Search

A pervasive dilemma in neuroimaging is whether to prioritize sample size or scan time given fixed resources. Here, we systematically investigate this trade-off in the context of brain-wide association studies (BWAS) using functional magnetic resonance imaging (fMRI). We find that total scan duration (sample size x scan time per participant) robustly explains individual-level phenotypic prediction accuracy via a logarithmic model, suggesting that sample size and scan time are broadly interchangeable up to 20-30 min of data. However, the returns of scan time diminish relative to sample size, which we explain with principled theoretical derivations. When accounting for fixed overhead costs associated with each participant (e.g., recruitment, non-imaging measures), prediction accuracy in many small-scale and some large-scale BWAS might benefit from longer scan time than typically assumed. These results generalize across phenotypic domains, scanners, acquisition protocols, racial groups, mental disorders, age groups, as well as resting-state and task-state functional connectivity. Overall, our study emphasizes the importance of scan time, which is ignored in standard power calculations. Standard power calculations maximize sample size, at the expense of scan time, which can result in sub-optimal prediction accuracies and inefficient use of resources. Our empirically informed reference is available for future study design: WEB_APPLICATION_LINK.

Background: Major depressive disorder (MDD) is notably underdiagnosed and undertreated due to its complex nature and subjective diagnostic methods. Biomarker identification would help provide a clearer understanding of MDD aetiology. Although machine learning (ML) has been implemented in previous studies to study the alteration of microRNA (miRNA) levels in MDD cases, clinical translation has not been feasible due to the lack of interpretability (i.e. too many miRNAs for consideration) and stability., Methods: This study applied logistic regression (LR) model to the blood miRNA expression profile to differentiate patients with MDD (n = 60) from healthy controls (HCs, n = 60). Embedded (L1-regularised logistic regression) feature selector was utilised to extract clinically relevant miRNAs, and optimized for clinical application., Results: Patients with MDD could be differentiated from HCs with the area under the receiver operating characteristic curve (AUC) of 0.81 on testing data when all available miRNAs were considered (which served as a benchmark). Our LR model selected miRNAs up to 5 (known as LR-5 model) emerged as the best model because it achieved a moderate classification ability (AUC = 0.75), relatively high interpretability (feature number = 5) and stability (ϕ̂Z=0.55) compared to the benchmark. The top-ranking miRNAs identified by our model have demonstrated associations with MDD pathways involving cytokine signalling in the immune system, the reelin signalling pathway, programmed cell death and cellular responses to stress., Conclusion: The LR-5 model, which is optimised based on ML design factors, may lead to a robust and clinically usable MDD diagnostic tool., Competing Interests: Declaration of competing interest There are no conflict of interests., (Copyright © 2024 Elsevier B.V. All rights reserved.)

Background: Major depressive disorder (MDD) is a leading cause of disability worldwide. At present, however, there are no established biomarkers that have been validated for diagnosing and treating MDD. This study sought to assess the diagnostic and predictive potential of the differences in serum amino acid concentration levels between MDD patients and healthy controls (HCs), integrating them into interpretable machine learning models. Methods: In total, 70 MDD patients and 70 HCs matched in age, gender, and ethnicity were recruited for the study. Serum amino acid profiling was conducted by means of chromatography-mass spectrometry. A total of 21 metabolites were analysed, with 17 from a preset amino acid panel and the remaining 4 from a preset kynurenine panel. Logistic regression was applied to differentiate MDD patients from HCs. Results: The best-performing model utilised both feature selection and hyperparameter optimisation and yielded a moderate area under the receiver operating curve (AUC) classification value of 0.76 on the testing data. The top five metabolites identified as potential biomarkers for MDD were 3-hydroxy-kynurenine, valine, kynurenine, glutamic acid, and xanthurenic acid. Conclusions: Our study highlights the potential of using an interpretable machine learning analysis model based on amino acids to aid and increase the diagnostic accuracy of MDD in clinical practice.

