Eligibility

Eligibility for entry into the Sloane Project is indicated in figure 1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Sloane eligibility flow chart. *Including DCIS with microinvasion (no more than 1 mm). **Further events include ipsilateral breast and nodal events, contralateral breast and nodal events and distant events, as well as deaths. ADH, atypical ductal hyperplasia; AIDEP, atypical intraductal epithelial proliferation; DCIS, ductal carcinoma in situ; FEA, flat epithelial atypia; LISN, lobular in situ neoplasia (including lobular carcinoma in situ and atypical lobular hyperplasia); NHSBSP, National Health Service Breast Screening Programme.

Case ascertainment

The Sloane prospective cohort study is based on patient-level data collected by the NHS as part of the care and support of patients and is voluntarily submitted by UK NHSBSP units. Sloane Project cases were therefore matched to the data collected through the annual NHSBSP/Association of Breast Surgery (ABS) breast screening audit,24 a national audit of all patients diagnosed through the NHSBSP, primarily designed to assess surgical performance with surgical Quality Assurance Standards, for the period 01 April 2003–31 March 2012, to understand the level of case ascertainment.

Governance

The project is managed, and the data are collated, maintained and quality assured by the Screening Quality Assurance Service (SQAS), based in Birmingham, UK. It is overseen by a Project Steering Group (see the Acknowledgements section) comprising radiologists, pathologists, surgeons, managers, statisticians, oncologists and patient advocates.

More recently, the study has been permitted to process personally identifiable data without consent under Regulation 5 of Statutory Instrument 2002 No. 1438: The Health Service (Control of Patient Information)25 25 (15/CAG/0207) in line with the following clause: ‘quality assuring screening services to ensure they are effective and safe, and that any incidents are investigated and managed appropriately’. This statutory exemption to common law permits the processing of personally identifiable data, as part of the core remit of population screening.

Data are securely held within the organisation with patient identifiers recorded. There is restricted access to a certified group of individuals and a full audit trail for access. A system-level security policy is in place to cover the Sloane Project as a system asset as well as a data protection impact assessment. Figure 2 describes example of data flows in and out of the Sloane Project, along with the controls in place to mitigate for risks.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Example of data flow diagram and key for Sloane Project data. *Screening Quality Assurance Service (SQAS). **Screening Research Advisory Committee (RAC). (1) Ensure follow SQAS guidelines for secure posting or email. (2) Use detailed internal protocols and policies for all aspects of Sloane work. (3) Avoid duplication by checking at least two identifiers. (4) Assign tumour and patient ID to minimise the use of personal data when working with the data. (5) Use system integrity checks and validation checks to indicate problems with data or quality assurance purposes. (6) System-level security policy in place, which includes audit trail, user access limits, server security among other things. Data protection impact assessment also in place. (7) With whom we have an existing data sharing contact. (8) Ensure publication is seen and approval given by Sloane Steering Group before publishing. Data release functions to change from 01 October 2021, but the process will remain similar, ODR, Office for Data Release.

Patient and public involvement

Patients from Independent Cancer Patients’ Voice26 joined the Sloane Project Steering Group in 2012. Since this time, they have actively contributed to publications guidance, communications and promotional material, including the website, as well as supporting funding applications. They are also involved, as part of a wider international collaboration, with the research that has followed from the cohort study.

Clinical follow-up data from the English programme are accessed via a data sharing agreement (DSA) between the National Disease Registration Service (NDRS) and SQAS. Follow-up and mortality data from Scotland, Wales and Northern Ireland need specific permissions. The agreements (from these ‘devolved nations’) do not allow for further onward sharing of data at present. Apart from the limitation described above, access to the Sloane Project data from external parties is governed by application to the breast screening Research Advisory Committee and Office for Data Release. Data will only be released by the Sloane Project to researchers under approval and in an anonymised or depersonalised format and under a DSA.

Primary data capture

Demographic data and data deriving from radiology, pathology, surgery and adjuvant treatments at the time of diagnosis and initial treatment were originally captured via individual, pathology, surgery and radiotherapy paper proformas (see online supplemental file 1), collected at NHSBSP unit and hospital level. More recently, the proformas have been shortened and combined into one form, which can be completed electronically (see online supplemental file 2).27 The forms are submitted to the Sloane Project team and data are then entered on a secure database held on an SQL server that generates an individual patient and tumour identifier.

