Improved survival for patients with AIDS-related Kaposi's sarcoma.

1994 ◽  
Vol 12 (9) ◽  
pp. 1910-1916 ◽  
Author(s):  
S A Miles ◽  
H Wang ◽  
R Elashoff ◽  
R T Mitsuyasu
Keyword(s):  
Prognostic Factors ◽  
Kaposi's Sarcoma ◽  
Proportional Hazards ◽  
Cox Regression ◽  
Cox Model ◽  
Medical Center ◽  
Kaposi’S Sarcoma ◽  
Cell Number ◽  
Cox Proportional Hazards ◽  
Cd4 Cell

PURPOSE We retrospectively analyzed all patients with AIDS-related Kaposi's sarcoma (AIDS-KS) seen at one large California medical center to delineate factors that may have contributed to a relative decline in survival. METHODS Potential prognostic factors were analyzed individually, using the Cox proportional hazards regression model, for their association with survival. After a stepwise Cox regression procedure was applied to those factors that showed a significant effect on survival, a subset of factors that best predicted survival was identified. We then quantified the effect of the year of diagnosis on survival using a univariate Cox model. Next, we combined the year of diagnosis with the subset of prognostic factors previously identified into the Cox model to examine survival after adjustment for the prognostic factors. Survival distribution was estimated by the Kaplan-Meier method, and the 95% confidence interval for the median survival was computed using the modified reflected method. RESULTS In 688 patients, we identified four baseline variables that best predicted survival: CD4 cell number, hematocrit, number of KS lesions, and body mass index (BMI). Adjusted for these predictive factors, there was a significant improvement in survival for patients with AIDS-KS over the last 6 years. CONCLUSION Contrary to prior reports, survival has increased for patients with AIDS-KS. The apparent increase in observed mortality is most likely due to a decline in the CD4 cell number at presentation.

Body weight loss (BWL) as a prognostic/predictive factor in previously treated patients (pts) with metastatic gastric or gastroesophageal junction cancer (mGC/GEJC): Post-hoc analyses of the phase III tags trial.

2021 ◽  
Vol 39 (3_suppl) ◽  
pp. 476-476
Author(s):  
Michele Ghidini ◽  
Howard S. Hochster ◽  
Toshihiko Doi ◽  
Eric Van Cutsem ◽  
Lukas Makris ◽  
...  
Keyword(s):  
Body Weight ◽  
Prognostic Factors ◽  
Proportional Hazards ◽  
Cox Model ◽  
Gastroesophageal Junction ◽  
Body Weight Loss ◽  
Cox Proportional Hazards ◽  
Phase Iii ◽  
Body Weight Data ◽  
Grade 3

476 Background: Nutritional status is closely linked to cancer mortality, and BWL has been shown to be prognostic for survival in curative, first-, and second-line settings in mGC/GEJC. In the phase III TAGS trial, trifluridine/tipiracil (FTD/TPI) showed clinical benefit versus placebo (PBO) and manageable safety in pts with mGC/GEJC who had received ≥2 prior chemotherapy regimens. The association of early BWL with survival outcomes in TAGS was examined in retrospective analyses. Methods: The TAGS intent-to-treat (ITT) population was categorized into pts who experienced <3% or ≥3% BWL from the start of treatment until day 1 of cycle 2 (each cycle: 28 days). Overall survival (OS), and progression-free survival (PFS) were compared between subgroups within each treatment arm due to significant imbalances of early BWL between treatment arms. The effect of early BWL on OS was assessed by a univariate Cox proportional hazards (PH) model as well as a multivariate Cox PH model that adjusted for baseline prognostic factors identified in the original ITT analysis. Results: Body weight data were available for 451 of 507 (89%) pts in the study (n=304, FTD/TPI; n=147, PBO). In the FTD/TPI and PBO arms, respectively, 74% (224/304 pts) and 65% (95/147) experienced <3% BWL, whereas 26% (80/304) and 35% (52/147) experienced ≥3% BWL at the end of cycle 1. Pts with <3% BWL had longer OS than those with ≥3% BWL in both FTD/TPI (median [m] OS: 6.5 vs 4.9 months [mo]; hazard ratio [HR], 0.75; 95% CI, 0.55–1.02) and PBO arms (mOS: 6.0 vs 2.5 mo; HR, 0.32; 95% CI, 0.21–0.49). The PFS HR for pts with <3% BWL vs ≥3% BWL was 0.95 (95% CI, 0.71–1.25; mPFS, 2.1 vs 1.9 mo) in the FTD/TPI group and 0.49 (95% CI, 0.34–0.72; mPFS, 1.9 vs 1.7 mo) in the PBO group. In the pooled ITT population, the unadjusted HR for the <3% vs ≥3% BWL group calculated using a univariate Cox model was 0.58 (95% CI, 0.46–0.73), indicating a strong prognostic effect of early BWL. Results of multivariate analyses were consistent with univariate analyses and suggested that early BWL was both a prognostic ( P<0.0001) and predictive (interaction P=0.0003) factor for OS in pts with mGC/GEJC. Grade ≥3 adverse events (AEs) of any cause were reported in 77% and 82% of FTD/TPI-treated pts in the <3% and ≥3% BWL subgroups, respectively, and in 45% and 67% of placebo-treated pts in the <3% and ≥3% BWL subgroups. Conclusions: To our knowledge, this is the first analysis to show that BWL is negatively associated with survival in pts with mGC/GEJC receiving third- or later-line treatment. In TAGS, early BWL (≥3% BWL at the end of cycle 1) was a strong negative prognostic factor for OS regardless of FTD/TPI or PBO treatment. Grade ≥3 AE frequencies were similar in FTD/TPI-treated pts with <3% or ≥3% BWL. The relationship of BWL to other prognostic factors will be explored further. Clinical trial information: NCT02500043.

