Artificial intelligence-powered real-time multimodal model for predicting recurrence and survival in head and neck cancer: a multicenter, multinational study

BACKGROUNDColorectal cancer (CRC) accounts for 9.4% of overall cancer deaths, ranking second after lung cancer. Despite the large number of factors tested to predict their outcome, most patients with similar variables show big differences in survival. Moreover, right-sided CRC (RCRC) and left-sided CRC (LCRC) patients exhibit large differences in outcome after surgical intervention as assessed by preoperative blood leukocyte status. We hypothesised that stronger indexes than circulating (blood) leukocyte ratios to predict RCRC and LCRC patient outcomes will result from combining both circulating and infiltrated (tumour/peritumour fixed tissues) concentrations of leukocytes.AIMTo seek variables involving leukocyte balances in peripheral blood and tumour tissues and to predict the outcome of CRC patients.METHODSSixty-five patients diagnosed with colon adenocarcinoma by the Digestive Surgery Service of the La Paz University Hospital (Madrid, Spain) were enrolled in this study: 43 with RCRC and 22 with LCRC. Patients were followed-up from January 2017 to March 2021 to record overall survival (OS) and recurrence-free survival (RFS) after surgical interventions. Leukocyte concentrations in peripheral blood were determined by routine laboratory protocols. Paraffin-fixed samples of tumour and peritumoural tissues were assessed for leukocyte concentrations by immunohistochemical detection of CD4, CD8, and CD14 marker expression. Ratios of leukocyte concentration in blood and tissues were calculated and evaluated for their predictor values for OS and RFS with Spearman correlations and Cox univariate and multivariate proportional hazards regression, followed by the calculation of the receiver-operating characteristic and area under the curve (AUC) and the determination of Youden’s optimal cutoff values for those variables that significantly correlated with either RCRC or LCRC patient outcomes. RCRC patients from the cohort were randomly assigned to modelling and validation sets, and clinician-friendly nomograms were developed to predict OS and RFS from the respective significant indexes. The accuracy of the model was evaluated using calibration and validation plots.RESULTSThe relationship of leukocyte ratios in blood and peritumour resulted in six robust predictors of worse OS in RCRC: CD8+ lymphocyte content in peritumour (CD8pt, AUC = 0.585, cutoff < 8.250, P = 0.0077); total lymphocyte content in peritumour (CD4CD8pt, AUC = 0.550, cutoff < 10.160, P = 0.0188); lymphocyte-to-monocyte ratio in peritumour (LMRpt, AUC = 0.807, cutoff < 3.185, P = 0.0028); CD8+ LMR in peritumour (CD8MRpt, AUC = 0.757, cutoff < 1.650, P = 0.0007); the ratio of blood LMR to LMR in peritumour (LMRb/LMRpt, AUC = 0.672, cutoff > 0.985, P = 0.0244); and the ratio of blood LMR to CD8+ LMR in peritumour (LMRb/CD8MRpt, AUC = 0.601, cutoff > 1.485, P = 0.0101). In addition, three robust predictors of worse RFS in RCRC were found: LMRpt (AUC = 0.737, cutoff < 3.185, P = 0.0046); LMRb/LMRpt (AUC = 0.678, cutoff > 0.985, P = 0.0155) and LMRb/CD8MRpt (AUC = 0.615, cutoff > 1.485, P = 0.0141). Furthermore, the ratio of blood LMR to CD4+ LMR in peritumour (LMRb/CD4MRpt, AUC = 0.786, cutoff > 10.570, P = 0.0416) was found to robustly predict poorer OS in LCRC patients. The nomograms showed moderate accuracy in predicting OS and RFS in RCRC patients, with concordance index of 0.600 and 0.605, respectively.CONCLUSIONEasily obtainable variables at preoperative consultation, defining the status of leukocyte balances between peripheral blood and peritumoural tissues, are robust predictors for OS and RFS of both RCRC and LCRC patients.

PurposeTo assess the impact of neighborhood-level disadvantage using the area disadvantage index (ADI) on survival outcomes in head and neck squamous cell carcinoma (HNSCC) patients.MethodsPatients diagnosed with previously untreated HNSCC from a single institutional study at a large, tertiary care hospital between 2008 and 2014 were provided self-administered questionnaires in a prospective longitudinal cohort study. Area Deprivation Index (ADI) was the primary exposure of interest, calculated using Federal Information Processing Standard (FIPS) codes that assess a neighborhood’s socioeconomic conditions, where a higher ADI indicates a disadvantaged neighborhood and lower socioeconomic status. The primary outcomes of interest were overall survival (OS), disease-specific survival (DSS), and recurrence-free survival (RFS). Survival outcomes were assessed using Kaplan–Meier analysis and Cox proportional hazard models.ResultsThe study included 792 patients. Patients with a higher ADI score were more likely to live in a less populous area (p < 0.01) and have a higher comorbidity score (p < 0.01), were heavy smokers (p < 0.01), and most cases (80.8%) tested negative for Human Papillomavirus (HPV) infection (p < 0.01). Higher terciles of ADI were associated with lower 5-year OS (p < 0.01), DSS (p = 0.01), and RFS (p = 0.03), with each 10-point increase in ADI being associated with a 1.1 times increase in hazard of death, disease-specific death, or recurrence (p < 0.01 for all). Patients in the highest tercile of ADI had significantly higher hazards of death (HR: 1.8 [1.3, 2.4], p < 0.01) and recurrence (1.4 [1.1, 1.9], p = 0.04) compared to the lowest tercile. In multivariable models, ADI was not significantly associated with OS, DSS, or RFS. Predictors of OS and DSS included HPV, stage, age, BMI, pack years, and comorbidity score, while RFS was predicted by HPV, stage, and comorbidity.ConclusionsHigher ADI scores were linked to poorer survival outcomes in HNSCC. These findings underscore the importance of considering social determinants of health, particularly ADI components like income, employment, housing quality, and access to care, in influencing HNSCC mortality and recurrence rates.Supplementary InformationThe online version contains supplementary material available at 10.1007/s10552-025-02067-3.

