📄 medicine

Risk factor analysis and predictive model construction and evaluation of fractional uric acid excretion in Chinese inpatients with Type 2 Diabetes

This study developed and validated a practical linear regression model using sex, serum uric acid, serum creatinine, and fasting plasma glucose to accurately predict fractional uric acid excretion in Chinese inpatients with Type 2 Diabetes, offering a convenient alternative to complex 24-hour urine collection methods.

Original authors: Lin Zhuo, Xiantong Zou, Wenjia Yang, Ling Chen, Linong Ji, Lingli Zhou, Xueyao Han

Published 2026-07-22

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Lin Zhuo, Xiantong Zou, Wenjia Yang, Ling Chen, Linong Ji, Lingli Zhou, Xueyao Han

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine your body is a bustling city, and your kidneys are the sanitation department, working around the clock to filter out waste and keep the streets clean. One of the most important things they need to manage is a substance called uric acid. Think of uric acid as a type of trash that your body produces when it breaks down certain foods. Usually, the city's sanitation crew (your kidneys) does a great job, keeping just the right amount of trash in the bloodstream and flushing the rest out through your urine.

But sometimes, the system gets a bit glitchy. In people with Type 2 Diabetes, the rules of the game change. Their bodies might produce too much trash, or the sanitation crew might get confused and start flushing out too much (or too little) of it. This balance is measured by something scientists call the "Fractional Excretion of Uric Acid" (FEUA). It's basically a score that tells us what percentage of the uric acid in the blood the kidneys decide to throw away. If the score is too high or too low, it can be a warning sign that the kidneys are under stress or that the city is heading toward a traffic jam of waste, which can lead to serious damage over time.

The problem is, checking this score is a pain. To get an accurate reading, a patient has to stop taking certain medications for two weeks, eat a very strict diet, and collect all their urine for a full 24 hours. It's like asking a busy city to stop all traffic for a week just to count the cars. It's accurate, but it's not practical for everyday life or for looking at huge databases of medical records. So, scientists have been hunting for a shortcut: a way to guess this score using simple blood tests that doctors already do every day.

The Hunt for a Crystal Ball

This is where the team from Peking University People's Hospital steps in. They decided to build a "crystal ball"—a mathematical model—that could predict a patient's FEUA score using only four pieces of information that are already sitting in their medical chart: their sex, their blood uric acid level, their blood creatinine level (a measure of kidney function), and their fasting blood sugar.

They gathered data from over 2,300 patients with Type 2 Diabetes who were hospitalized in Beijing. They split the group into two teams: a "training team" of about 1,510 people to teach the model how to think, and a "test team" of about 791 people (from different years) to see if the model could actually pass the exam without cheating.

The Magic Formula

After crunching the numbers, the researchers found that these four simple factors were the secret sauce. They built a formula that looks a bit like a math homework problem, but here's the gist:

Sex: Being male or female changes the baseline.
Blood Uric Acid (SUA): If your blood is full of uric acid, your kidneys tend to hold onto it more (lowering the FEUA score).
Blood Creatinine (SCr): If your kidneys are working harder or are under stress, the score goes up.
Blood Sugar (FPG): If your blood sugar is high, it seems to push the kidneys to flush out more uric acid.

When they plugged these numbers into their formula, the results were surprisingly accurate. The model's predictions matched the actual, hard-to-measure urine tests about 79% of the time in terms of correlation. Even better, when they tried to use the model to spot patients with "high" FEUA (defined as 11% or more), the model was right about 95% of the time. It's like having a weather app that predicts rain with 95% accuracy, saving you from getting soaked without needing to stand outside for a whole day to check.

Why This Matters

The researchers didn't just stop at the math; they tested their model on patients from 2017 and 2025 to make sure it wasn't just a fluke that worked for one specific group. It held up! The model remained stable and accurate, even when the patients were different or the data came from a different time period.

The big takeaway here is that we don't always need a 24-hour urine collection to understand how a diabetic patient's kidneys are handling uric acid. By using a simple equation based on common blood tests, doctors can now get a reliable estimate of this "sanitation score." This could help identify patients who are at risk for kidney damage earlier, allowing for better care without the hassle of strict diets and long urine collections. It turns a complex, time-consuming puzzle into a quick, easy calculation that fits right into a routine doctor's visit.

Technical Summary: Risk Factor Analysis and Predictive Model Construction for Fractional Uric Acid Excretion in Chinese Inpatients with Type 2 Diabetes

Problem Statement
Fractional excretion of uric acid (FEUA) is increasingly recognized as a critical risk factor for nephropathy and a precise indicator of uric acid excretion (UAE). However, the accurate measurement of FEUA is clinically arduous, requiring a 24-hour urine collection following a strict 2-week medication washout period and a 5-day purine-restricted diet. These requirements limit its application in routine clinical practice and hinder large-scale research utilizing public databases that often lack specific urinary variables. Furthermore, no existing model currently exists to predict FEUA levels based on easily accessible clinical data. This study addresses the need for a simple, reliable method to estimate FEUA in patients with Type 2 Diabetes Mellitus (T2DM), a population prone to dysregulated uric acid metabolism and kidney damage.