Differential Impact of Combination Antiretroviral Therapy in Preventing Kaposi's Sarcoma With and Without Visceral Involvement

2006 ◽  
Vol 24 (21) ◽  
pp. 3408-3414 ◽  
Author(s):  
Sophie Grabar ◽  
Bruno Abraham ◽  
Aba Mahamat ◽  
Pascal Del Giudice ◽  
Eric Rosenthal ◽  
...  
Keyword(s):  
Kaposi's Sarcoma ◽  
Proportional Hazards ◽  
Specific Effect ◽  
Kaposi’S Sarcoma ◽  
Cox Proportional Hazards ◽  
Nucleoside Analog ◽  
Large Hospital ◽  
Visceral Involvement ◽  
The Impact

Purpose To study the impact of different potent combined antiretroviral treatment (cART) on the incidence of HIV-associated Kaposi's sarcoma (KS) with and without visceral involvement. Patients and Methods Patients were selected from the French Hospital Database on HIV, a large hospital cohort. The risk of KS was estimated by using Cox proportional hazards models adjusting for age, the CD4 cell nadir, the HIV exposure category, prior AIDS, cART, and the type of cART regimen. cART regimens were distinguished according to whether they contained protease inhibitor (PI), non-nucleoside analog (NNRTI), both, or only nucleoside analog (NRTI). Separate analyzes were conducted according to the initial visceral involvement of KS. Results Among the 54,999 patients included in this study (182,756 person-years of follow-up), 1,634 patients were diagnosed with KS during follow-up, of whom 421 had visceral involvement at diagnosis. The KS incidence rate fell from 32 per 1,000 person-years in 1993 to 1994 to 3 per 1,000 person-years after 1999. PI-containing and NNRTI-containing cART regimens were associated with similar reductions in the risk of KS (hazard ratio, 0.68; 95%CI, 0.61 to 0.75; HR, 0.62; 95% CI, 0.54 to 0.71, respectively). The risk of visceral KS fell more strongly than the risk of cutaneous KS (> 50% and < 30%, respectively). Conclusion The incidence of KS, and especially visceral KS, has fallen sharply since the advent of cART. This effect is likely due to immune restoration rather than to a specific effect on the tumoral process, as PI-containing and NNRTI-containing regimens had similar preventive efficacy.

scholarly journals Prognostic Factors and a Nomogram Predicting Overall Survival in Patients with Limb Chondrosarcomas: A Population-Based Study

2021 ◽  
Vol 2021 ◽  
pp. 1-17
Author(s):  
Xinjie Wu ◽  
Yanlei Wang ◽  
Wei Sun ◽  
Mingsheng Tan
Keyword(s):  
Overall Survival ◽  
Prognostic Factors ◽  
Clinical Decision Making ◽  
Proportional Hazards ◽  
Cox Regression ◽  
Training Group ◽  
Clinical Decision ◽  
The United States ◽  
Cox Proportional Hazards ◽  
Predictive Values

