Objective:To explore the value of an integrated modeling approach combining radiomics, dosiomics, and clinical factors in the prediction of the locoregional recurrence (LRR) risk after radiotherapy for patients with locoregionally advanced hypopharyngeal squamous cell carcinoma (HPSCC), in order to provide supplementary clinical evidence and decision-making basis for personalized treatment for this rare disease characterized by low incidence and poor prognosis.Methods:The clinical images and pathological data were retrospectively enrolled from 76 HPSCC patients treated at the Peking University Cancer Hospital from October 2011 to July 2020. The planning gross tumor volumes (PGTVs) were taken as the volumes of interest (VOIs). A total of 1 316 radiomic and dosiomic features were extracted from the planning CT and dose distribution images. After stability testing, feature dimensionality reduction was achieved using least absolute shrinkage and selection operator (LASSO) regression and principal component analysis (PCA), with radiomic principal components (RPCs) and dosiomic principal components (DPCs) obtained, respectively. Using various combinations of RPCs, DPCs, and clinical variables as predictors, multivariate Cox regression models were developed after 5-fold cross-validation 100 times. The model performance was evaluated based on the Akaike information criterion (AIC) and concordance index (C-index).Results:Using two RPCs and three DPCs selected, dosiomics and radiomic Cox proportional hazards models were constructed, with C-index values of 0.781 and 0.778 and AIC values of 94.44 and 92.27, respectively. The result indicated that one RPC and three DPCs showed significant associations in Cox regression ( P < 0.05). Other prediction models were established by integrating the clinical data of patients with radiomic features, dosiomic features, or both. The prediction result demonstrated that compared to models based on individual factors or dual components, the multi-omics model yielded the highest prediction accuracy (C-index: 0.823, AIC: 84.94). Conclusions:Integrated models that combine radiomic features, dosiomic features, and clinical factors demonstrate great potential for enhancing the accuracy of LRR risk prediction. These models are expected to provide decision-making support for devising personalized treatment strategies and ultimately improve the prognosis of HPSCC patients.

Objective:To explore the value of an integrated modeling approach combining radiomics, dosiomics, and clinical factors in the prediction of the locoregional recurrence (LRR) risk after radiotherapy for patients with locoregionally advanced hypopharyngeal squamous cell carcinoma (HPSCC), in order to provide supplementary clinical evidence and decision-making basis for personalized treatment for this rare disease characterized by low incidence and poor prognosis.Methods:The clinical images and pathological data were retrospectively enrolled from 76 HPSCC patients treated at the Peking University Cancer Hospital from October 2011 to July 2020. The planning gross tumor volumes (PGTVs) were taken as the volumes of interest (VOIs). A total of 1 316 radiomic and dosiomic features were extracted from the planning CT and dose distribution images. After stability testing, feature dimensionality reduction was achieved using least absolute shrinkage and selection operator (LASSO) regression and principal component analysis (PCA), with radiomic principal components (RPCs) and dosiomic principal components (DPCs) obtained, respectively. Using various combinations of RPCs, DPCs, and clinical variables as predictors, multivariate Cox regression models were developed after 5-fold cross-validation 100 times. The model performance was evaluated based on the Akaike information criterion (AIC) and concordance index (C-index).Results:Using two RPCs and three DPCs selected, dosiomics and radiomic Cox proportional hazards models were constructed, with C-index values of 0.781 and 0.778 and AIC values of 94.44 and 92.27, respectively. The result indicated that one RPC and three DPCs showed significant associations in Cox regression ( P < 0.05). Other prediction models were established by integrating the clinical data of patients with radiomic features, dosiomic features, or both. The prediction result demonstrated that compared to models based on individual factors or dual components, the multi-omics model yielded the highest prediction accuracy (C-index: 0.823, AIC: 84.94). Conclusions:Integrated models that combine radiomic features, dosiomic features, and clinical factors demonstrate great potential for enhancing the accuracy of LRR risk prediction. These models are expected to provide decision-making support for devising personalized treatment strategies and ultimately improve the prognosis of HPSCC patients.

Objective To compare the efficacy of highly active anti-retroviral treatment (HAART) with and with AZT. Methods Data of 235 HIV patients accepted from Oct. 2010 to Nov. 2013 who took AZT (133) or TDF (102) containing regimen as first-line HAART are analyzed retrospectively. CD4+ T cell counts acted as the base line after 12 months of HAART. Increase in CD4+ T cell count number after the HAART, virological failures and drug resistance were compared between the two groups. Results The two groups had comparable baseline CD4+T cell count, gender ratio, and HIV transmision mode; after 12 months of HARRT, no statistically differences were found between the two groups with regard to CD4 + T cell count and increase in CD4+ T cell count after the 12-month HAART (P > 0.05); AZT-containing group had more virological failure (3/0). Meanwhile AZT-containing group had one thymidine analog mutation (TAMs) which confers resistance to AZT (P > 0.05 for both). Conculsion The two HAARTs have same immunological effects; AZT-containing group exhibits 2.2% viralogial failure, but its direct relationship with AZT has not been confirmed.