To determine the prognostic role of metastatic lymph node (LN) number and the minimal number of LNs for optimal staging of patients with pancreatic neuroendocrine tumors (pNETs). Prognosis relative to number of LN metastasis (LNM), and minimal number of LNs needed to evaluate for accurate staging, have been poorly defined for pNETs. Number of LNM and total number of LN evaluated (TNLE) were assessed relative to recurrence-free survival (RFS) and overall survival (OS) in a multi-institutional database. External validation was performed using Surveillance, Epidemiology and End Results (SEER) registry. Among 854 patients who underwent resection, 233 (27.3%) had at least 1 LNM. Patients with 1, 2, or 3 LNM had a comparable worse RFS versus patients with no nodal metastasis (5-year RFS, 1 LNM 65.6%, 2 LNM 68.2%, 3 LNM 63.2% vs 0 LNM 82.6%; all P < 0.001). In contrast, patients with ≥4 LNM (proposed N2) had a worse RFS versus patients who either had 1 to 3 LNM (proposed N1) or node-negative disease (5-year RFS, ≥4 LNM 43.5% vs 1-3 LNM 66.3%, 0 LNM 82.6%; all P < 0.05) [C-statistics area under the curve (AUC) 0.650]. TNLE ≥8 had the highest discriminatory power relative to RFS (AUC 0.713) and OS (AUC 0.726) among patients who had 1 to 3 LNM, and patients who had ≥4 LNM in the multi-institutional and SEER database (n = 2764). Regional lymphadenectomy of at least 8 lymph nodes was necessary to stage patients accurately. The proposed nodal staging of N0, N1, and N2 optimally staged patients.

6071 Background: Phase III data suggests a benefit of HDC in the adjuvant setting, but the effect of HDC and WC on long term survival and for HPV+ HNSCC is unknown. Methods: Data from a published retrospective study (Geiger Oral Onc 2013) of HDC vs WC in resected HNSCC was updated. Overall survival (OS) and recurrence-free survival (RFS) was assessed by Kaplan-Meier method for all pts and by HPV status. Multivariate analyses were performed to assess impact of HPV status, smoking, age, HDC vs WC, and cumulative cisplatin dose ( &lt; 200mg/m2 vs ≥200 mg/m2). Results: 51 patients (pts) received HDC and 53 WC. Median follow-up was 8.7 yrs (0.8-13.7). For the whole cohort, HDC had significantly improved OS over WC (p = 0.0095; 5- and 10-yr OS 84% and 80% vs 72% and 60%). No OS benefit for HDC was seen in pts with HPV+HNSCC (5- and 10-yr OS 90% and 87% for HDC and 81% and 81% for WC; p = 0.51). For HPV-negative HNSCC, OS had borderline significance with HDC vs WC (5- and 10-yr OS 73% and 68% vs 65% and 44%; p = 0.06). For the whole cohort, there was no difference in 5- and 10-yr RFS (78% and 74% for HDC vs 72% and 62% for WC; p = 0.32). When analyzed by HPV status, there was no difference in RFS with HDC or WC for either HPV+ (p = 0.43) or HPV-negative HNSCC (p = 0.97). On multivariate analyses of OS for all pts, only HPV status was significant (p = 0.0011; HR 0.27, CI 0.12-0.62). For HPV+ HNSCC, there was no significant predictor of OS. For non-HPV HNSCC, the benefit of HDC approached significance with a decreased risk of death (HR 0.38; p = 0.07). For all pts, those who received ≥200mg/m2 had significantly improved OS (5-yr 90% vs 72% and 10-yr 86% vs 61%; p = 0.004). By HPV status, cumulative dose had no significant effect on OS. Conclusions: OS is better with HDC and with cumulative dose &gt; 200 mg/m2 in unselected patients. The benefit of cisplatin is likely higher for non-HPV HNSCC. A difference in OS with no difference in RFS suggests non cancer-related causes of death in the WC cohort. Ability to receive HDC could be a surrogate marker of comorbidity. [Table: see text]

