Original ArticleIntegrated prediction and decision models are valuable in informing personalized decision making
Introduction
In the last years, personalized health care has gained increasing attention. Personalized health care may be thought of as the tailoring of medical treatment to individual characteristics, needs, and preferences of a patient during all stages of care [1]. The evidence-based medicine approach historically focuses on randomized clinical trials. In such comparative studies, which can include economic evaluations, often little attention is paid to patient heterogeneity. As a result, reimbursement decisions are generally based on average (cost-)effectiveness of the intervention in a selected population [2]. This is also reflected in clinical guidelines, in which recommendations are often based on studies that pertain to the average population. They often do not guide in dealing with individual patient characteristics [3], [4].
In recent years, evidence-based medicine research increasingly takes into account patient heterogeneity [5]. For example, prediction models are developed to estimate a probability of a certain outcome in an individual, given his or her personal or disease characteristics [6], [7]. These probability estimates can guide care providers as well as the individuals themselves in deciding upon further management [8]. Such models are becoming increasingly abundant in medical literature. Often multiple outcomes (e.g., survival and side effects) are relevant for one patient, and therefore, multiple prediction models are relevant for such an individual patient. This implies a trade-off between different outcomes, which makes the use of these prediction models for clinical decision making difficult.
This is, for example, the case in early-stage oral cavity squamous cell carcinoma (OCSCC) patients. If they have no clinically detected lymph node metastases in the neck (cN0), the decision whether to remove the lymph nodes or not implies a trade-off between survival, quality of life, and costs. In a previous study, we used decision analytical modeling to inform this trade-off for an average cohort of patients [9]. Sensitivity analysis showed that predictions for individual patients, based on their characteristics, such as a patient's probability of occult metastases or his quality of life after surgery, influence which strategy is deemed optimal. Therefore, it seems worthwhile to include patient characteristics into this decision model to weigh risks, benefits, and costs for individual patients, to allow for more personalized evidence-based treatment decisions.
The aim of this study was to show how prediction models can be incorporated into decision models and to assess the value of personalized decisions based on these prediction models using the case of the management of the neck in early-stage OCSCC as an example.
Section snippets
The case
In about 30% of patients with early-stage (T1-2) OCSCC, who have no clinically detected regional lymph node metastases (cN0), occult metastases are present in the lymph nodes of the neck [10]. Therefore, the decision needs to be made whether to remove those lymph nodes at risk for involvement by metastatic cancer or to follow a watchful waiting (WW) strategy. Removing the lymph nodes at risk by performing a neck dissection (ND) increases prognosis [11]. However, this invasive procedure is
Results
The results of the three approaches are presented in Table 2. The SLN strategy was the most cost-effective strategy using population averages for the parameters. On average, the SLN strategy resulted in 4.9158 QALYs and costs €8,675 per patient, which represent the results of the population-based approach.
In the perfectly predicted approach, each subgroup of patients will receive the treatment under the assumption that probabilities in the different strategies could be perfectly predicted. With
Discussion
This study showed how prediction models can be incorporated into decision models and assessed the value of personalized decisions regarding the management of the neck in early-stage cN0 OCSCC. With perfect predictions, only 9% of the patients would receive END. Patients for which END was most cost-effective all had occult metastases. However, for some of the patients with occult metastases, it was not cost-effective to perform END. These patients would suffer from shoulder complaints after END
References (36)
U.S. Food and drug administration: personalized medicine
Subgroups and heterogeneity in cost-effectiveness analysis
Pharmacoeconomics
(2008)- et al.
Scientific American: how clinical guidelines can fail both doctors and patients
(2013) - et al.
Clinical practice guidelines and patient decision aids. An inevitable relationship
J Clin Epidemiol
(2012) - et al.
Prognosis research strategy (PROGRESS) 4: stratified medicine research
BMJ
(2013) - et al.
Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement
BMJ
(2015) - et al.
Prognosis research strategy (PROGRESS) 3: prognostic model research
PLoS Med
(2013) - et al.
Diagnostic and prognostic prediction models
J Thromb Haemost
(2013) - et al.
Management of the N0 neck in early stage oral squamous cell cancer: a modeling study of the cost-effectiveness
Oral Oncol
(2013) - et al.
Improved survival with elective neck dissection in thick early-stage oral squamous cell carcinoma
Head Neck
(2012)
Elective neck dissection in early-stage oral squamous cell carcinoma--does it influence recurrence and survival?
Head Neck
Impact of shoulder complaints after neck dissection on shoulder disability and quality of life
Otolaryngol Head Neck Surg
Evaluation of clinical and histomorphological parameters as potential predictors of occult metastases in sentinel lymph nodes of early squamous cell carcinoma of the oral cavity
Ann Surg Oncol
Risk factors of metastasis in oral squamous cell carcinomas
Oncology
Established prognostic variables in N0 oral carcinoma
Otolaryngol Head Neck Surg
Multivariate analysis of occult lymph node metastasis as a prognostic indicator for patients with squamous cell carcinoma of the oral cavity
Cancer
Micrometastasis in oral/oropharyngeal squamous cell carcinoma: incidence, histopathological features and clinical implications
Br J Oral Maxillofac Surg
The prognostic significance of histological features in oral squamous cell carcinoma
J Oral Pathol Med
Cited by (15)
A critical view of the revolution of precision medicine: genetics, epigenetics, sex, and gender
2023, Sex, Gender, and Epigenetics: From Molecule to BedsideThe potential use of big data in oncology
2019, Oral OncologyCitation Excerpt :These integrated multi-omics data will for example provide more comprehensive insight into biological behavior and mechanisms that underlie growth patter, metastatic potential as well as response to (targeted) treatment of HNSCC. From the perspective of turning our current understanding and available data into actionable insights that can be used to improve treatment outcome, personalized medicine is absolutely dependent on big data [15]. The amount of data available for the biomedical community exponentially increases, especially with advancing technologies generating terabytes of data, notably in sequencing and imaging.
Decision analytic modeling was useful to assess the impact of a prediction model on health outcomes before a randomized trial
2019, Journal of Clinical EpidemiologyCitation Excerpt :Few other DAM-based assessments have been previously performed that assess the potential impact of prediction models before an empirical impact study has been executed. One study assessed the value of a prediction model for predicting shoulder pain in patients with early-stage oral cavity squamous cell carcinoma after surgical removal of lymph nodes [48]. Although the analysis did focus on specific scenarios regarding the accuracy of predictions, compliance or additional clinical expertise on top of the prediction model were not evaluated.
Confounding obscures our view, effect modification is part of reality
2019, Journal of Clinical EpidemiologyImpact of big data on cancer care and research
2023, Big Data in Oncology: Impact, Challenges, and Risk AssessmentImplications of big data in the targeted therapy of cancer
2023, Big Data in Oncology: Impact, Challenges, and Risk Assessment
Conflict of interest: None.