目的: 系统评价健康筛查发生高尿酸血症风险预测模型,为其防治和进一步研究提供参考依据。方法: 计算机检索The Cochrane Library、PubMed、Web of Science、维普、万方和知网数据库中有关健康筛查发生高尿酸血症风险预测模型的文献,检索时限为建库至2021年12月10日。由2名研究者独立进行文献筛选和数据提取,并参照预测模型研究的偏倚风险评估工具PROBAST对纳入研究进行偏倚风险和适用性评价,对纳入研究的研究设计、样本量、预测因子、建模方法、验模方法、模型性能等结果进行比较和分析。结果: 本研究共纳入14项研究,12项为开发研究。14项研究的ROC曲线下面积在0.512～0.827,其中有3项模型区分度极佳(＞0.80)。模型纳入前5的预测因子包括:身体质量指数、年龄、甘油三酯、性别、总胆固醇。12项开发研究的适用性较好,但14项研究偏倚风险高,主要原因是变量转化尚未进行规范处理、忽略缺失值、单因素分析筛选预测因子、尚未汇报数据复杂性问题、缺乏模型性能评估。结论: 健康筛查发生高尿酸血症风险预测模型整体偏倚风险较高,研究尚处于发展阶段,需要参照PROBAST评估工具和TRIPOD报告声明开发高质量的模型,以便今后能尽早识别高尿酸血症风险并及时采取预防措施。
Objective: To systematically evaluate the risk prediction models for hyperuricemia in health screening, with the purpose of providing references for the prevention and treatment of hyperuricemia. Methods: The studies on risk prediction model for hyperuricemia in health screening were searched from the Cochrane Library, PubMed, Web of Science, VIP, WanFang, and CNKI databases from database inception to December 10, 2021. Literature screening and data extraction were done by two researchers independently, and the risk of bias and applicability of the included studies were evaluated using the Prediction Model Risk of Bias Assessment Tool (PROBAST). The outcomes of study design, sample size, predictors, modeling techniques, model testing techniques, and model performance of the included studies were analyzed and compared. Results: Fourteen studies of risk prediction model for hyperuricemia in health screening were included in this study, including 12 development studies. There were 14 studies had the AUC under the ROC curve ranging from 0.512 to 0.827, of which 3 models showing good discrimination (＞0.80). Body mass index, age, triglycerides, sex, and total cholesterol were the top 5 predictors included in the models. Even though the 12 development studies had good applicability, the 14 studies were of high risk of bias, which was primarily due to variables undergoing no normative processing, missing values being ignored, predictors being screened out using univariate analysis, complexity of non-reported data, and a lack of model performance evaluation. Conclusion: The study is still in the early stages, and the overall risk of bias of risk prediction models for hyperuricemia in health screening is high. To enable early detection of hyperuricemia risk and take prompt prevention measures, high-quality models with reference to the PROBAST assessment tool and transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD) should be created.
[1] 陈剑锋, 马斌. 老年体检人群高尿酸血症流行现况及影响因素分析研究[J]. 中国全科医学, 2020, 23(5):604-610.
[2] 孙沙沙, 洛佳坤, 马宇飞, 等. 高尿酸血症与高血压关系的研究进展[J]. 中华老年心脑血管病杂志, 2020, 22(3):321-322.
[3] 李林, 朱小霞, 戴宇翔, 等. 中国高尿酸血症相关疾病诊疗多学科专家共识[J]. 中华内科杂志, 2017, 56(3):235-248.
[4] Chen S, Guo XF, Dong SY, et al. Association between the hypertriglyceridemic waist phenotype and hyperuricemia:a cross-sectional study[J]. Clin Rheumatol, 2017, 36(5):1111-1119.
[5] Janssens ACJW, Martens FK. Reflection on modern methods:revisiting the area under the ROC Curve[J]. Int J Epidemiol, 2020, 49(4):1397-1403.
[6] Nezic DG. Assessing the performance of risk prediction models[J]. Eur J Cardiothorac Surg, 2020, 58(2):401-401.
[7] 李琪, 周洁, 覃盛媚, 等. TRIPOD规范解读及对护理研究的启示[J]. 护理研究, 2020, 34(20):3678-3684.
[8] Alba AC, Agoritsas T, Walsh M, et al. Discrimination and calibration of clinical prediction models users' guides to the medical literature[J]. JAMA, 2017, 318(14):1377-1384.
[9] 王俊峰, 章仲恒, 周支瑞, 等. 临床预测模型:模型的验证[J]. 中国循证心血管医学杂志, 2019, 11(2):141-144.