Head and neck squamous cell carcinoma (HNSCC) is a debilitating and deadly disease. The main risk factors are tobacco and alcohol use and human papillomavirus (HPV) infection. Early detection tests are needed because the majority of patients present in late stage when cure rates reach only 40%. Our group has developed a simple, inexpensive, noninvasive diagnostic test based on soluble CD44 (solCD44) and total protein levels. We used a case-control design to evaluate soluble markers for HNSCC in 150 oropharyngeal (OP) and lip/oral cavity (Lip/OC) patients and 150 controls frequency matched for age, gender, race, ethnicity, tobacco use and socioeconomic status. We compared patient groups with respect to important covariates using the chi-square, Fisher's exact test, or t-test. Markers’ mean levels were compared either by t-test or ANOVA, followed by pairwise multiple comparisons. Logistic regression analysis was used to evaluate predictivity of the salivary markers univariately and multivariately with adjustment for demographics and risk factors. We report odds ratio (OR) estimates with corresponding 95% confidence interval (95% CI) and area under the curve (AUC) of the operating characteristic curve (ROC) for fitted models. The case-control groups did not differ in regards to age, gender, race, ethnicity, history of ever vs. never smoking, current alcohol use, number of teeth removed, or county vs. private hospital system. The mean log2CD44 and protein levels were elevated in cases (log2 CD44= 1.94 ng/ml, protein=0.93 mg/ml) compared to controls (log2 CD44= 1.28 ng/ml, protein= 0.76 mg/ml), (p&amp;lt;.0001; p=.003, respectively). Log2CD44 levels were significantly elevated in older vs. younger cases (p&amp;lt;0.05). There were no significant mean log2 CD44 level differences between Lip/OC and OP cases, TNM or HPV status in cancer patients. In the cases for which HPV status was available (as measured by the surrogate marker p16), log 2 CD44 levels varied by smoking status (lower in never smokers), by drinking status in HPV + cases (lower in non-drinkers) and N-stage (higher levels in N0, Nx vs. N1-N3 in HPV-positive and the opposite effect in HPV-negative tumors). For protein, there were no differences in either the case or the control group based on demographic or risk variables. When we stratified by HPV status, race/ethnicity and drinking status did have an effect (higher levels in blacks vs. WNH and higher in current vs. former and never/mild drinkers in HPV positive tumors only). A specific model was developed for each of the 3 race/ethnicity groups with the highest AUC for blacks (AUC= .835) and WNH (AUC= .831) followed by HW (AUC= .777). Models included log2CD44, protein, smoking, gender and age; strongly suggesting that these factors play an important role for HNSCC early detection studies. Citation Format: Lutecia H. Mateus Pereira, Isildinha Reis, Robert Duncan, Judy Wen, Erika Reategui, Stephanie Bayers, Laurian Walters, Aymee Perez, Jennifer Hu, W Jarrard Goodwin, Elizabeth J. Franzmann. CD44, protein, demographics and risk factor data: A combined approach to detect head and neck squamous cell carcinoma. [abstract]. In: Proceedings of the 104th Annual Meeting of the American Association for Cancer Research; 2013 Apr 6-10; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2013;73(8 Suppl):Abstract nr 3548. doi:10.1158/1538-7445.AM2013-3548

BackgroundThe prognostic utility of the red cell distribution width-to-albumin ratio (RAR) in non-muscle-invasive bladder cancer (NMIBC) has not been established. This study aimed to evaluate the associations between preoperative RAR and both recurrence-free survival (RFS) and overall survival (OS) in patients with NMIBC.MethodsA retrospective review was performed for 240 individuals with NMIBC having undergone transurethral resection of bladder tumor (TURBT) at Shijiazhuang People’s Hospital from November 2013 to January 2024. Demographic characteristics, hematological parameters, pathological data, and information on instillation therapy were collected. The optimal RAR cutoff was identified by applying receiver operating characteristic (ROC) analysis. Survival curves were generated via the Kaplan–Meier method. The relationships of RAR with both RFS and OS were examined using univariate and multivariate Cox proportional hazards regression models. A nomogram was created using the identified independent prognostic factors from multivariate analysis to predict RFS. The discriminative ability and clinical usefulness of the nomogram were assessed by the concordance index (C-index), the calibration plots, time-dependent ROC analysis, and decision curve analysis (DCA).ResultsPatients with higher preoperative RAR showed significantly poorer RFS and OS. Multivariate analysis identified high RAR as an independent prognostic factor for both RFS (HR: 1.731, 95% CI: 1.012 - 2.959) and OS (HR: 3.425, 95% CI: 1.196 - 9.806) in NMIBC patients. Based on these findings, RAR was incorporated into a nomogram model for predicting RFS. Compared to the baseline model without RAR, the new model exhibited an improved C-index (from 0.704 to 0.728). The calibration plots demonstrated excellent consistency of the nomogram-predicted probabilities for 1-, 3-, and 5-year RFS with the actual survival rates. The time-dependent ROC analysis revealed that the areas under the curve (AUC) values for RFS predictions at 1-, 3-, and 5- years were 0.806, 0.797 and 0.806. DCA validated that the nomogram yielded a superior net benefit within threshold probability ranges of 10% to 45% when compared to traditional staging systems.ConclusionsThe findings suggest that preoperative RAR serves as a novel and independent prognostic factor for predicting RFS and OS in NMIBC cases.

