[https://doi.org/10.5061/dryad.4mw6m90m8

1. Dataset Overview

This dataset was collected as part of a meta-analysis to evaluate the diagnostic accuracy of various nutritional screening tools used in patients with digestive system tumors. The primary objective was to assess the performance of tools such as the Patient-Generated Subjective Global Assessment (PG-SGA) and the Nutritional Risk Screening 2002 (NRS-2002) in diagnosing malnutrition and assessing nutritional risk in cancer patients.

The dataset includes:

• Extracted diagnostic performance data: true positives (TP), false positives (FP), true negatives (TN), and false negatives (FN).
• Risk of bias assessments using the QUADAS-2 tool.
• Bayesian latent class analysis results to further evaluate the clinical utility of these tools.
• Stata commands used for analysis to ensure reproducibility.

The data were collected through a systematic review of studies that met predefined inclusion criteria. All information was extracted and verified independently by two reviewers, with discrepancies resolved by a third reviewer.

2. Files and Variables

2.1 Data_Extraction_Form.csv

Description: This file contains raw data extracted from studies, including diagnostic outcomes for various nutritional screening tools.

Variables:

• Study ID: A unique identifier for each study.
• Author: The first author of the study.
• Year: The year the study was published.
• Country: The country where the study was conducted.
• Sample Size: The number of participants in the study.
• Male-Female Ratio: The ratio of male to female participants in the study.
• Age: The average age of participants included in the study.
• Diagnose: Clinical diagnosis of patients in each study.
• Reference Standard: The gold standard against which the screening tool was compared.
• Index Test: The nutritional screening tool being evaluated (e.g., PG-SGA, NRS-2002).
• TP (True Positives): Patients correctly identified by the screening tool as malnourished or at risk.
• FP (False Positives): Patients incorrectly identified as malnourished or at risk.
• FN (False Negatives): Patients incorrectly identified as not malnourished or at risk.
• TN (True Negatives): Patients correctly identified as not malnourished or at risk.

2.2 included_excluded_studies.csv

Description: This file lists all studies considered for inclusion in the meta-analysis and provides reasons for inclusion or exclusion.

Variables:

• Study ID: Unique identifier for each study.
• Author (Year): First author and year of publication.
• Included/Excluded: Indicates whether the study was included in the final analysis.
• Reason for Exclusion: If excluded, the reason for exclusion (e.g., data incomplete, not a diagnostic study).

2.3 Result_of_meta-analysis_and_Bay.csv

Description:
This file contains the final results of the meta-analysis and Bayesian evaluation for various nutritional screening tools used to assess malnutrition and nutritional risks in patients with digestive system tumors.

Variables:

• Normative: The names of nutritional screening tools being evaluated (e.g., MUST, MST, NRS-2002, MNA-SF, NRI, PG-SGA).
• Sensitivity (95% CI): The sensitivity (true positive rate) of each nutritional screening tool, with 95% confidence intervals.
  • Format: Numerical values ranging from 0 to 1, with corresponding confidence intervals in parentheses.
  • Example: 0.911 (95% CI: 0.866–0.942) for PG-SGA.
• Specificity (95% CI): The specificity (true negative rate) of each nutritional screening tool, with 95% confidence intervals.
  • Format: Numerical values ranging from 0 to 1, with corresponding confidence intervals in parentheses.
  • Example: 0.805 (95% CI: 0.674–0.891) for PG-SGA.
• DOR (95% CI): The diagnostic odds ratio, which summarizes the overall diagnostic accuracy of the tool, with 95% confidence intervals.
  • Format: Numerical value greater than 0, with confidence intervals in parentheses.
  • Example: 41.987 (95% CI: 18.870–93.424) for PG-SGA.
• LR+ (95% CI): Positive likelihood ratio, indicating how much the odds of the disease increase when a test is positive, with 95% confidence intervals.
  • Format: Numerical value greater than 0, with confidence intervals in parentheses.
  • Example: 4.66 (95% CI: 2.680–8.108) for PG-SGA.
• LR- (95% CI): Negative likelihood ratio, indicating how much the odds of the disease decrease when a test is negative, with 95% confidence intervals.
  • Format: Numerical value between 0 and 1, with confidence intervals in parentheses.
  • Example: 0.111 (95% CI: 0.072–0.171) for PG-SGA.
• 1/LR- (95% CI): The inverse of the negative likelihood ratio, representing how unlikely the disease is when a test is negative, with 95% confidence intervals.
  • Format: Numerical value greater than 1, with confidence intervals in parentheses.
  • Example: 9.007 (95% CI: 5.861–13.842) for PG-SGA.

2.4 risk_of_bias_assessment.xml

Description: This file provides the risk of bias assessment for each study, evaluated using the QUADAS-2 tool.

Variables:

• Study ID: Unique identifier for each study.
• Patient Selection: Risk of bias in patient selection (Low/Unclear/High).
• Index Test: Risk of bias for the index test (Low/Unclear/High).
• Reference Standard: Risk of bias for the reference standard (Low/Unclear/High).
• Flow and Timing: Risk of bias in study flow and timing (Low/Unclear/High).
• Applicability Concerns: Concerns about the applicability of results to the population of interest (Low/Unclear/High).

3. Code/Software

stata_commands_for_diagnostic_analysis.xml

Description: This file contains Stata commands used to perform diagnostic meta-analysis and ensure reproducibility.

Commands:

• metandi tp fp fn tn: Computes summary diagnostic performance statistics.
• metandi tp fp fn tn, plot nob noh: Plots the HSROC curve.
• midas tp fp fn tn, pubbias: Tests for publication bias.
• midas tp fp fn tn, fagan: Generates a Fagan plot for Bayesian analysis.

4. Explanation of Placeholders

• null: Indicates that data for a specific variable were not collected or are unavailable.
• n/a: Indicates that the variable does not apply to the specific study context.
• NA: Indicates either missing data or not applicable data; the exact meaning depends on the context and is described in the variable section of the dataset.

5. Access Information

This dataset has not been made available in other public repositories and is exclusively accessible through this submission. Future updates or alternative repositories will be announced when applicable.

6. Data Sources

The dataset was derived from published studies included in systematic reviews and meta-analyses of nutritional screening tools for digestive system cancers. Data were extracted from:

• Databases: PubMed, Embase, Cochrane Controlled Clinical Trials Database, China Knowledge Network, Wanfang Database, and Wikipedia Database.
• Screening Tools: Evaluations focused on tools like the PG-SGA and NRS-2002 to assess malnutrition and nutritional risk in gastrointestinal cancer patients.