Objectives: To characterize longitudinal changes in Epstein-Barr virus (EBV) DNA post-radiotherapy in nasopharyngeal carcinoma (NPC) patients, and investigate whether an early (0-2 weeks) or delayed (8-12 weeks) EBV DNA result better predicts for disease-free survival (DFS)., Materials and Methods: Histologically-confirmed NPC patients with ≥1 EBV DNA test quantified using the harmonized BamHI-W polymerase chain reaction-based assay at 0-2 and 8-12 weeks post-radiotherapy were included., Results: We identified 302 patients with EBV DNA measured at 0-2 weeks post-radiotherapy; of which, 110 (36.4 %) underwent a repeat test at 8-12 weeks post-treatment. Patients harboring a detectable EBV DNA at 0-2 weeks experienced an inferior DFS (adjusted HR 1-264 copies 1.72 [95 %CI: 1.05-2.83], P = 0.031; AHR ≥265 copies 4.39 [95 %CI: 1.68-11.44], P = 0.002 relative to 0 copies/mL). At 8-12 weeks, we observed substantial shifts in EBV DNA readings from 0 to 2 weeks; 76/110 (69.1 %) and 34/110 (30.9 %) patients at 0-2 weeks versus 90/110 (81.8 %) and 20/110 (18.2 %) at 8-12 weeks recorded undetectable and detectable EBV DNA, respectively. Positive EBV DNA at 8-12 weeks was strongly associated with relapse (73.3 % [11/15] for 1-264; 80.0 % [4/5] for ≥265 subgroups had relapses versus 15.6 % [14/90] for 0 copies/mL). Area under receiver operating curve values for 2-year relapse rates were 0.817 (95 %CI: 0.725-0.909) for stage + EBV DNA 8-12w versus 0.654 (95 %CI: 0.542-0.765) for stage + EBV DNA 0-2w. CONCLUSION: EBV DNA is dynamic post-radiotherapy, and delayed EBV DNA testing better enriched for higher-risk NPC patients. This implicates trials investigating adjuvant chemotherapy intensification based on early EBV DNA testing., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2023 The Author(s). Published by Elsevier Ltd.. All rights reserved.)

There is significant interest in using neuroimaging data to predict behavior. The predictive models are often interpreted by the computation of feature importance, which quantifies the predictive relevance of an imaging feature. Tian and Zalesky (2021) suggest that feature importance estimates exhibit low split-half reliability, as well as a trade-off between prediction accuracy and feature importance reliability across parcellation resolutions. However, it is unclear whether the trade-off between prediction accuracy and feature importance reliability is universal. Here, we demonstrate that, with a sufficient sample size, feature importance (operationalized as Haufe-transformed weights) can achieve fair to excellent split-half reliability. With a sample size of 2600 participants, Haufe-transformed weights achieve average intra-class correlation coefficients of 0.75, 0.57 and 0.53 for cognitive, personality and mental health measures respectively. Haufe-transformed weights are much more reliable than original regression weights and univariate FC-behavior correlations. Original regression weights are not reliable even with 2600 participants. Intriguingly, feature importance reliability is strongly positively correlated with prediction accuracy across phenotypes. Within a particular behavioral domain, there is no clear relationship between prediction performance and feature importance reliability across regression models. Furthermore, we show mathematically that feature importance reliability is necessary, but not sufficient, for low feature importance error. In the case of linear models, lower feature importance error is mathematically related to lower prediction error. Therefore, higher feature importance reliability might yield lower feature importance error and higher prediction accuracy. Finally, we discuss how our theoretical results relate with the reliability of imaging features and behavioral measures. Overall, the current study provides empirical and theoretical insights into the relationship between prediction accuracy and feature importance reliability., Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2023 The Author(s). Published by Elsevier Inc. All rights reserved.)