Objective: This study aimed to analyze the prognostic risk factors for hepatoid adenocarcinoma of the stomach (HAS) and construct two nomogram-based clinical prediction models to predict overall survival (OS) and recurrence-free survival (RFS) in patients with HAS. Methods: Data were retrospectively collected from 82 patients (64 males, 18 females; mean age 60.3 ± 9.4 years) who underwent radical gastrectomy and were pathologically diagnosed with gastric hepatoid adenocarcinoma at the First Medical Center of the PLA General Hospital between February 2006 and September 2023. Statistical analyses were conducted using SPSS 25.0 and R 4.3.2. Survival analyses were performed using the Kaplan-Meier method, and univariate analyses were used to identify clinical and pathological factors associated with prognosis. Variables with P<0.05 in the univariate analysis were included in multivariate Cox regression models to identify independent risk factors for OS and RFS. These factors were incorporated into the prediction models to construct nomograms. The discriminatory power of the models was assessed using the area under the curve (AUC) of receiver operating characteristic (ROC) analyses, while calibration curves, decision curve analysis (DCA), and comparisons with the 8th edition of the TNM staging system of the American Joint Committee on Cancer (AJCC) were employed to evaluate model performance. Results: Among the 82 patients, 36 (43.9%) exhibited vascular infiltration, 61 (74.4%) had nerve infiltration, and lymph node metastasis was observed in 60 cases (73.2%). Pathological stages I, II, III, and IV were distributed as 11 (13.4%), 26 (31.7%), 44 (53.7%), and 1 (1.2%) cases, respectively. Inflammatory markers included neutrophil-to-lymphocyte ratio (NLR) ≥ 4.33 in 22 cases (26.8%), platelet-to-lymphocyte ratio (PLR) ≥ 142.2 in 50 cases (61.0%), monocyte-to-lymphocyte ratio (MLR) ≥ 0.411 in 22 cases (26.8%), α-fetoprotein (AFP) ≥ 2.48 µg/L in 64 cases (78.0%), and C-reactive protein (CRP) ≥ 7.506 mg/L in 12 cases (14.6%). Among the 82 patients, 3 cases (3.6%) were lost to follow-up. The median follow-up time was 52 (range: 8-147) months, with a median OS of 61(2-147) months. The 1-year and 3-year OS rates were 78.5% and 58.5%, respectively, while the 1-year and 3-year RFS rates were 77.3% and 60.3%, respectively. Multivariate analysis identified several independent risk factors influencing OS in patients with HAS: advanced pathological stage, MLR ≥ 0.411, AFP ≥ 2.545 µg/L, and CRP ≥ 7.51 mg/L. The hazard ratios (HRs) and 95% confidence intervals (CIs) were as follows: 5.218 (1.230-22.143), 2.610 (1.287-5.294), 2.950 (1.013-8.589), and 2.594 (1.145-5.877), respectively (all P < 0.05). For RFS, advanced pathological stage, PLR ≥ 152.0, and MLR ≥ 0.411 were independent risk factors, with HRs (95% CIs) of 4.735 (1.080-20.760), 3.759 (1.259-11.226), and 2.714 (1.218-6.048), respectively (all P < 0.05). The AUC values for OS prediction at 1 year, 3 years, and 5 years were 0.7765, 0.7525, and 0.7702, respectively. For RFS, the AUC values were 0.7304, 0.8137, and 0.8307 at 1 year, 3 years, and 5 years, respectively. The calibration curves demonstrated strong agreement between nomogram- predicted outcomes and observed survival data. DCA indicated that both TNM staging and the nomogram-based clinical prediction models provided a net positive benefit in predicting OS and RFS in HAS patients, with the nomogram model demonstrating superior performance. Conclusion: The nomogram-based clinical prediction models developed in this study demonstrated robust performance in predicting long-term OS and RFS in patients with HAS.

BackgroundMost prognostic signatures for colorectal cancer (CRC) are developed to predict overall survival (OS). Gene signatures predicting recurrence-free survival (RFS) are rarely reported, and postoperative recurrence results in a poor outcome. Thus, we aim to construct a robust, individualized gene signature that can predict both OS and RFS of CRC patients.MethodsPrognostic genes that were significantly associated with both OS and RFS in GSE39582 and TCGA cohorts were screened via univariate Cox regression analysis and Venn diagram. These genes were then submitted to least absolute shrinkage and selection operator (LASSO) regression analysis and followed by multivariate Cox regression analysis to obtain an optimal gene signature. Kaplan–Meier (K–M), calibration curves and receiver operating characteristic (ROC) curves were used to evaluate the predictive performance of this signature. A nomogram integrating prognostic factors was constructed to predict 1-, 3-, and 5-year survival probabilities. Function annotation and pathway enrichment analyses were used to elucidate the biological implications of this model.ResultsA total of 186 genes significantly associated with both OS and RFS were identified. Based on these genes, LASSO and multivariate Cox regression analyses determined an 8-gene signature that contained ATOH1, CACNB1, CEBPA, EPPHB2, HIST1H2BJ, INHBB, LYPD6, and ZBED3. Signature high-risk cases had worse OS in the GSE39582 training cohort (hazard ratio [HR] = 1.54, 95% confidence interval [CI] = 1.42 to 1.67) and the TCGA validation cohort (HR = 1.39, 95% CI = 1.24 to 1.56) and worse RFS in both cohorts (GSE39582: HR = 1.49, 95% CI = 1.35 to 1.64; TCGA: HR = 1.39, 95% CI = 1.25 to 1.56). The area under the curves (AUCs) of this model in the training and validation cohorts were all around 0.7, which were higher or no less than several previous models, suggesting that this signature could improve OS and RFS prediction of CRC patients. The risk score was related to multiple oncological pathways. CACNB1, HIST1H2BJ, and INHBB were significantly upregulated in CRC tissues.ConclusionA credible OS and RFS prediction signature with multi-cohort and cross-platform compatibility was constructed in CRC. This signature might facilitate personalized treatment and improve the survival of CRC patients.