Methodology
This retrospective, single-center study was conducted at Peking University People's Hospital involving 2,301 hospitalized patients with T2DM.

Data Source: Data were extracted from the Hospital Information System (HIS) for patients admitted between 2017 and 2025.
Cohort Division:
- Training/Internal Validation Set (n=1,510): Patients admitted between 2018 and 2024.
- External Validation Set (n=791): Patients admitted in 2017 and 2025 (temporal validation) to test robustness across different time periods.
Inclusion/Exclusion: Patients were excluded if they had malignant tumors, specific endocrine or renal diseases, were pregnant, or were taking medications known to affect uric acid metabolism (e.g., diuretics, SGLT2 inhibitors, ARBs, ACEIs, uric acid-lowering therapies). Extreme values for serum creatinine (SCr), fasting plasma glucose (FPG), or serum uric acid (SUA) were also excluded.
Statistical Approach:
- Variable Selection: Univariate analysis identified significant factors. A multivariate backward-stepwise linear regression was performed.
- Transformation: FEUA, $\beta_2$ -microglobulin, and Urine Albumin-to-Creatinine Ratio (UACR) underwent natural logarithmic transformation to correct for skewness.
- Bias Correction: Duan's Smearing Factor was applied to correct for re-transformation bias when converting log-predicted values back to the original scale.
- Validation: The model underwent internal validation (5-fold, 10-fold, and leave-one-out cross-validation) and external temporal validation. Performance was assessed using Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Spearman correlation, Receiver Operating Characteristic (ROC) curves, calibration plots, and Decision Curve Analysis (DCA).

Key Contributions and Model Construction
The study successfully developed a linear regression model to predict FEUA using four routinely available serum and demographic variables: Sex, Serum Uric Acid (SUA), Serum Creatinine (SCr), and Fasting Plasma Glucose (FPG).

The final predictive formula for the natural logarithm of FEUA is:
$\ln(\text{FEUA}) = 1.9408 + 0.1462 \times \text{Gender} - 0.0031 \times \text{SUA} + 0.00112 \times \text{SCr} + 0.0068 \times \text{FPG}$
(Where Gender is coded as Male=1, Female=2; SUA in $\mu$ mol/L; SCr in $\mu$ mol/L; FPG in mmol/L)

The predicted FEUA is calculated as:
$\text{Predicted FEUA} = \exp(\ln(\text{FEUA})) \times 1.0265$
(1.0265 is the Duan's Smearing Factor)

Results

Internal Validation: The model demonstrated strong agreement between observed and predicted FEUA.
- Correlation: Spearman correlation coefficient was 0.7870 ( $P < 0.001$ ).
- Error Metrics: RMSE was 1.690%, and the mean difference between observed and predicted values was negligible (-0.005%).
- Classification Accuracy: For predicting elevated FEUA ( $\ge 11\%$ ), the model achieved an overall accuracy of 94.70%. The Area Under the Curve (AUC) was 0.9552 (95% CI: 0.9413–0.9691). At a cut-point of 8.0257%, the sensitivity was 95.41% and specificity was 82.73%.
- Stability: Cross-validation indices (5-fold, 10-fold, LOO) remained stable, with RMSE values ranging between 1.6898 and 1.7081. Subgroup analyses (based on renal function and glycemic status) showed consistent performance with ROC values > 0.949.
External Validation: The model maintained robust performance in the independent temporal dataset (2017 and 2025).
- Correlation: Spearman correlation was 0.7451 ( $P < 0.001$ ).
- Classification: The AUC for predicting elevated FEUA was 0.8946 (95% CI: 0.8545–0.9347), with an overall accuracy of 94.18%.
- Calibration: Calibration curves and decision curves confirmed the model's clinical utility and agreement between predicted probabilities and actual outcomes.

Significance and Claims
The authors claim that this study provides the first effective and practical predictive model for FEUA in T2DM inpatients using only four common clinical indexes. The significance of this work lies in:

Clinical Utility: It offers a non-invasive, convenient alternative to the cumbersome 24-hour urine collection and washout protocols, aiding physicians in identifying patients with elevated FEUA (a risk factor for kidney damage) and guiding preventive or therapeutic measures for hyperuricemia and diabetic kidney disease.
Research Expansion: By enabling the estimation of FEUA from standard serum data, the model facilitates the utilization of large-scale public databases that previously lacked urinary variables, potentially expanding the scope of research into uric acid metabolism and kidney disease.
Mechanistic Insight: The study reinforces the relationships between FEUA and its predictors (e.g., the negative association with SUA and positive association with SCr and FPG), discussing potential physiological mechanisms such as oxidative stress scavenging, renal transporter regulation, and hormonal influences.

The authors maintain modesty regarding the study's limitations, acknowledging its retrospective, single-center, and cross-sectional nature, and noting that future multicenter external validations are required to confirm generalizability to wider populations.

More like this