Purpose: The purpose of this study was to evaluate the radiotherapy planning feasibility of dose escalation with intensity-modulated proton therapy (IMPT) to hypoxic tumor regions identified on 18F-Fluoromisonidazole (FMISO) positron emission tomography and computed tomography (PET-CT) in NPC., Materials and Methods: Nine patients with stages T3-4N0-3M0 NPC underwent 18F-FMISO PET-CT before and during week 3 of radiotherapy. The hypoxic volume (GTVhypo) is automatically generated by applying a subthresholding algorithm within the gross tumor volume (GTV) with a tumor to muscle standardized uptake value (SUV) ratio of 1.3 on the 18F-FMISO PET-CT scan. Two proton plans were generated for each patient, a standard plan to 70 Gy and dose escalation plan with upfront boost followed by standard 70GyE plan. The stereotactic boost was planned with single-field uniform dose optimization using two fields to deliver 10 GyE in two fractions to GTVhypo. The standard plan was generated with IMPT with robust optimization to deliver 70GyE, 60GyE in 33 fractions using simultaneous integrated boost technique. A plan sum was generated for assessment., Results: Eight of nine patients showed tumor hypoxia on the baseline 18F-FMISO PET-CT scan. The mean hypoxic tumor volume was 3.9 cm 3 (range .9-11.9cm 3 ). The average SUVmax of the hypoxic volume was 2.2 (range 1.44-2.98). All the dose-volume parameters met the planning objectives for target coverage. Dose escalation was not feasible in three of eight patients as the D0.03cc of temporal lobe was greater than 75GyE., Conclusions: The utility of boost to the hypoxic volume before standard course of radiotherapy with IMPT is dosimetrically feasible in selected patients. Clinical trials are warranted to determine the clinical outcomes of this approach., (© 2023 John Wiley & Sons Australia, Ltd.)

Purpose or Objective: The COVID-19 pandemic has resulted in significant healthcare implications, with care for cancer patients compromised due to resource diversion towards battling the pandemic. We aim to investigate the impact of the peak wave of the pandemic in 2020 on the delivery of cancer care in Singapore, specifically via our nasopharyngeal carcinoma (NPC) treatment data. This study applies real world numbers to the impact of COVID-19 on cancer care delivery in Singapore. The choice of nasopharyngeal cancer allows a good direct estimate of common treatment measures such as time to biopsy, time to staging scans, time to treatment commencement, due to its clear protocol and algorithms for staging and treatment; thus serving as an excellent surrogate for the effectiveness and timeliness of the different aspects of cancer care delivery., Materials and Methods: In this retrospective study, we included all patients with newly diagnosed NPC from 1st January to 31st May from 2017 to 2020 at our centre. This time period was chosen as it coincided with the period in 2020 during the COVID-19 pandemic where there was the most strain on healthcare resources and the most restrictions on population movement within Singapore, which may impact on healthcare seeking behaviour. Narrowing down the time period to the first 5 months of the 4 respective years also allowed us to reduce the effect of annual seasonal variation in patient numbers seen as a result of holidays and festive periods such as the Lunar New Year and scheduled school holidays. Electronic medical records (EMR) were accessed. Only newly diagnosed NPC cases were included in our analysis. Patients with second synchronous primary malignancies or NPC disease recurrence were excluded. Data analysis was carried out using a combination of SPSS and Microsoft Excel., Results: Significantly, there was a reduction of 37-46.3% in newly diagnosed NPC cases during the peak of the COVID-19 pandemic from January to end May 2020 compared to the preceding three years. Despite the reduction in numbers of newly diagnosed NPC, there was no statistically significant differences in delay from biopsy to the first radiation oncology visit and from biopsy to the first day of treatment in 2020 compared to the preceding years. All the patients treated in our centre also received the standard NPC treatment for their disease stage as per international guidelines., Conclusion: We recommend a heightened awareness of the dangers of delaying cancer presentation and care in healthcare policies and resource allocation and at the same time, encourage patient's confidence in their ability to seek care. With the resurgence of new COVID-19 variants and case numbers worldwide and in Singapore, this study focuses upon the need to be aware of the exigencies of other clinical groups in resource utilization. It would be instructive to compare this study with future long term follow up to investigate the trajectory of our cancer care delivery, as well as survival outcomes., Competing Interests: Competing interestsThe authors declare that they have no competing interests., (© The Author(s) 2023.)

