Prediction models for fear of cancer recurrence in adults with cancer: a systematic review

TL;DR

This paper reviews prediction models for fear of cancer recurrence in cancer patients, finding that while models show acceptable performance, most have methodological flaws and need improvement.

Contribution

The study systematically evaluates the risk of bias and applicability of FCR prediction models, highlighting gaps in validation and methodology.

Findings

7 studies with 18 models were included, showing discrimination ranging from 0.660 to 0.996.

All studies had high risk of bias due to poor data handling and limited validation.

Common predictors included age, social support, income, and fatigue.

Abstract

Numerous studies have developed or validated prediction models to estimate the likelihood of Fear of Cancer Recurrence (FCR) among patients with cancer. The quality of these models and evaluations of their applicability to clinical practice and future research remain unclear. This study systematically evaluated the risk of bias and applicability of prediction models for FCR in oncology populations.

We searched PubMed, Embase, Web of Science, the Cochrane Library, Cumulative Index to Nursing and Allied Health Literature (CINAHL), China National Knowledge Infrastructure (CNKI), China Science and Technology Journal Database (VIP), Wanfang Data, and Chinese Biomedical Literature Database (CBM) from inception to August 1, 2025. Two reviewers independently screened studies and extracted data. We used the Prediction model Risk Of Bias ASsessment Tool (PROBAST) checklist to assess risk of bias and applicability.

We identified 702 records and ultimately included 7 studies encompassing 18 models. Most were published between 2022 and 2025. Sample sizes in the included studies ranged from 200 to 918, and the reported discrimination in model development cohorts ranged from 0.660 to 0.996. Applicability was judged to be good across all studies. However, all studies exhibited a high risk of bias, mainly due to suboptimal data sources, inadequate handling of predictors and missing data, and limited model validation. Among the included studies, the prevalence of FCR ranged from 47.7% to 63.4%. The most frequent predictors were age, social support, household/per-capita monthly income, occupation or employment status, and fatigue.

Prediction modeling for FCR in patients with cancer remains in an early stage, with both shared and heterogeneous predictors. Although overall model performance appears acceptable, most studies have methodological shortcomings, and few models have been validated. Future studies should design, conduct, and report models in accordance with the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) guideline. In addition, studies with larger sample sizes and multicenter external validation are needed to improve model robustness.

Figures