Introduction. We aimed to develop and validate a nomogram for predicting the overall survival of patients with limb chondrosarcomas. Methods. The Surveillance, Epidemiology, and End Results (SEER) program database was used to identify patients diagnosed with chondrosarcomas, from which data was extracted from 18 registries in the United States between 1973 and 2016. A total of 813 patients were selected from the database. Univariate and multivariate analyses were performed using Cox proportional hazards regression models on the training group to identify independent prognostic factors and construct a nomogram to predict the 3- and 5-year survival probability of patients with limb chondrosarcomas. The predictive values were compared using concordance indexes ( C -indexes) and calibration plots. Results. All 813 patients were randomly divided into a training group ( n = 572 ) and a validation group ( n = 241 ). After univariate and multivariate Cox regression, a nomogram was constructed based on a new model containing the predictive variables of age, site, grade, tumor size, histology, stage, and use of surgery, radiotherapy, or chemotherapy. The prediction model provided excellent C -indexes (0.86 and 0.77 in the training and validation groups, respectively). The good discrimination and calibration of the nomograms were demonstrated for both the training and validation groups. Conclusions. The nomograms precisely and individually predict the overall survival of patients with limb chondrosarcomas and could assist personalized prognostic evaluation and individualized clinical decision-making.

scholarly journals Elevated levels of serum urokinase plasminogen activator predict poor prognosis in hepatocellular carcinoma after resection

2019 ◽  
Author(s):  
Ming-Chao Tsai ◽  
Yi-Hao Yen ◽  
Kuo-Chin Chang ◽  
Chao-Hung Hung ◽  
Chien-Hung Chen ◽  
...  
Keyword(s):  
Prognostic Factors ◽  
Plasminogen Activator ◽  
Proportional Hazards ◽  
Cox Regression ◽  
Urokinase Plasminogen Activator ◽  
Cox Proportional Hazards ◽  
Rank Test ◽  
Enzyme Linked Immunosorbent Assay ◽  
Chang Gung Memorial Hospital ◽  
Pretreatment Serum

Abstract Background Urokinase plasminogen activator (uPA) is an extracellular matrix-degrading protease that is involved in the invasiveness and progression of cancer. There is good evidence that uPA expression is a clinically relevant biomarker in some solid tumors, but its role in hepatocellulcar carcinoma (HCC) is uncertain. We evaluated the prognostic value of serum uPA before surgery in HCC patients receiving curative resection.Methods Serum uPA levels were determined by enzyme-linked immunosorbent assay in 282 HCC patients who received complete liver resections at Kaohsiung Chang Gung Memorial Hospital. Overall survival (OS) curves were constructed using the Kaplan-Meier method and compared using the log-rank test. A Cox proportional -hazards regression model was used to identify independent prognostic factors. The median follow-up time was 52 months.Results Patients with higher pretreatment serum uPA (≥1 ng/ml) had significantly shorter OS (p = 0.002). Patients with liver cirrhosis, hypoalbuminemia, and thrombocytopenia were significantly more likely to present with elevated uPA levels. Multivariate Cox regression analyses indicated that high pretreatment serum uPA [hazard ratio (HR), 1.848, p = 0.006], vascular invasion (HR, 2.940, p <0.001), and pathology stage III/IV (HR, 3.517, p<0.001) were independent prognostic factors for OS. In further stratified analyses, the combination of serum uPA and AFP had more capacity to predict OS.Conclusions We conclude that uPA is a clinically relevant biomarker in HCC patients receiving curative resection, with higher expression of uPA being associated with higher mortality. This also highlights the potential utility of uPA as a therapeutic target for improved treatment strategies.

Prognostic factors and survival of patients with brain metastasis (BM) from breast cancer (BC) who underwent craniotomy.