[10] 陈香萍, 张奕, 庄一渝, 等. PROBAST:诊断或预后多因素预测模型研究偏倚风险的评估工具[J]. 中国循证医学杂志, 2020, 20(6):737-744.
[11] Moons KGM, Wolff RF, Riley RD, et al. PROBAST:A tool to assess risk of bias and applicability of prediction model studies:explanation and elaboration[J]. Ann Intern Med, 2019, 170(1):W1-W33.
[12] 杨媛, 周光清, 李宛霖, 等. 不同肥胖指标在高尿酸血症风险预测中的应用价值比较研究[J]. 中国全科医学, 2021:1-9.
[13] Shi JC, Chen XH, Yang Q, et al. A simple prediction model of hyperuricemia for use in a rural setting[J]. Sci Rep, 2021, 11(1):1-10.
[14] Gao YH, Jia SC, Li DH, et al. Prediction model of random forest for the risk of hyperuricemia in a Chinese basic health checkup test[J]. Biosci Rep, 2021, 41(4):BSR20203859.
[15] Chen MQ, Wang HY, Shi WR, et al. Estimate of prevalent hyperuricemia by systemic inflammation response index:results from a rural Chinese population[J]. Postgrad Med, 2020, 133(2):242-249 .
[16] 梁冰倩, 黄志碧, 赖银娟, 等. 随机森林模型和Logistic回归模型在高尿酸血症预测中的应用效果比较[J]. 广西医学, 2020, 42(6):729-733.
[17] Lee SW, Choe EK, Park B. Exploration of machine learning for hyperuricemia prediction models based on basic health checkup tests[J]. J Clin Med, 2019, 8(172):1-10.
[18] Cao J, Wang CX, Zhang G, et al. Incidence and simple prediction model of hyperuricemia for urban Han Chinese adults:a prospective cohort study[J]. Int J Environ Res Public Health, 2017, 14(67):1-7.
[19] Huang ZP, Huang BX, Zhang H, et al. Waist-to-height ratio is a better predictor of hyperuricemia than body mass index and waist circumference in Chinese[J]. Ann Nutr Metab, 2019, 75(3):187-194.
[20] 马明珠, 唐兰, 钟立, 等. 基于列线图的高尿酸血症患病风险个体化预测[J]. 现代医药卫生, 2021, 37(22):3782-3787.
[21] Zeng J, Zhang JG, Li ZY, et al. Prediction model of artificial neural network for the risk of hyperuricemia incorporating dietary risk factors in a Chinese adult study[J]. Food Nutr Res, 2020, 64(3712):1-9.
[22] Chen MQ, Shi WR, Shi CN, et al. Impact of monocyte to high-density lipoprotein ratio on prevalent hyperuricemia:findings from a rural Chinese population[J]. Lipids Health Dis, 2020, 19(48):1-9.
[23] 查晓娟, 江明菲, 李荣, 等. 高尿酸血症的列线图模型建立[J]. 皖南医学院学报, 2019, 38(2):180-182, 186.
[24] Shi WR, Xing LY, Jing L, et al. Usefulness of triglyceride-glucose index for estimating hyperuricemia risk:insights from a general population[J]. Postgrad Med, 2019, 131(5):348-356.
[25] 杨兴华, 陶秋山, 詹思延, 等. 中国台湾35~74岁健检人群高尿酸血症5年发病风险预测模型[J]. 中国预防医学杂志, 2013, 14(9):655-659.
[26] 曹煜隆, 单娇, 龚志忠, 等. 个体预后与诊断预测模型研究报告规范——TRIPOD声明解读[J]. 中国循证医学杂志, 2020, 20(4):492-496.
[27] Reilly JJ, El-Hamdouchi A, Diouf A, et al. Determining the worldwide prevalence of obesity[J]. Lancet, 2018, 391(10132):1773-1774.
[28] Salem CB, Slim R, Fathallah N, et al. Drug-induced hyperuricaemia and gout[J]. Rheumatology (Oxford), 2017, 56(5):679-688.
[29] Ali N, Rahman S, Islam S, et al. The relationship between serum uric acid and lipid profile in Bangladeshi adults[J]. BMC Cardiovasc Disord, 2019, 19(1):1-7.
[30] Siontis GCM, Tzoulaki I, Castaldi PJ, et al. External validation of new risk prediction models is infrequent and reveals worse prognostic discrimination[J]. J Clin Epidemiol 2015, 68(1):25-34.