As all Sloane patients are diagnosed within the NHSBSP, adherence to NHSBSP guidelines28 and participation in the relevant quality assurance programmes29 are mandatory. Participating units are required to follow a pathology protocol (see online supplemental files 3, 4)30 that includes guidance on the handling and reporting of specimens, definitions for DCIS and atypia, microinvasion, cytonuclear grade, comedo necrosis and assessment of excision margins to NHSBSP pathology standards.31 Radiology guidelines (see online supplemental files 3, 4)32 encourage participating radiologists to complete the radiology proforma. Radiologists participate in the NHSBSP Personal Performance in Mammographic Screening external quality assurance scheme.33

Data handling

Data are manually transcribed into the database, which has built-in integrity checks run on a regular basis, along with extra validation checks for each subsequent data request. Missing data items are subsequently added as far as possible. If the data cannot be sourced from information already supplied or from the NHSBSP unit or other data sources, then unknown/unrecorded status is assigned. Best efforts are made to minimise the number of missing data items (see online supplemental file 5).

Validation of data is undertaken by cross-checking with original NHSBSP unit source documents for those patients with recurrence and more generally, for the overall dataset, against the ABS national audits.24 The Sloane Project data, as completed by clinicians, are taken as the primary source when there is conflicting information. However, the ABS national audit is used to supplement missing data items.

Classification of subsequent events

Given the potential difficulties in distinguishing local recurrence from a new primary lesion in the same breast, terminology has been developed to classify subsequent events (online supplemental table 1).

Death classification

Date and cause of death have historically been obtained directly from the NHS Trusts or Health Boards. Since 2013, the majority are now sourced from the NDRS/NHS Wales Cancer Network Information System Cymru data provided by the Office for National Statistics (ONS),34 National Records of Scotland35 and Northern Ireland patient records. Rules for coding the underlying cause of death are based on the rules used by ONS34 that apply the condition or conditions entered in the lowest completed line of part I of the Medical Certificate Cause of Death. Women who die of breast cancer but who have no intervening breast events recorded are deemed to have had distant metastases on the date they died. The definitions for cause of death in online supplemental table 2 are used for the Sloane dataset.

Follow-up and recurrence data capture

A methodology has been developed to identify further events, including ipsilateral breast/nodal, contralateral breast/nodal and distant events. This is run annually for the English patients, who comprise 87.1% of the Sloane cohort, and on an ad hoc basis for those from Scotland, Wales and Northern Ireland, who make up 10.1%, 2.5% and 0.3% of patients, respectively.

Further events are identified by matching women by their unique NHS number (in England and Wales) and Community Health Index number (in Scotland and Northern Ireland) and information provided by NHSBSP units to routinely collected UK datasets, as detailed in table 1.

The current version of the International Classification of Diseases and Related Health Problems36 was used when ascertaining diagnostic codes alongside Office of Population Censuses and Surveys Classification of Interventions and Procedures (OPCS-4)37 when ascertaining procedure codes.

The censor date used in time-related analyses is the earliest of date of death, date of first further event (DCIS, invasive breast cancer, metastatic breast cancer) or end of time period of matched datasets if no further event has occurred and the woman is still alive.

Follow-up methodology

The follow-up methodology was developed to test the proof of concept of using the English National Cancer Registration and Analysis Service (NCRAS) data to identify further events and decrease the burden of local data collection within NHSBSP units and NHS Trusts.

The methodology is based on the following variables from the Sloane database:

• Date of mammogram.

• Date of surgery.

• Laterality of primary lesion.

These are then linked (via NHS number) to patients registered within the datasets and data items as described in table 1.

Validation of the follow-up methodology

The methodology was validated on a subset of 3711 (39%) of the cohort of Sloane patients for whom recurrence information was independently collected directly from the NHSBSP units prior to the generation of the methodology using NCRAS data from 2013. These data were compared with NCRAS data focusing on invasive status (invasive/non-invasive/distant), laterality and date of diagnosis for the primary and any further event.

To confirm a match of either primary or recurrence, data were considered to be:

• Acceptable—same laterality, same invasive status, date of diagnosis within 6 months (either prior or after).

• Not acceptable—different laterality, different invasive status (with the exception of microinvasion which is coded as invasive on Cancer Analysis System (CAS)), dates >12 months apart.

• Query acceptable—for example, dates different between 6 and 12 months; for example, in one case the diagnosis date and NCRAS chemotherapy date (the only event) differed by 8 months, which was considered to be acceptable.

In order to support this proof-of-concept project and provide further validation, we sent a list of 100 Sloane Project patients to University Hospitals of Derby and Burton NHS Foundation Trust (UHDB). Some of these patients we had identified as having a further event using the methodology. The Trust was blinded to our records for each patient and was requested to identify those who they believed to be patients with a further event, and provide date and type, but also to indicate those who had not had a further event according to their records.

