br Introduction br The English National Health Service Breas
The English National Health Service Breast Screening Programme (NHSBSP) started in 1987 following the Forrest report.1 The programme currently undertakes two-view mammography on women aged 50e70 years at intervals of 3 years. Following the AgeX trial, invitations now occur from age range 47e73 years as the trial involves random-ising women aged 47e49 and 71e73 to receive an invitation to screening.2
There has been controversy for many years related to the optimum balance between benefits and harms in screening for breast cancer.3 In the UK, invasive cancer detection rate targets are informed by the Swedish Two-County rando-mised controlled trial, which detected nearly all cancers as invasive and are age standardised.4,5 In Europe, targets are set at three times the underlying incidence rate for first screen and 1.5 for subsequent screens.6
Targets for recall rates range from 2% in Holland7 to a recommended upper threshold of 12% in the United States.8 Europe and the NHSBSP set separate targets for recall rate. At prevalent (first) screens, <5% in Europe with a minimum standard of <7% (but <7% and <10% respectively in the NHSBSP). At incident (subsequent) screens the European
Percutaneous needle biopsy has essentially replaced open surgical biopsy for the diagnosis of both palpable and impalpable abnormalities.10 The NHSBSP has targets to reduce the rate of women referred for benign surgical bi-opsy, repeat operation rates, and targets to avoid early surveillance/follow-up9; however, no target is set for needle (cytology/core) biopsy rates. There has been some upward drift in needle biopsy rates over the years as units try to increase cancer detection and ensure diagnosis of malig-nancy prior to referral for surgery. In addition, there is marked variation in unit needle biopsy rates by a factor of three times at prevalent screens and a factor of about two times at incident screens.11
The aim of the present study was to examine the asso-ciation between cancer detection and needle biopsy rates and provide information that could potentially enable the setting of evidence-based needle biopsy rate ranges.
Materials and methods
Information on cancer detection, recall, and needle biopsy rates were taken from the national Korner (KC62) returns sent to Public Health England (PHE) annually and published by NHS Digital.10 Data in this Dalbavancin paper are from 80 screening units over the 7 screening years 2009/10 to 2015/16. The KC62 datasets are based on women as the denominator and the needle biopsy rate is interpreted as the number of screened women who have at least one needle biopsy. There is no information on repeat or multiple biopsies in these returns. Similarly the KC62 returns record no information on multiple or bilateral cancers. The models presented here, mostly look at the associa-tion between needle biopsy rates (from referral to cytology/ core biopsy) and cancer detection rate where cancer detection rates are based on all cancers. In practice, 97% of all cancers are detected following needle biopsies, the other 3% of cancers are almost all detected via open biopsy. Alternative models can be produced looking at the associ-ation between cancers detected only from needle biopsy against needle biopsy rates or the association between any biopsy (including open biopsy) and cancer detection rate. The rationale for the choice of models is discussed later, although gonorrhea should be noted that in practice the model re-sults are very similar.
Linear models have been used to examine the association between biopsy rate and recall rate and non-linear (two parameter negative exponential) models have been used to examine the association between biopsy rates and cancer detection rates. All models are based on data from each of the 80 screening units and weighted by the number of women screened by each unit. Full details of the rationale behind the use of two parameter negative exponential models are given in Electronic Supplementary Material Appendix A.
The false-positive needle biopsy rate (referred to in this paper for clarity as the non-malignant/benign biopsy rate) is estimated from the non-linear models and defined as the number of biopsies minus the number of cancers (invasive and non-invasive cancers) from the model.
Relationship between biopsy rates and recall rate using linear regression
Prevalent (first) screens (women aged 45e52 years) and incident (subsequent) screens (53e70 years)
The association between needle biopsy rate and recall rate over the range of observed values has been modelled using linear regression weighted by the number of women screened by each unit. The model is biopsy rate (per 1,000 women screened) ¼ 3.44 recall rate (%) þ 6.74, p<0.001 with an Adj-R2 of 50.3% and correlation coefficient of 0.71. Therefore, for every 1% increase in recall rate at prevalent screens the biopsy rate increases by 0.344%. The fitted regression model for incident screens is biopsy rate (per 1,000) ¼ 2.67 recall rate (%) þ 5.41, p<0.001 with an Adj-R2 of 52% and correlation coefficient of 0.73. Therefore, for every 1% increase in recall rate at incident screens the bi-opsy rate increases by 0.267%. Fig 1 shows scatterplots of the biopsy rate and recall rate for prevalent and incident screens, respectively, with the 95% confidence band. There are no major departures from the line of best fit although a few units have high or low biopsy rates for their recall rate.