2013 ◽  
Vol 31 (15_suppl) ◽  
pp. e22017-e22017
Author(s):  
Jose Pablo Leone ◽  
Diana E. Cunningham ◽  
Adrian Lee ◽  
Rohit Bhargava ◽  
Ronald L. Hamilton ◽  
...  
Keyword(s):  
Breast Cancer ◽  
Prognostic Factors ◽  
Proportional Hazards ◽  
Cox Regression ◽  
Primary Diagnosis ◽  
Cox Proportional Hazards ◽  
Factors Associated ◽  
Brain Radiotherapy ◽  
Whole Brain Radiation ◽  
First Recurrence

e22017 Background: BC is the second most frequent cause of BM after lung cancer, with metastases occurring in 10-16% of all patients. BM in patients with BC is a catastrophic event that results in poor prognosis. Identification of prognostic factors associated with breast cancer brain metastases (BCBM) could help to identify patients at risk. The aim of this study was to assess clinical characteristics, prognostic factors and survival of patients with BCBM who had craniotomy and resection in a series of patients treated with modern multimodality therapy. Methods: We analyzed 42 patients with BCBM who underwent resection. Patients were diagnosed with BC between April 1994 and May 2010. Cox proportional hazards regression was selected to describe factors associated with time to BM, survival from the date of first recurrence, and overall survival (OS). Results: Median age was 51 years (range 24-74). Median follow-up was 4.2 years (range 0.6-18.5). The mean time to BM from primary diagnosis was 49 months (range 0-206.22). Patients had a median of 2 BM with a median size of 3.25 cm. The proportion of the biological subtypes of BC was ER+/HER2- 25%, ER+/HER2+ 15%, ER-/HER2+ 30% and ER-/HER2- 30%. Brain radiotherapy was given to 28 patients, of which 10 had stereotactic radiosurgery, 7 whole brain radiation, and 11 both. Median OS from the date of primary diagnosis was 5.74 years. Median survival after diagnosis of BM was 1.33 years. In multivariate Cox regression analyses, stage was the only factor associated with shorter time to the development of BM (P=0.059), whereas age was the only factor associated with survival from the date of recurrence (P=0.027) and with OS (P=0.037). Controlling for age and stage, neither the biological subtype of cancer, the radiation modality nor the site of first recurrence showed any impact on survival. Conclusions: Stage at primary diagnosis correlated with shorter time to the development of BM, while age at diagnosis was associated with shorter survival in BCBM. None of the other clinical factors had influence on survival.

scholarly journals Comprehensive analysis of ceRNA network related to lincRNA in glioblastoma and prediction of clinical prognosis

BMC Cancer ◽  
2021 ◽  
Vol 21 (1) ◽  
Author(s):  
Guangdong Liu ◽  
Danian Liu ◽  
Jingjing Huang ◽  
Jianxin Li ◽  
Chuang Wang ◽  
...  
Keyword(s):  
Proportional Hazards ◽  
Cox Regression ◽  
Cox Model ◽  
Differential Expression Analysis ◽  
Cox Proportional Hazards ◽  
Gene Set Enrichment Analysis ◽  
Normal Brain ◽  
Tissue Samples ◽  
Clinical Prognosis ◽  
Cerna Network

Abstract Background Long intergenic non-coding RNAs (lincRNAs) are capable of regulating several tumours, while competitive endogenous RNA (ceRNA) networks are of great significance in revealing the biological mechanism of tumours. Here, we aimed to study the ceRNA network of lincRNA in glioblastoma (GBM). Methods We obtained GBM and normal brain tissue samples from TCGA, GTEx, and GEO databases, and performed weighted gene co-expression network analysis and differential expression analysis on all lincRNA and mRNA data. Subsequently, we predicted the interaction between lincRNAs, miRNAs, and target mRNAs. Univariate and multivariate Cox regression analyses were performed on the mRNAs using CGGA data, and a Cox proportional hazards regression model was constructed. The ceRNA network was further screened by the DEmiRNA and mRNA of Cox model. Results A prognostic prediction model was constructed for patients with GBM. We assembled a ceRNA network consisting of 18 lincRNAs, 6 miRNAs, and 8 mRNAs. Gene Set Enrichment Analysis was carried out on four lincRNAs with obvious differential expressions and relatively few studies in GBM. Conclusion We identified four lincRNAs that have research value for GBM and obtained the ceRNA network. Our research is expected to facilitate in-depth understanding and study of the molecular mechanism of GBM, and provide new insights into targeted therapy and prognosis of the tumour.

scholarly journals Elevated levels of serum urokinase plasminogen activator predict poor prognosis in hepatocellular carcinoma after resection