To the Editor: Retroperitoneal liposarcoma (RLS) pertains to one of the rare malignant tumors originated from the retroperitoneum. It is evaluated as the most common type of retroperitoneal sarcoma. RLS can be divided into four subtypes according to the pathological classification, which includes well-differentiated liposarcoma (WDL), dedifferentiated liposarcoma (DDL), myxoid cell liposarcoma (MLS), and pleomorphic liposarcoma (PLS).[1] Giant RLS refers to the RLS with long diameter two times larger than the specified length of pathologic tumor stage-4 (pT4) category (≥30 cm) according to the eighth version of the American Joint Committee on Cancer (AJCC) Staging Manual.[2] Despite relatively low incidence, it brings great health burdens for patients and the standard therapeutic strategies remain to be explored. There are specific features of giant RLS according to clinical experience. For instance, giant RLS can occupy almost the entire abdominal cavity, close to or even wrapping around the inferior vena cava. Due to its location in the deep abdomen, patients usually have untypical symptoms at the early stage. The obvious signs and symptoms can be perceived only when the tumor becomes giant. These features increase difficulties in achieving complete resection during surgery and satisfactory patient outcomes. In addition, there is a lack of comprehensive research on the characteristics of giant RLS, which hinders the further progress of its diagnosis and targeted therapies. Identification of giant RLS characteristics may have great significance in clinical practice. Herein, we conducted a single-center retrospective study to reveal the clinicopathological features and outcomes of patients with giant RLS. This study was approved by the Protection of Human Subjects Committee of the Chinese People's Liberation Army (PLA) General Hospital (No. S2015-106-01). Informed written consent was provided by all participants included in this study. A total of 61 patients with giant RLS who received treatment at the First Medical Center, PLA General Hospital from January 2000 to December 2015 were enrolled in this study. Those with remote metastasis or perioperative death, or who received chemoradiotherapy were excluded from this study. The complete surgical margin referred to the resected margin without observable tumors. The complete resection was defined as the operation that macroscopically resected RLS without residual tumors. The time point of recurrence was when the recurrent region of RLS was detected. Overall survival (OS) referred to the time from surgical complete resection to the end of 5-year follow-up or death. Recurrence-free survival (RFS) was defined as the time from surgical complete resection to the onset of recurrence or death within 5 years.[3] Univariate analysis was performed for the selection of factors that were correlated with OS or RFS. The candidate factors were next included in the multivariate analysis to determine independent prognostic factors using Cox regression model. The categorical data were expressed as number (percentage); continuous variables were expressed as median (Q1–Q3) since they are not normally distributed. All data were analyzed using IBM SPSS Statistics for Windows, Version 25.0 (IBM Corp, Armonk, New York, USA). A two-sided P-value < 0.05 was considered statistically significant. In this retrospective study, there were 34 male (56%) and 27 female (44%) patients. The median age of them was 55 (46–63) years. The median size of tumors was 35 (31–40) cm [Supplementary Table 1, https://links.lww.com/CM9/B211], and the giant tumors could occupy the entire abdominal cavity [Supplementary Figure 1, https://links.lww.com/CM9/B211]. Considering the proportions and differentiation status of the four pathological subtypes of giant RLS, the patients were divided into two subgroups: WDL and non-WDL groups. WDL group accounted for the largest proportions among the four subtypes (34/61, 56%); the remaining 27 cases with DDL, MLS, and PLS subtypes were classified as the non-WDL group (44%). A total of 41 patients underwent organ resection (67%) and 17 cases were subjected to the resection of three and more organs (28%) [Supplementary Table 1, https://links.lww.com/CM9/B211]. The colon was the most frequently removed organ [Supplementary Figure 2, https://links.lww.com/CM9/B211]. Postoperative complications were evaluated according to the Clavien–Dindo classification.[4] There were 19 cases scored as Grade I (31%); while 42 cases underwent more serious complications were scored as Grades II–IV (69%) [Supplementary Table 1, https://links.lww.com/CM9/B211]. Subgroup analysis showed that 52% of cases in WDL group who received complete resection suffered from the Grade II–IV complications (13/25), while 81% of cases in non-WDL group who received complete resection had the Grade II–IV complications (13/16) [Supplementary Figure 3, https://links.lww.com/CM9/B211]. The median OS of all patients was 40 months during the 5-year follow up. The 1-, 3-, and 5-year OS rates were 83.8%, 51.7%, and 31.3%, respectively [Figure 1A]. To explore prognostic factors of 5-year OS for patients with giant RLS, univariate analysis was performed and indicated that resection times, resection method, intraoperative bleeding volume, total transfusion volume, postoperative complication, tumor morphology, tumor number, completeness of tumor capsule, pathological subtype, and status of surgical margin had significant correlations with 5-year OS (all P < 0.05) [Supplementary Table 1, https://links.lww.com/CM9/B211]. Variables with statistical significance were further examined using multivariate analysis. The results showed that postoperative complication and status of surgical margin were the prognostic factors of 5-year OS (both P < 0.05) [Supplementary Table 1, https://links.lww.com/CM9/B211].Figure 1: The 5-year OS and RFS analysis of patients with giant RLS. (A) The 5-year OS analysis of all 61 giant RLS patients included in this study. (B and C) The 5-year OS analysis of patients with WDL subtype (B) and non-WDL subtype (C) undergoing complete and incomplete resection. (D and E) The 5-year OS analysis of patients based on postoperative complication grades (D) and surgical margin status (E). (F) The 5-year RFS analysis of all 41 patients who received complete resection. (G and H) The 5-year RFS analysis of patients based on tumor capsule completeness (G) and resection times (H). OS: Overall survival; RFS: Recurrence-free survival; RLS: Retroperitoneal liposarcoma; WDL: Well-differentiated liposarcoma.The patients were next stratified to investigate the 5-year OS of subpopulations. Patients with WDL subtype of giant RLS who received complete resection had significantly better 5-year OS than those who received incomplete resection (P < 0.05) [Figure 1B]; whereas, for those with non-WDL subtypes, the difference between the two groups with complete or incomplete resections was not significant (P > 0.05) [Figure 1C]. In addition, we found that the patients who had milder postoperative complications or negative surgical margins obtained better 5-year OS (P < 0.05) [Figures 1D and 1E]. For the investigations into prognostic factors of RFS, data of 41 of the 61 patients, who received complete resection, were analyzed [Supplementary Table 2, https://links.lww.com/CM9/B211]. A total of 33 patients experienced local recurrence within 5 years. The median RFS was 15 (8–25) months. The 1-, 3-, and 5-year RFS rates of patients were 61.6%, 18.7%, and 13.4%, respectively [Figure 1F]. According to the univariate analysis, resection times, pathological subtype, and completeness of tumor capsule were correlated with 5-year RFS (all P < 0.05). After multivariate analysis, only the pathological subtype (P < 0.05) showed as an independent prognostic factor of 5-year RFS [Supplementary Table 2, https://links.lww.com/CM9/B211]. In addition, we found that the 5-year RFS of patients who had incomplete tumor capsules or multiple surgeries (resection times >3) was worse (both P < 0.05) [Figures 1G and 1H]. Few studies focused on the clinicopathological characteristics and survival outcomes of patients with giant RLS. Our study aimed to investigate the clinical features and risk factors of giant RLS, providing basic evidence for the development of giant RLS therapy. Compared with the survival outcomes of patients with non-giant RLS reported by other studies, those who suffered from giant RLS may have a worse prognosis.[5] Moreover, the processes of giant RLS resection are complicated, which may increase the risks of postoperative complications. In this study, we found that serious postoperative complication was significantly associated with the poor prognosis of patients with giant RLS. The importance of postoperative management of complications should be emphasized for patients with giant RLS. The role of pathological subtype in predicting the survival time of patients with ordinary RLS has been revealed in recent years. In this study, however, the pathological subtype was not proved to be an independent prognostic factor of 5-year OS. This contradiction may be attributable to the progression of giant RLS. The pathological subtype presented in this study was identified during the last surgery before follow-up. Well-differentiated giant RLS may progress to poorly-differentiated subtype, promoting tumor invasion, and attenuating survival benefits of patients. Additionally, the follow-up endpoint for RFS is tumor recurrence or a patient's death. RFS can reflect the effects of detected pathological subtype on survival time, free from potential influences of subsequent progression. Our findings proved that pathological subtype served as an independent factor of 5-year RFS, which further verifies our assumption that giant RLS may have a natural tendency to malignant progression. The completeness of surgical resection has been reported as an independent prognostic factor of patients with RLS. Supported by this evidence, wide application of complete RLS resection has been recommended. However, this opinion may be inappropriate for the treatment of giant RLS. In this study, giant tumors in many cases were found wrapping around abdominal organs and large vessels, which increased difficulties in achieving complete resection. Rigid conduction of complete resection may conversely elevate surgical risks and impair benefits from operations. Hence, a complete resection strategy for a giant RLS requires cautious consideration. Patients with WDL subtype of giant RLS could obtain significant survival benefits from complete resection when compared with those receiving incomplete resection. However, there were no significant differences in survival time between non-WDL patients who received complete and incomplete resections, and they were more likely to have severe postoperative complications. This suggests that pathological subtype may serve as an indicator for surgical decision-making. Radical resection may fit for patients with WDL subtype of giant RLS, while the cases with non-WDL subtype may need staging or palliative operations. Further clinical investigations should be conducted to prove the effectiveness of this strategy. It has to be admitted that there are several limitations. First, the study only included the cases from one center and the sample size is small due to the extremely low incidence of giant RLS. It impairs the representativeness of this research. Second, as a retrospective study, the range of collected information is limited and some potential biases cannot be eliminated. Third, several indicators of pathological examination, such as tumor necrosis and mitotic count, were not reported in some cases. It increases difficulties in more deeply identifying mechanisms underlying giant RLS progression. In conclusion, the status of surgical margins and postoperative complications were independent factors of OS, and the pathological subtype was proved as an independent marker for the recurrence of giant RLS. Complete resection for the WDL subtype of giant RLS contributed to prognosis improvement. However, survival benefits from complete resection were not obtained for the patients with non-WDL subtype of giant RLS. The pathological subtype may serve as an important reference indicator for the surgical decision-making process of giant RLS. Funding This study was supported by the National Natural Science Foundation of China (No. 82073192) and the National Key Research and Development Project of China (No. 2019YFB1311505). Conflicts of interest None.

