (50 % vs. 7 %, respectively, p=0.01). Four out of 5 patients who had KPS 90–100 and developed nonhematologic toxicities (80 %) received standard-dose chemotherapy in the first cycle. Both patients with KPS 80 developed grade 3 peripheral neuropathy even with a reduced-dose TC regimen in the first cycle.Table 4:Ability of CARG score vs. physician-rated KPS to predict chemotherapy toxicity.Overall cohort n=76 n, %aTotal toxicity n=18 n (%)a, (%)b, (%)cp-ValuedAUCeHematologic toxicity n=14 n (%)a, (%)b, (%)cp-ValuedAUCeNonhematologic toxicity n=7 n (%)a, (%)b, (%)cp-ValuedAUCeCARG score0.320.560.190.670.860.39Low (0–5)26 (34.2)4 (5.2), (15.4), (22.2)2 (2.6), (7.7), (14.2)3 (3.9), (11.5), (42.9)Intermediate (6–9)35 (46.1)11 (14.5), (31.4), (61.1)9 (11.8), (25.7), (64.3)3 (3.9), (8.6), (42.9)High (10–19)15 (19.7)3 (3.9), (20.0) (16.7)3 (3.9), (20.0), (21.4)1 (1.3), (6.7), (14.2)KPS0.040.560.220.550.010.6190–10071 (93.4)15 (19.7), (21.1), (83.3)12 (15.8), (16.9), (85.7)5 (6.5), (7.1), (71.4)804 (5.2)3 (3.9), (75.0), (16.7)2 (2.6), (50.0), (14.3)2 (2.6), (50.0), (28.6)701 (1.3)000AUC, area under curve; CARG, cancer aging research group; KPS, Karnofsky performance status. a% of the total cohort, b% of the CARG or KPS group (i.e., horizontal order), c% of the toxicity group (i.e., vertical order), dusing Pearson Chi-Square test, ethe area under the receiver operation characteristic (ROC) curve.The AUC-ROC of the two models were similar for total chemotherapy toxicities (CARG: 0.562, 95 % CI 0.40–0.69 vs. KPS: 0.565, 95 % CI 0.40–0.72). However, subanalysis revealed that AUC-ROC of the KPS model is larger for nonhematologic toxicity (KPS: 0.61 [95 % CI 0.37–0.86] vs. CARG: 0.39 [95 % CI 0.21–0.66]), and AUC-ROC of CARG model is larger for hematologic toxicities (CARG: 0.67 [95 % CI 0.48–0.78] vs. KPS: 0.55 [95 % CI 0.37–0.72]). At the best cutoff point equal to 7, the sensitivity and specificity of the CARG model for hematologic toxicity were 0.69 and 0.63, respectively (Table 4 and Figure 2). As noted earlier, four patients required G-CSF support during the chemotherapy course. Among these, no one was CARG-low risk (one high risk and three intermediate risk).Figure 2:ROC curves of CARG and KPS models for predicting total (A), hematologic (B), and nonhematologic (C) chemotherapy toxicities in older patients. Abbreviations: AUC, area under curve; CARG, cancer aging research group; KPS, Karnofsky performance status; ROC, receiver operating characteristic.DiscussionIn this cohort, the rate of chemotherapy-related toxicities was 23.6 %, more in the form of hematologic toxicity (18.4 %). The most common hematologic and nonhematologic toxicities were leukopenia (11.8 %) and peripheral neuropathy (5.2 %), respectively. Four patients who required G-CSF support amid the chemotherapy course were intermediate-high risk per the CARG model. The CARG model’s ability to predict hematologic toxicity was acceptable (AUC-ROC=0.67); however, it had poor discrimination in predicting total and nonhematologic toxicities (AUC-ROC=0.56 and 0.39, respectively) . Compared with the CARG model, physician-rated KPS has a similar value for predicting total chemotherapy toxicity (AUC-ROC=0.56), poor discrimination for hematologic toxicity (AUC-ROC=0.55 vs. 0.67), and acceptable discrimination for nonhematologic toxicity (AUC-ROC=0.61 vs. 0.39).Compared to the main CARG and the Indian validation studies, we recorded significantly lower rates of chemotherapy-related toxicity. The total grade 3–5 toxicity rate was 23 % vs. 53 % in the CARG development cohort, 58 % in the CARG validation cohort, and 52 % in the Indian cohort. We found no grade 5 toxicity; however, it was reported in 2 % and 4 % of patients in the main CARG and Indian studies, respectively. Dislike the main CARG study, the nonhematologic toxicities in this cohort were limited to diarrhea, neuropathy, and oral mucositis, and no grade 3–5 toxicities were recorded for fatigue, nausea, dehydration, thrombosis, syncope, and electrolytes imbalance. In the CARG, Indian, and Japanese cohorts, the risk of toxicity was increased with an increasing CARG risk score. However, the current study did not reflect this correlation in the evaluated Iranian patients. In contrast to the current study, in the development and validation CARG cohorts as well as the Indian cohort, CARG score had acceptable discrimination to predict total chemotherapy toxicity (AUC-ROC: 0.56 vs. 0.72, 0.65, and 0.63, respectively). On the other hand, the AUC-ROC of physician-rated KPS was similar and not satisfactory in predicting total toxicity in the current and the development CARG, validation CARG, and Indian cohorts (0.56 vs. 0.53, 0.54, and 0.52, respectively). In contrast to the current study, the Japanese cohort found that the CARG model was predictive for nonhematologic toxicity but not for hematologic toxicity .The discrepancy between the main CARG/Indian/Japanese cohorts and current study findings is likely multifactorial. First, participants in the current study had younger mean age compared with the CARG development and CARG validation studies). It has been established that chemotherapy-related toxicity increases with age . Second, the percentage of women in the current study was lower than in CARG development and validation studies (40 % vs. 56 % and 55 %). Strong evidence has noted that women are more susceptible to chemotherapy toxicities, including hematologic toxicities (e.g., anemia, leukopenia, and neutropenia) and nonhematologic toxicities (e.g., nausea and vomiting, constipation, stomatitis, and alopecia) , , , . Third, fewer patients in the present cohort received standard-dose chemotherapy at the initial cycle (48 % vs. 76 % in the main CARG and