2019 ◽  
Author(s):  
Ming-Chao Tsai(Former Corresponding Author) ◽  
Yi-Hao Yen ◽  
Kuo-Chin Chang ◽  
Chao-Hung Hung ◽  
Chien-Hung Chen ◽  
...  
Keyword(s):  
Prognostic Factors ◽  
Plasminogen Activator ◽  
Proportional Hazards ◽  
Cox Regression ◽  
Urokinase Plasminogen Activator ◽  
Cox Proportional Hazards ◽  
Rank Test ◽  
Enzyme Linked Immunosorbent Assay ◽  
Chang Gung Memorial Hospital ◽  
Pretreatment Serum

Abstract Background Urokinase plasminogen activator (uPA) is an extracellular matrix-degrading protease that is involved in the invasiveness and progression of cancer. There is good evidence that uPA expression is a clinically relevant biomarker in some solid tumors, but its role in hepatocellulcar carcinoma (HCC) is uncertain. We evaluated the prognostic value of serum uPA before surgery in HCC patients receiving curative resection.Methods Serum uPA levels were determined by enzyme-linked immunosorbent assay in 282 HCC patients who received complete liver resections at Kaohsiung Chang Gung Memorial Hospital. Overall survival (OS) curves were constructed using the Kaplan-Meier method and compared using the log-rank test. A Cox proportional -hazards regression model was used to identify independent prognostic factors. The median follow-up time was 52 months.Results Patients with higher pretreatment serum uPA (≥1 ng/ml) had significantly shorter OS ( p = 0.002). Patients with liver cirrhosis, hypoalbuminemia, and thrombocytopenia were significantly more likely to present with elevated uPA levels. Multivariate Cox regression analyses indicated that high pretreatment serum uPA [hazard ratio (HR), 1.848, p = 0.006], vascular invasion (HR, 2.940, p <0.001), and pathology stage III/IV (HR, 3.517, p <0.001) were independent prognostic factors for OS. In further stratified analyses, the combination of serum uPA and AFP had more capacity to predict OS.Conclusions We conclude that uPA is a clinically relevant biomarker in HCC patients receiving curative resection, with higher expression of uPA being associated with higher mortality. This also highlights the potential utility of uPA as a therapeutic target for improved treatment strategies.

scholarly journals Comprehensive Analysis of lincRNA-Related ceRNA Network in Glioblastoma

2020 ◽  
Author(s):  
Guangdong Liu ◽  
Danian Liu ◽  
Jingjing Huang ◽  
Jianxin Li ◽  
Chuang Wang ◽  
...  
Keyword(s):  
Proportional Hazards ◽  
Cox Regression ◽  
Cox Model ◽  
Differential Expression Analysis ◽  
Enrichment Analysis ◽  
Cox Proportional Hazards ◽  
Gene Set Enrichment Analysis ◽  
Normal Brain ◽  
Tissue Samples ◽  
Cerna Network

Abstract BackgroundLong intergenic non-coding RNAs (lincRNAs) are capable of regulating several tumours, while competitive endogenous RNA (ceRNA) networks are of great significance in revealing the biological mechanism of tumours. Currently, there is a dearth of studies on the ceRNA network of lincRNAs in glioblastoma (GBM), which we aimed to assess in the present study. MethodsWe obtained GBM and normal brain tissue samples from TCGA, GTEx, and GEO databases, and performed weighted gene co-expression network analysis and differential expression analysis on all lincRNA and mRNA data. Subsequently, we predicted the interaction between lincRNAs, miRNAs, and target mRNAs. Univariate and multivariate Cox regression analyses were performed on the mRNAs using CGGA data, and a Cox proportional hazards regression model was constructed. ResultsAccording to the Cox model, we assembled a ceRNA network consisting of 23 lincRNAs, 14 miRNAs, and 13 mRNAs. Gene Set Enrichment Analysis was carried out on four lincRNAs with obvious differential expressions and relatively few studies in GBM. ConclusionWe identified four lincRNAs that have the most research values for GBM and obtained the ceRNA network. Our research is expected to facilitate in-depth understanding and study of the molecular mechanism of GBM, and provide new insights into targeted therapy and prognosis of the tumour.

scholarly journals Hierarchical Neural Regression Models for Customer Churn Prediction