Lymph node (LN) metastasis is significantly associated with worse prognosis for patients with intrahepatic cholangiocarcinoma (ICC). Improvement in preoperative assessment on LN metastasis helps in treatment decision-making. We aimed to investigate the role of radiomics-based method in predicting LN metastasis for patients with ICC. A total of 296 patients with ICC who underwent curative-intent hepatectomy and lymphadenectomy at two centers in China were analyzed. Radiomic features, including histogram- and wavelet-based features, shape and size features, and texture features were extracted from four-phase computerized tomography (CT) images. The clinical and conventional radiological variables which were independently associated with LN metastasis were also identified. A combined nomogram predicting LN metastasis was developed, and its performance was determined by discrimination, calibration, and stratification of long-term prognosis. The results were validated by the internal and external validation cohorts. Twenty-four radiomic features were selected into the nomogram. The established nomogram demonstrated good discrimination and calibration, with areas under the curve (AUCs) of 0.98 [95% confidence interval (CI) 0.96-0.99], 0.93 (0.88-0.98), and 0.89 (0.81-0.96) in the training and two validation cohorts, respectively. The 5-year overall survival (OS) and recurrence-free survival (RFS) rates of patients with high risk of LN metastasis as grouped by nomogram were poorer than those of patients with low risk in the training cohort (OS 28.8% versus 53.9%, p < 0.001; RFS 26.3% versus 44.2%, p = 0.001). Similar results were observed in the two validation cohorts. Radiomics-based method provided accurate prediction of LN metastasis and prognostic assessment for ICC patients, and might aid the preoperative surgical decision.