Aims: Ultra-hypofractionated radiotherapy (UHF-RT) is widely utilized in men with localized prostate cancer (PCa). There are limited data in Asian cohorts. We report the outcomes of a single-arm, phase II trial of UHF-RT from an Asian center., Methods: We recruited men with histologically confirmed, nonmetastatic localized PCa. UHF-RT regimens were 36.25 Gy (Cohort A) and 37.5 Gy (Cohort B) delivered in five fractions every other day over 1.5-2.5 weeks. Primary endpoint was physician-scored late genitourinary (GU) and gastrointestinal (GI) adverse events (AEs). Quality-of-life (QoL) was assessed by Expanded Prostate Cancer Index Composite (EPIC) at baseline, 1- and 2-year post-UHF-RT., Results: Between March 2014 and August 2019, 105 men were recruited; four were subsequently excluded from analysis. Median age was 68.0 (Interquartile range (IQR): 63.8-73.0) years. 26 (24.8%) and 68 (64.8%) men had NCCN-defined low-and intermediate-risk PCa, respectively. No late ≥G3 GU or GI toxicities were reported in both cohorts. Peak incidence of acute ≥G2 GU AEs at 14 days post-UHF-RT was 23.6% (17/72) and 24.0% (6/25) in Cohorts A and B, respectively; ≥G2 GI AEs were observed in 9.7% (7/72) and 36.0% (9/25), respectively. Late ≥G2 GU and GI AEs occurred in 4.7% and 3.1% of Cohort A patients, and 5.0% in Cohort B at 12 months, with no AEs at 24 months. EPIC scores changed minimally across all domains. At a median follow-up of 44.9 months, we recorded one (1.3%) biochemical relapse by the Phoenix criteria (Cohort A)., Conclusion: UHF-RT is well tolerated in Asian men and can be a recommended fractionation schema for localized PCa., (© 2021 The Authors. Asia-Pacific Journal of Clinical Oncology published by John Wiley & Sons Australia, Ltd.)

Background: This study aims to explore the role of four-dimensional (4D) transperineal ultrasound (TPUS) in the contouring of prostate gland with planning computed tomography (CT) images, in the absence of magnetic resonance imaging (MRI)., Materials and Methods: Five radiation oncologists (ROs) performed two rounds of prostate gland contouring (single-blinded) on CT-alone and CT/TPUS datasets obtained from 10 patients who underwent TPUS-guided external beam radiotherapy. Parameters include prostate volume, DICE similarity coefficient (DSC) and centroid position. Wilcoxon signed-rank test assessed the significance of inter-modality differences, and the intraclass correlation coefficient (ICC ) reflected inter- and intra-observer reliability of parameters., Results: Inter-modality analysis revealed high agreement (based on DSC and centroid position) of prostate gland contours between CT-alone and CT/TPUS. Statistical significant difference was observed in the superior-inferior direction of the prostate centroid position (p = 0.011). All modalities yielded excellent inter-observer reliability of delineated prostate volume with ICC > 0.9, mean DSC > 0.8 and centroid position: CT-alone (ICC = 1.000) and CT/TPUS (ICC = 0.999) left-right (L/R); CT-alone (ICC = 0.999) and CT/TPUS (ICC = 0.998) anterior-posterior (A/P); CT-alone (ICC = 0.999) and CT/TPUS (ICC = 1.000) superior-inferior (S/I). Similarly, all modalities yielded excellent intra-observer reliability of delineated prostate volume, ICC > 0.9 and mean DSC > 0.8. Lastly, intra-observer reliability was excellent on both imaging modalities for the prostate centroid position, ICC > 0.9., Conclusion: TPUS does not add significantly to the amount of anatomical information provided by CT images. However, TPUS can supplement planning CT to achieve a higher positional accuracy in the S/I direction if access to CT/MRI fusion is limited., Competing Interests: Conflict of interest The authors do not have any conflict of interest to declare., (© 2022 Greater Poland Cancer Centre.)

