In this blog, Dr Hannah Harrison, NIHR Systematic Review Methods Fellow at the Prevention Group, introduces research on approaches to screening for kidney cancer.
Hannah and her colleagues looked at the usefulness of mathematical models predicting the likelihood of an individual developing kidney cancer in a recently conducted systematic review, published on 14 July 2020 in European Urology Focus. They found several suitable models, using a range of risk factors (such as age and smoking) to predict the risk for individuals.
Kidney cancer is the 15th most common cancer worldwide, with a rising incidence, especially in developed countries. Across Europe, kidney cancer is responsible for 50,000 annual deaths. Diagnosing kidney cancer at an early stage has been shown to significantly reduce mortality, however, more than 20% of individuals diagnosed with kidney cancer have metastatic disease.
A national screening programme could help increase rates of early detection; however, the relatively low prevalence of kidney cancer in the general population makes such an approach unlikely. The potential benefits would not be balanced by the potential harms of such an approach.
One possible solution is to stratify the population into risk categories and target a screening program only at those who are at the highest risk. This approach could reduce the burden of kidney cancer for both patients and the healthcare system. Risk stratification requires a model that can accurately predict the risk of an individual developing kidney cancer.
We conducted a systematic review to identify and evaluate published risk prediction models for kidney cancer. In order to identify suitable models, we carried out a literature search in MEDLINE and EMBASE that identified around 15,000 references. We screened these, according to PRISMA guidelines, in order to identify the most relevant publications. To be included in the review a model had to be applicable to the general population and use two or more risk factors. Data extraction and quality assessment was carried out for the final list of included models.
The systematic review identified 62 models that could be used to predict an individual’s risk of kidney cancer. The risk factors used ranged from simple demographic and lifestyle risk factors (for example, sex and smoking status), to biomarker and genetic risk factors. Only 11 of the models identified had published performance measures, which show how well the model distinguishes individuals who develop kidney cancer from the rest of the population. Many of the models were shown to have a high risk of bias, often due to development in small case-control studies. In order to assess their suitability for risk-stratified screening, the performance of the models should be tested in a large population cohort; this has not yet been done for the vast majority of the models.
The authors recommend that this work should initially focus on the models using demographic and lifestyle risk factors that are easily accessible from medical records or self-assessment questionnaires, as these would be the easiest to implement in practice. This systematic review also identified promising models using biomarker and genetic risk factors. However, these areas of research, in particular the use of genetic risk factors, are largely unexplored.
H Harrison, R Thompson, Z Lin, S H Rossi, G D Stewart, S J Griffin, J A Usher-Smith, Risk Prediction Models for Kidney Cancer: A Systematic Review European Urology Focus, 14 July 2020
This blog was written by Dr Hannah Harrison.
About the Prevention Group.
Contact Lucy Lloyd, Communications, Primary Care Unit
Image arranged by Eliza Monnington