TY - JOUR

T1 - Systematic mapping review of statistical methods applied to the relationships between cancer diagnosis and geographical level factors in UK

AU - Mendes, Jessica Andretta

AU - Keegan, Thomas

AU - Jones, Lisa

AU - Atkinson, Peter M

AU - Sedda, Luigi

PY - 2025/7/6

Y1 - 2025/7/6

N2 - ObjectivesWe examined studies that analysed the spatial association of cancers with demographic, environmental, behavioural and/or socioeconomic factors and the statistical methods applied.DesignSystematic mapping review.Data sourcesWeb of Science (SSCI) (search on 28 July 2022), MEDLINE, SocINDEX and CINAHL (search on 4 August 2022), additional searches included grey literature.Eligibility criteria for selecting studies(1) Focused on the constituent countries of the UK (England, Wales, Scotland and Northern Ireland) and its major regions (eg, the North West); (2) compared cancer(s) outcomes with demographic, environmental, behavioural and socioeconomic characteristics by applying methods to identify their spatial association; (3) reported cancer prevalence, incidence rates, relative risk or ORs for a risk factor or to an average level of cancer.Data extraction and synthesisA standardised data extraction form was developed and for all studies, core data were extracted including bibliographic information, study design, geographical factors analysed, data aggregation level, methods applied and main findings. We described and synthesised the characteristics of the studies using summary tables, charts and graphs.Results52 studies were included covering a variety of objectives and geographical scales. These studies considered different types of cancer, with the most common cancer types analysed being blood and lymphoid cell cancers. The most common methods used to assess the association between cancers and geographical level factors were regression analyses, with the majority being Poisson regression, then logistic and linear regression. Studies were usually conducted at ward and local authority level, or by exact point location when distances from putative risk sources were considered. The results were usually presented in plots or as tables, instead of maps.ConclusionOur results highlight the lack of consideration of spatially explicit models in the analysed studies, with the risk of having failed the assumption of independence in the data.Prospero registration numberCRD42022349165.

AB - ObjectivesWe examined studies that analysed the spatial association of cancers with demographic, environmental, behavioural and/or socioeconomic factors and the statistical methods applied.DesignSystematic mapping review.Data sourcesWeb of Science (SSCI) (search on 28 July 2022), MEDLINE, SocINDEX and CINAHL (search on 4 August 2022), additional searches included grey literature.Eligibility criteria for selecting studies(1) Focused on the constituent countries of the UK (England, Wales, Scotland and Northern Ireland) and its major regions (eg, the North West); (2) compared cancer(s) outcomes with demographic, environmental, behavioural and socioeconomic characteristics by applying methods to identify their spatial association; (3) reported cancer prevalence, incidence rates, relative risk or ORs for a risk factor or to an average level of cancer.Data extraction and synthesisA standardised data extraction form was developed and for all studies, core data were extracted including bibliographic information, study design, geographical factors analysed, data aggregation level, methods applied and main findings. We described and synthesised the characteristics of the studies using summary tables, charts and graphs.Results52 studies were included covering a variety of objectives and geographical scales. These studies considered different types of cancer, with the most common cancer types analysed being blood and lymphoid cell cancers. The most common methods used to assess the association between cancers and geographical level factors were regression analyses, with the majority being Poisson regression, then logistic and linear regression. Studies were usually conducted at ward and local authority level, or by exact point location when distances from putative risk sources were considered. The results were usually presented in plots or as tables, instead of maps.ConclusionOur results highlight the lack of consideration of spatially explicit models in the analysed studies, with the risk of having failed the assumption of independence in the data.Prospero registration numberCRD42022349165.

U2 - 10.1136/bmjopen-2024-098379

DO - 10.1136/bmjopen-2024-098379

M3 - Journal article

C2 - 40623758

VL - 15

JO - BMJ Open

JF - BMJ Open

SN - 2044-6055

IS - 7

M1 - e098379

ER -