Indian validation studies and 70 % in the Japanese cohort). Fourth, the most applied chemotherapy regimen in the current study was weekly paclitaxel/carboplatin, which is a well-tolerated regimen . In addition, weekly patient visits for clinical examination and toxicity evaluation during the chemotherapy course can reduce the rate of chemotherapy toxicity. Although no subanalysis was done, this point might be considered as the possible contributing factor to fewer toxicities in this cohort. Fifth, participants in the CARG studies were mainly White (85 %); however, the current study evaluated chemotherapy toxicity in Iranian patients of a different race from the Middle East and North Africa region. Increasing evidence has put forward the patient’s race as a predictive factor of chemotherapy toxicity, which can be mediated by genetic and epigenetic factors , , , . A pooled analysis of clinical trials showed that cisplatin-based chemotherapy led to more neutropenia and anemia in Asian patients with lung cancer than non-Asian patients . Another study found similar results for patients with breast cancer receiving FEC (fluorouracil, epirubicin, plus cyclophosphamide) regimen . In a randomized trial on patients with colon cancer treated with 5-fluorouracil-based chemotherapy, Black patients experienced less nausea, vomiting, diarrhea, and oral mucositis than White patients . These studies were a few examples to illustrate the importance of evaluating the chemotherapy toxicity prediction tools in populations of different races before their general application.These differences might arise from the genetic polymorphisms between different races. For example, patients with lung cancer with MTHFR rs1801131 and MDM2 rs1690924 polymorphisms tend to experience more platinum-induced gastrointestinal toxicity. MTHFR rs1801133CT/TT carriers are prone to platinum-induced hematological toxicity . In patients with cervical cancer, ERCC1 rs3212986 polymorphism is associated with cisplatin-induced gastrointestinal toxicity . Therefore, genetic polymorphisms in different communities can be an independent factor in developing chemotherapy toxicity.Overall, CARG has several limitations. For example, it cannot precisely identify each risk factor for chemotherapy toxicity as a CGA can. Besides, CARG is merely a diagnostic tool. However, CGA is made to precisely identify the frailties of patients and, above all, to set up specific geriatric interventions in the face of the identified elements of fragility to correct the reversible factors, which promote toxicity, to limit their risk of toxicity, and thus to allow the treatment feasibility [18, 37]. The comparison between the current and previous studies reveal that CARG (and possibly other) chemotherapy prediction tool requires validation before its clinical application. Besides, this cohort had several learning points for the clinical practice of physicians in Iran. First, we realized that the CARG model for Iranian patients was predictive only for hematologic toxicity. Second, the cut-off point to distinguish low-risk patients for hematologic toxicities is 6 instead of 5 in the main study. Third, in this cohort, 39 patients received reduced-dose chemotherapy at the initial cycle based on the physician’s clinical judgment. In spite of a reduced dose, 30 % of patients experienced chemotherapy toxicity. This notion illustrates the importance of improving clinical judgment and developing a rationally-designed prediction tool for our clinical practice.These findings should be interpreted in light of the current study’s limitations: First, several interpreting factors were not analyzed in this study, including comorbidities, medications, nutritional status, and psychological state, all may affect the toxicities; second, this study did not evaluate the patients’ ethnicity.; third, the data of this study were per the real-world management of patients. Hence, there were cases who required growth factor support from the beginning due to intensive chemotherapy regimens (12 patients, 15.7 %) or during the course of chemotherapy due to the development of neutropenia (4 patients, 5.2 %). This inevitable issue can affect the study results. Despite these limitations, our study possesses several strengths: First, the prospective design. Second, separate analysis and report per hematologic and nonhematologic toxicities. Third, the clinical evaluation of patients by a single physician to improve the internal validity. In the main study, patients were reviewed for chemotherapy toxicities by two physicians, which can lead to inter-observer bias. This study was conducted to examine the CARG model in Iranian patients, not to validate it. We are aiming to continue this study to evaluate the validity of this model with more participants.ConclusionsThis prospective cohort was designed to examine the reproducibility of the CARG prediction tool in older Iranian patients with cancer. We found that the CARG score had an acceptable ability to predict hematologic toxicities; however, its efficacy for total and nonhematologic toxicities was limited. This issue reflects the importance of evaluating chemotherapy prediction tools in different populations and, if required, developing rationally designed, race-specific toxicity prediction models to improve toxicity prediction in patients with cancer. Future studies with larger sample sizes and patients from different races are invited to delineate this notion. Also, meta-analyses of available studies can be helpful to this end.
ONCOLOGIE – de Gruyter
Published: May 1, 2023
Keywords: cancer; chemotherapy; geriatric assessment; geriatric oncology
Access the full text.
Sign up today, get DeepDyve free for 14 days.