Background: Androgen deprivation therapy (ADT) and radiotherapy (RT) are the mainstay treatment for localized prostate cancer and recurrence after surgery. Cardiovascular (CV) toxicity of ADT is increasingly recognized, and the risk relates to pre-existing risk factors and ADT modalities. Despite ethnic differences in the prevalence of CV risk factors and variations of CV mortality, data on ADT-related cardiotoxicities in the Asian population remain inconclusive. Our registry-based study investigated ADT-related major adverse cardiovascular events (MACE) after primary or salvage RT., Methods: Our study combined two prospectively established registry databases from National Cancer Center Singapore and National Heart Center Singapore. The primary endpoint is time to first MACE after treatment. MACE is defined as myocardial infarction, stroke, unstable angina, or cardiovascular death. Two types of propensity score adjustments, including ADT propensity score as a covariate in the multivariable regression model and propensity score weighting, were applied to balance baseline features and CV risk factors between RT alone and RT + ADT groups., Results: From 2000 to 2019, 1940 patients received either RT alone (n = 494) or RT + ADT (n = 1446) were included. After a median follow-up of 10 years (RT) and 7.2 years (RT+ ADT), the cumulative incidence of MACE at 1, 3 and 9 years was 1.2, 5 and 16.2% in RT group, and 1.1, 5.2 and 17.6% in RT + ADT group, respectively. There were no differences in the incidence of MACE between 2 groups (HR 1.01, 95% CI 0.78-1.30, p = 0.969). Pre-treatment CV risk factors were common (80%), and CV disease (15.9%) was the second leading cause of death after prostate cancer (21.1%). On univariate analysis, older age, Indians and Malays, pre-existing CV risk factors, and history of MACE were associated with higher MACE risk. After propensity score adjustments, there remained no significant differences in MACE risk between RT + ADT and RT group on multivariable analysis., Conclusions: In our registry-based study, ADT is not associated with increased risk of major cardiovascular events among Southeast Asian men with prostate cancer after curative radiotherapy., (© 2022. The Author(s).)

Background and Objective: The magnitude of intra-fractional prostate displacement (change from initial position over time) is associated with the duration of the patient lying on the radiotherapy treatment couch. This study reports a minute-by-minute association and calculates the impact of this displacement on duration-dependent margins using real-time intra-fractional position data monitored by four-dimensional transperineal ultrasound (4D TPUS)., Materials and Methods: A total of 55 patients were recruited prospectively. Intra-fractional position of the prostate was monitored in real-time using a 4D TPUS Clarity® system. A total of 1745 monitoring sessions were analysed. Van Herk's margin recipe (2.5∑ + 1.64((σ 2  + σ p 2 ) 1/2  - σ p )) was used to estimate the duration-dependant margins for every minute, up to the 15th minute. Linear regression analysis was then performed on the overall margins against time and direction., Results: The mean intra-fractional position was 0.76 mm Inferior (Inf), 0 mm Lateral (Lat) and 0.94 mm Posterior (Post) at the 15th minute. A minimum margin expansion of 2.42 mm (Superior/Inf), 1.02 mm (Left/Right) and 2.65 mm (Anterior/Post) was required for an 8‑minute treatment compared to 4.29 mm (Sup/Inf), 1.84 mm (Lt/Rt) and 4.63 mm (Ant/Post) for a 15-minute treatment. The required margin expansion increased linearly (R 2  = 0.99) in all directions (p < 0.01). However, while there was no statistically significant difference (p = 0.10) in the required margin expansion in the Sup/Inf and Ant/Post directions respective of the time duration, the margins were much bigger compared to those in the Lt/Rt direction (p < 0.01)., Conclusion: We report our experience in deriving the minimum duration-dependant margin to generate the required planning target volume for prostate radiotherapy. The required margin increases linearly in all directions within the 15-min duration; thus, the margin will depend on the duration of the technique chosen (IMRT/VMAT/3DCRT/proton).