BackgroundProliferative hepatocellular carcinoma (PHCC) is an aggressive subtype of hepatocellular carcinoma (HCC) characterized by high recurrence rates and poor prognosis. Accurate preoperative identification of PHCC is essential for prognostic assessment and individualized treatment planning. However, conventional imaging methods often fail to reliably diagnose PHCC. This study aimed to develop an effective deep learning (DL)-based model using multiphasic gadoxetate disodium (Gd-EOB-DTPA)-enhanced magnetic resonance imaging (MRI) to preoperatively predict PHCC.MethodsThis retrospective multicenter study included 1,111 HCC patients who underwent Gd-EOB-DTPA-enhanced MRI followed by curative resection between January 2015 and June 2021. Patients were allocated to training (n=818), internal (n=150), and external (n=143) validation cohorts. A hybrid model combining DL and clinical-radiological features was developed using the AutoGluon framework. The clinical-radiological model was constructed using multivariate logistic regression, whereas the DL model was developed utilizing DL confidence scores derived from arterial and venous phase images. Model performance was assessed by the area under the curve (AUC) and 95% confidence intervals (CIs), and recurrence-free survival (RFS) was analyzed using the log-rank test.ResultsThe hybrid model demonstrated superior performance in both the training and internal validation cohorts, achieving AUCs of 0.922 (95% CI: 0.900–0.939) and 0.894 (95% CI: 0.834–0.934), respectively. In the external validation cohort, the hybrid and DL models demonstrated comparable performance (AUC: 0.771 vs. 0.788; P=0.065). Notably, the DL model outperformed the clinical-radiological model in predicting PHCC, with AUCs of 0.857 (95% CI: 0.781–0.912) vs. 0.747 (95% CI: 0.652–0.825) in the internal validation cohort and 0.788 (95% CI: 0.686–0.865) vs. 0.625 (95% CI: 0.505–0.735) in the external validation cohort (all P<0.05). Furthermore, patients classified as high-risk by the hybrid model had significantly shorter RFS compared to those in the low-risk group (P<0.05).ConclusionsThe hybrid model showed potential for predicting PHCC, which may assist clinicians in making personalized treatment decisions.

Background: This retrospective study was to evaluate the efficacy and toxicity of gemcitabine plus carboplatin (GC regimen) and paclitaxel plus carboplatin (PC regimen) combination chemotherapy administered as an adjuvant therapy after complete resection of non-small cell lung cancer. Methods: Forty-four patients (GC regimen, n = 29; PC regimen, n = 15) received gemcitabine at a dose of 1000 mg/m2 on days 1 and 8, and carboplatin with the target dose of area under the curve (AUC) of 4 on day 8 every 28 days and paclitaxel at a dose of 70 mg/m2 on days 1, 8 and 15, and carboplatin with the target dose of AUC of 5 on day 1 every 28 days. Results: A total of 130 cycles of the treatment were administered (averaged, 3.1 in GC arm and 2.7 cycles in PC arm). Forty-three patients (97.7%) completed the scheduled cycles. One patient (2.3%) was discontinued due to grade 4 pneumonia. The dose was reduced in 2 patients (4.5%) due to grade 4 thrombocytopenia. Grade 3/4 neutropenia was significantly observed in the PC group (GC: 12/29, 41.4%; PC: 11/15, 73.3%, p = 0.0443). The nonhematological toxicities were mild. Grade 1/2 alanine aminotransferase and aspartate aminotransferase in the GC group was significantly observed higher compared to those of the PC group (GC: 20/29, 69.0%; PC: 4/15, 26.7%, p = 0.0076). Grade 1/2 alopecia was significantly observed in the PC group (GC: 0/25, 0.0%; PC: 13/15, 86.7%, p 0.0001). There was no treatment-related death. The median survival time (MST) of the entire GC group was 784 days, the 3-year overall survival (OS) was 75.9%, and 3-year recurrence-free survival (RFS) was 65.5%. The MST of the entire PC group was 963 days, the 3-year OS was 80.0%, and the 3-year RFS was 60.0%. Conclusion: These results demonstrate that the GC and PC combination chemotherapies are efficacious and feasible regimens, which should be considered as one of the standard therapies for adjuvant therapy.