2013 ◽  
Vol 2013 ◽  
pp. 1-9 ◽  
Author(s):  
Golshan Mohammadi ◽  
Reza Tavakkoli-Moghaddam ◽  
Mehrdad Mohammadi
Keyword(s):  
Data Mining ◽  
Hierarchical Models ◽  
Proportional Hazards ◽  
Cox Regression ◽  
Cox Model ◽  
Cox Proportional Hazards ◽  
Churn Prediction ◽  
Data Mining Techniques ◽  
Customer Churn ◽  
Customer Churn Prediction

As customers are the main assets of each industry, customer churn prediction is becoming a major task for companies to remain in competition with competitors. In the literature, the better applicability and efficiency of hierarchical data mining techniques has been reported. This paper considers three hierarchical models by combining four different data mining techniques for churn prediction, which are backpropagation artificial neural networks (ANN), self-organizing maps (SOM), alpha-cut fuzzyc-means (α-FCM), and Cox proportional hazards regression model. The hierarchical models are ANN + ANN + Cox, SOM + ANN + Cox, andα-FCM + ANN + Cox. In particular, the first component of the models aims to cluster data in two churner and nonchurner groups and also filter out unrepresentative data or outliers. Then, the clustered data as the outputs are used to assign customers to churner and nonchurner groups by the second technique. Finally, the correctly classified data are used to create Cox proportional hazards model. To evaluate the performance of the hierarchical models, an Iranian mobile dataset is considered. The experimental results show that the hierarchical models outperform the single Cox regression baseline model in terms of prediction accuracy, Types I and II errors, RMSE, and MAD metrics. In addition, theα-FCM + ANN + Cox model significantly performs better than the two other hierarchical models.

Prognostic factors and survival of patients with brain metastasis (BM) from breast cancer (BC) who underwent craniotomy.

2013 ◽  
Vol 31 (26_suppl) ◽  
pp. 143-143
Author(s):  
Jose Pablo Leone ◽  
Diana E. Cunningham ◽  
Adrian Lee ◽  
Rohit Bhargava ◽  
Ronald L. Hamilton ◽  
...  
Keyword(s):  
Breast Cancer ◽  
Prognostic Factors ◽  
Proportional Hazards ◽  
Cox Regression ◽  
Primary Diagnosis ◽  
Cox Proportional Hazards ◽  
Factors Associated ◽  
Brain Radiotherapy ◽  
Whole Brain Radiation ◽  
First Recurrence

143 Background: BC is the second most frequent cause of BM after lung cancer, with metastases occurring in 10% - 16% of all patients. BM in patients with BC is a catastrophic event that results in poor prognosis. Identification of prognostic factors associated with breast cancer brain metastases (BCBM) could help to identify patients at risk. The aim of this study was to assess clinical characteristics, prognostic factors and survival of patients with BCBM who had craniotomy and resection in a series of patients treated with modern multimodality therapy. Methods: We analyzed 42 patients with BCBM who underwent resection. Patients were diagnosed with BC between April 1994 and May 2010. Cox proportional hazards regression was selected to describe factors associated with time to BM, survival from the date of first recurrence, and overall survival (OS). Results: Median age was 51 years (range 24-74). Median follow-up was 4.2 years (range 0.6-18.5). The mean time to BM from primary diagnosis was 49 months (range 0-206.22). Patients had a median of 2 BM with a median size of 3.25 cm. The proportion of the biological subtypes of BC was ER+/HER2- 25%, ER+/HER2+ 15%, ER-/HER2+ 30% and ER-/HER2- 30%. Brain radiotherapy was given to 28 patients, of which 10 had stereotactic radiosurgery, 7 whole brain radiation, and 11 both. Median OS from the date of primary diagnosis was 5.74 years. Median survival after diagnosis of BM was 1.33 years. In multivariate Cox regression analyses, stage was the only factor associated with shorter time to the development of BM (P=0.059), whereas age was the only factor associated with survival from the date of recurrence (P=0.027) and with OS (P=0.037). Controlling for age and stage, neither the biological subtype of cancer, the radiation modality nor the site of first recurrence showed any impact on survival. Conclusions: Stage at primary diagnosis correlated with shorter time to the development of BM, while age at diagnosis was associated with shorter survival in BCBM. None of the other clinical factors had influence on survival.

Sign in / Sign up

Export Citation Format

Share Document