Background and Purpose: To compare physician and patient reported xerostomia and correlate xerostomia with dosimetric and clinical parameters for nasopharyngeal cancer (NPC) patients treated with intensity modulated radiotherapy (IMRT) and chemotherapy., Patients and Methods: We analyzed the data of 172 patients with locally advanced NPC. Xerostomia was evaluated via physician-rated xerostomia based on RTOG morbidity score (E1), patient-rated dry mouth (E2) and patient-rated sticky saliva (E3) based on EORTC QLQ-HN35 questionnaire. Primary endpoint was the presence of moderate to severe xerostomia at 2-year after completion of IMRT., Results: The levels of physician reported xerostomia (E1) were consistently lower than patient reported dry mouth (E2) over time. The incidence of patients with xerostomia at 3-month post RT was 58% based on E1, 70% based on E2, and 51% based on E3. The corresponding incidence rates at 2-year post RT was 26% (E1), 36% (E2) and 21% (E3). The incidence of patients with xerostomia at 1-year post RT was close to that at 2-year post RT for all the 3 endpoints. The average Dmean of parotid glands was 41.5 Gy (range: 31.0 Gy-65.9 Gy, median: 40.7 Gy). No dosimetric parameters were significantly associated with xerostomia., Conclusion: Significant proportion of patients still experienced long term xerostomia with IMRT. Dose-effect relationships between xerostomia and the parotid glands were not observed in this study., (Copyright © 2019 Elsevier B.V. All rights reserved.)

Objective: Prostate cancers (PCa) in Asian individuals are molecularly distinct from those found in their Caucasian counterparts. There is no risk stratification tool for Asian men with rapid biochemical recurrence (BCR) following radical prostatectomy (RadP). This study aims to assess the detection rate of 68 Ga-prostate-specific membrane antigen-positron emission tomography/computed tomography (PSMA-PET/CT) for diagnosis of clinical recurrence and as a treatment decision making tool in Asian patients with BCR post-RadP., Methods: 68 Ga PSMA-PET and CT body with/without bone scan [conventional workup (CWU)] were performed in 55 Asian patients with BCR within 36 months post-RadP. Two blinded reviewers assessed the images. Detection rates of 68 Ga PSMA-PET/CT were evaluated, and impact on management was reviewed by comparison with CWU., Results: Median time to BCR post-RadP was 8.1 months. Detection rate for 68 Ga PSMA-PET/CT was 80% (44/55). A positive scan was significantly associated with increasing prostate-specific antigen (PSA) level [odds ratio (OR) = 1.13 (95% CI 1.05-1.30), P = 0.017], but not with higher Gleason grade or shorter PSA doubling time. Compared to CWU, 68 Ga PSMA-PET/CT detected an additional 106 lesions in 33/44 patients with a positive scan, resulting in a change in management in 25/44 (56.8%) patients: 10 to hormonal therapy (HT) and whole pelvis radiotherapy (RT) in addition to bed RT, and 15 to palliative HT alone., Conclusions: In the present report, we demonstrated the diagnostic and treatment decision utility of 68 Ga PSMA-PET/CT in Asian men with rapid BCR. Detection of small volume nodal and systemic recurrences at low PSA levels (< 1.0 ng/mL) highlights the role of the tool in assigning patients to treatment intensification with HT-RT or palliative HT in polymetastatic disease.

