Penttala Miika, Sorsa Timo, Thomas Julie Toby, Grigoriadis Andreas, Sakellari Dimitra, Sahni Vaibhav, Gupta Shipra, Pärnänen Pirjo, Pätilä Tommi, Räisänen Ismo T
Department of Oral and Maxillofacial Diseases, Head and Neck Center, University of Helsinki and Helsinki University Hospital, 00290 Helsinki, Finland.
Division of Oral Diseases, Department of Dental Medicine, Karolinska Institutet, 171 77 Stockholm, Sweden.
Diagnostics (Basel). 2025 Jun 2;15(11):1411. doi: 10.3390/diagnostics15111411.
We propose a framework for determining the stage of periodontitis in a personalized medicine context, building on our previously developed model for periodontitis detection. In this study, we improved the earlier model by incorporating additional components to form a comprehensive system for identifying both the presence and stage of periodontitis. Central to the home-use application is an active-matrix metalloproteinase-8 (aMMP-8) mouth rinse test (cut-off: 20 ng/mL), integrated with software delivered via a mobile application. First, using all the data, we modeled a single polynomial function to distinguish healthy and stage I periodontitis patients from stage II and III patients. Second, we used an already published periodontitis detection function to separate stage I patients from healthy patients. Third, one more function was created that divided stage II and III patients from each other. All functions were modeled by multiple logistic regression analysis from the patient data, which consisted of 149 adult patients visiting dental offices in Thessaloniki, Greece. The complete model demonstrated a sensitivity of 95.8% (95% CI: 92.1-99.4%) and a specificity of 71.0% (95% CI: 55.0-86.9%) for detecting periodontitis. Among those identified with periodontitis, the correct stage was determined in 61.1% of cases, with stage-specific accuracies of 64.3% for stage I, 60.5% for stage II, and 60.9% for stage III. All testing was performed on patient data with which the complete model was formed. The results of this study showed that with sufficient data and using multiple logistic regression analysis, a model can be created to simultaneously identify the presence and stage of periodontitis. Overall, in the complete model generated, a mouth rinse aMMP-8 test result with a cut-off value of 20 ng/mL, Visible Plaque Index (VPI) and information of patient's teeth number present were found to be important factors to determine the stage of periodontitis in a personalized medicine manner for everyone to use.
我们基于之前开发的牙周炎检测模型，提出了一个在个性化医疗背景下确定牙周炎阶段的框架。在本研究中，我们通过纳入额外的组件改进了早期模型，以形成一个用于识别牙周炎的存在和阶段的综合系统。家庭使用应用的核心是一种活性基质金属蛋白酶-8（aMMP-8）漱口水测试（临界值：20 ng/mL），并与通过移动应用程序提供的软件相结合。首先，使用所有数据，我们构建了一个单一多项式函数，以区分健康患者和I期牙周炎患者与II期和III期患者。其次，我们使用已发表的牙周炎检测函数将I期患者与健康患者区分开来。第三，创建了另一个函数，将II期和III期患者相互区分开来。所有函数均通过对患者数据进行多元逻辑回归分析构建，这些数据来自149名访问希腊塞萨洛尼基牙科诊所的成年患者。完整模型在检测牙周炎方面表现出95.8%的灵敏度（95%置信区间：92.1-99.4%）和71.0%的特异度（95%置信区间：55.0-86.9%）。在那些被诊断为牙周炎的患者中，正确分期在61.1%的病例中得以确定，I期的分期特异性准确率为64.3%，II期为60.5%，III期为60.9%。所有测试均在用于构建完整模型的患者数据上进行。本研究结果表明，通过足够的数据并使用多元逻辑回归分析，可以创建一个模型来同时识别牙周炎的存在和阶段。总体而言，在生成的完整模型中，临界值为20 ng/mL的漱口水aMMP-8测试结果、可见菌斑指数（VPI）以及患者现存牙齿数量信息被发现是以个性化医疗方式确定每个人牙周炎阶段的重要因素。