Background: p16 is a FDA approved surrogate biomarker for predicting human papillomavirus (HPV) associated oropharyngeal squamous cell carcinoma (OPSCC) with high sensitivity and moderate specificity. A p16 positive HPV negative phenotype indeed exists. Increased nuclear β-catenin expression has also been shown to correlate with HPV status in OPSCC. We, therefore, tested a novel model by combining β-catenin with p16 to increase the specificity for HPV prediction in OPSCC. Methods: Expression levels of β-catenin (nuclear and membrane immune-reactivity) and p16 were evaluated by immunohistochemistry (IHC) staining in 101 OPSCC tissues. HPV status was determined by HPV DNA in situ hybridization. Logistic regression models were used to estimate single or multiple biomarkers in HPV prediction. The prediction power, sensitivity, and specificity were determined by the Receiver Operating Characteristic (ROC) analysis. Results: All three biomarkers (p16, nuclear and membrane β-catenin) were significantly associated with HPV status (P &amp;lt; 0.01 for all). p16 alone showed the highest predictive power with area under the curve (AUC) of 0.9074 compared to 0.6762 for nuclear β-catenin and 0.7635 for membrane β-catenin. Multivariable analysis indicated that the odds ratios and 95% confidence interval (CI) for HPV positivity were 381.21 (17.14-8476.34), 4.93 (0.78-31.02), and 0.98 (0.97-1.00) for p16 (P &amp;lt; 0.001), nuclear β-catenin (P = 0.089) and membrane β-catenin (P = 0.085), respectively. The three-biomarker model had a sensitivity (99%) close to p16 alone (100%), but a higher specificity (89%) than p16 alone (81%) and showed prognosis value for overall survival (P = 0.0002) and disease-free survival (P = 0.0158). The model adjusting of clinical covariates only (AUC = 0.794) or adjusting of both clinical covariates and the three biomarkers (AUC = 0.852) did not produce a better HPV prediction as compared to the three-biomarker model (AUC = 0.962). Conclusion: Our findings suggests that the novel strategy of combining β-catenin with p16 increases the specificity for HPV prediction in OPSCC compared to p16 alone and deserves further validation in different clinical settings. (This study was supported by grants from Small Business Innovation Research (SBIR) Program (HHSN261201200097C) and National Institutes of Health (R33 CA161873) and Georgia Cancer Collation Distinguished Scholar Award to Dr. Zhuo Georgia Chen.) Citation Format: Guoqing Qian, Hong Xu, Zhongliang Hu, Sungjin Kim, Dongsheng Wang, Hongzheng Zhang, Zhengjia Chen, Susan Muller, Nabil Saba, Dong M. Shin, Andrew Wang, Zhuo Georgia Chen. Enhancing specificity in predication of human papillomavirus associated oropharyngeal squamous cell carcinoma by combining biomarkers p16 and β-catenin. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 1852. doi:10.1158/1538-7445.AM2014-1852

Prognostic Value of Baseline and Interim Total Metabolic Tumour Volume and Total Lesion Glycolysis Measured on 18f-FDG PET/CT in Patients with Follicular Lymphoma

Objectives To identify and validate baseline magnetic resonance imaging (b-MRI) radiomic features (RFs) as predictors of disease outcomes in effectively cured head and neck squamous cell carcinoma (HNSCC) patients. Materials and methods Training set (TS) and validation set (VS) were retrieved from preexisting datasets (HETeCo and BD2Decide trials, respectively). Only patients with both pre- and post-contrast enhancement T1 and T2-weighted b-MRI and at least 2 years of follow-up (FUP) were selected. The combination of the best extracted RFs was used to classify low risk (LR) vs. high risk (HR) of disease recurrence. Sensitivity, specificity, and area under the curve (AUC) of the radiomic model were computed on both TS and VS. Overall survival (OS) and 5-year disease-free survival (DFS) Kaplan–Meier (KM) curves were compared for LR vs. HR. The radiomic-based risk class was used in a multivariate Cox model, including well-established clinical prognostic factors (TNM, sub-site, human papillomavirus [HPV]). Results In total, 57 patients of TS and 137 of VS were included. Three RFs were selected for the signature. Sensitivity of recurrence risk classifier was 0.82 and 0.77, specificity 0.78 and 0.81, AUC 0.83 and 0.78 for TS and VS, respectively. VS KM curves for LR vs. HR groups significantly differed both for 5-year DFS (p<.0001) and OS (p=.0004). A combined model of RFs plus TNM improved prognostic performance as compared to TNM alone, both for VS 5-year DFS (C-index: 0.76 vs. 0.60) and OS (C-index: 0.74 vs. 0.64). Conclusions Radiomics of b-MRI can help to predict recurrence and survival outcomes in HNSCC.