Background: Inter-patient heterogeneity in radiation-induced DNA damage responses is proposed to reflect intrinsic variations in tumour and normal tissue radiation sensitivity, but the prediction of phenotype by a molecular biomarker is influenced by clinical confounders and assay reproducibility. Here, we characterised the intrapatient and inter-patient heterogeneity in biomarkers of DNA damage and repair and radiation-induced apoptosis., Methods: We enrolled 85 of 172 patients with locally advanced nasopharynx cancer from a randomised controlled phase II/III trial of induction chemotherapy added to chemo-radiotherapy. G 0 blood lymphocytes were harvested from these patients, and irradiated with 1, 4, and 8 Gy ex vivo. DNA damage induction (1 Gy 0.5 h) and repair (4 Gy 24 h) were assessed by duplicate γH2AX foci assays in 50-100 cells. Duplicate FLICA assays performed at 48 h post-8 Gy were employed as surrogate of radiation-induced apoptosis; %FLICA-positive cells were quantified by flow cytometry., Results: We observed limited intrapatient variation in γH2AX foci and %FLICA readouts; median difference of duplicate foci scores was - 0.37 (IQR = - 1.256-0.800) for 1 Gy 0.5 h and 0.09 (IQR = - 0.685-0.792) for 4 Gy 24 h; ICC of ≥0.80 was observed for duplicate %FLICA 0Gy and %FLICA 8Gy assays of CD4+ and CD8+ T lymphocytes. As expected, we observed wide inter-patient heterogeneity in both assays that was independent of intrapatient variation and clinical covariates, with the exception of age, which was inversely correlated with %FLICA background-corrected (Spearman R = - 0.406, P < 0.001 [CD4+]; R = - 0.220, P = 0.04 [CD8+]). Lastly, an exploratory case-control analysis indicates increased levels of γH2AX foci at 4 Gy 24 h in patients with severe late radiotherapy-induced xerostomia (P = 0.05)., Conclusion: Here, we confirmed the technical reproducibility of DNA damage response assays for clinical implementation as biomarkers of clinical radiosensitivity in nasopharynx cancer patients.

Purpose To investigate for a prognostic index (PI) to personalize recommendations for salvage intensity-modulated radiotherapy (IMRT) in patients with locally recurrent nasopharyngeal carcinoma (lrNPC). Methods Patients with lrNPC from two academic institutions (Sun Yat-Sen University Cancer Center [SYSUCC-A; n = 251 (training cohort)] and National Cancer Centre Singapore [NCCS; n = 114] and SYSUCC-B [n = 193 (validation cohorts)]) underwent salvage treatment with IMRT from 2001 to 2015. Primary and secondary clinical end points were overall survival (OS) and grade 5 toxicity-free rate (G5-TFR), respectively. Covariate inclusion to the PIs was qualified by a multivariable two-sided P < .05. Discrimination and calibration of the PIs were assessed. Results The primary PI comprised covariates that were adversely associated with OS in the training cohort (gross tumor volume recurrence hazard ratio [HR], 1.01/mL increase [ P < .001], age recurrence HR, 1.02/year increase [ P = .008]; repeat IMRT equivalent dose in 2-Gy fractions [EQD2] ≥ 68 Gy HR, 1.42 [ P = .03]; prior radiotherapy-induced grade ≥ 3 toxicities HR, 1.90 [ P = .001]; recurrent tumor [rT]-category 3 to 4 HR, 1.96 [ P = .005]), in ascending order of weight. Discrimination of the PI for OS was comparable between training and both validation cohorts (Harrell's C = 0.71 [SYSUCC-A], 0.72 [NCCS], and 0.69 [SYSUCC-B]); discretization by using a fixed PI score cutoff of 252 determined from the training data set yielded low- and high-risk subgroups with disparate OS in the validation cohorts (NCCS HR, 3.09 [95% CI, 1.95 to 4.89]; SYSUCC-B HR, 3.80 [95% CI, 2.55 to 5.66]). Our five-factor PI predicted OS and G5-TFR (predicted v observed 36-month OS and G5-TFR, 22% v 15% and 38% v 44% for high-risk NCCS and 26% v 31% and 45% v 46% for high-risk SYSUCC-B). Conclusion We present a validated PI for robust clinical stratification of radioresistant NPC. Low-risk patients represent ideal candidates for curative repeat IMRT, whereas novel clinical trials are needed in the unfavorable high-risk subgroup.