Case ascertainment

Overall, case ascertainment ranged from 0% to 93% by NHSBSP unit (median 32%, mean 38%). Individual attributes (including patient, tumour and treatment characteristics), such as age at screening (table 2), cytonuclear grade of DCIS, DCIS size, oestrogen receptor status, surgical and adjuvant treatment have been shown to be representative when comparing the voluntary Sloane prospective cohort study with the compulsory ABS/NHSBSP annual screening audit, although the more detailed Sloane data represent 38% of patients with DCIS diagnosed through the NHSBSP over the same time period.

Primary data completeness

Data completeness, assessed against a pragmatic desired completeness level of ≥95%, demonstrated the majority of key data items for the patients with DCIS within the Sloane Project database have excellent completeness levels. Data completeness was high for most data fields (table 3).

Follow-up methodology validation results

The results of the initial validation of the follow-up methodology, whereby recurrence (presence or absence of) information independently collected by Sloane was compared with routinely collected datasets from NCRAS, are shown in tables 4 and 5.

Table 4 shows that of 3711 Sloane patients with DCIS for whom follow-up data were received direct from the NHSBSP units, 91.5% were also in the NCRAS datasets and were an ‘acceptable match’. There was a 96.4% match between the Sloane and NCRAS datasets on fact of ‘no further event’. One hundred eight further events were identified through NCRAS datasets that had not been identified by the NHSBSP units. Further checking with the hospitals at which the event occurred confirms that these are correct. There are likely to be a number of reasons for this such as the patient’s further event being diagnosed at a different NHS Trust or Health Board, or the type of hospital systems that were checked for further event and the likelihood of finding the information on those systems.

Table 5 shows more detail about those 18 recurrences not matched according to our rules by NCRAS datasets. Of these, 11 were identified in Hospital Episode Statistics (HES) but the invasive status was recorded incorrectly (9 of 11) or laterality was incorrect (2 of 11). Five were not identified in HES because two had surgery performed outside the NHS, one had no surgery due to comorbidity, two remained unknown, although may have had their surgery in the private sector. Two were not found on CAS or HES but were identified on the Cancer Waiting Times dataset, but invasive status was incorrect.

We concluded that although 92% events were successfully matched for further event correctly on NCRAS datasets, a further 13 recurrences (6%) would now be picked up by recent improvements in NCRAS and linkage to other datasets. It is likely that using currently available matching to national datasets, 206 of 211 patients (97%) would have accurate identification of DCIS or invasive breast cancer recurrence on the same side, contralateral breast or distant metastatic disease.

The further validation on 100 UHDB patients showed that using the Sloane follow-up methodology, 20 matched exactly for further events, 75 matched as having no further events and 5 did not match. Thus, 95% of patients were correctly matched. Out of the five that did not match, three were because the Sloane methodology had resulted in a ‘query further event’ and UHDB confirmed ‘no further event’, one was ‘unknown invasive status’ and UHDB confirmed ‘invasive’ and one was recorded as ‘ipsilateral DCIS’ whereas UHDB had it recorded as ‘ipsilateral invasive’. These results can be used to improve the methodology.

Strengths

Long-term follow-up of patients is required to ensure the impact of diagnosis, treatment and subsequent care is optimised and changes in management evaluated. Large, prospective cohort studies of conditions such as DCIS where subsequent events are relatively rare and occur over decades present particular challenges. Even for smaller cohorts and especially for clinical trials, the benefits of accurately matching prospectively collected data to national datasets may be invaluable. The ambitious, prospective, long-term Sloane Project has identified and followed some 15 000 patients with NHSBSP-detected pre-invasive breast neoplasia over almost two decades. We demonstrate here a high level of concordance between subsequent disease events confirmed in the Sloane cohort data (provided directly by NHSBSP units) and those identified using a follow-up methodology applied to electronically stored national datasets. While events were rare, representing only 2% of the total cohort, concordance between data sources was strong. There was also a 96% level of concordance for freedom from further events. This is encouraging but requires future validation, before the use of NCRAS data to determine freedom from further events can be regarded as definitive.

This methodology, if validated in other settings, and potentially further refined, may be generalisable to other cohort studies and for clinical trial follow-up. The long-term follow-up of patients beyond an initial 5 years is increasingly important in early invasive breast cancer, as well as DCIS, given the low frequency of events over an extended time period in these groups.

Although Sloane patients comprise one-third of those diagnosed with DCIS within the NHSBSP during the relevant time period, the approach laid out here confirms that the data quality and completeness are high and appear to be representative of the overall population of patients with screen-detected DCIS in the UK. Thus, conclusions derived from analyses of care and clinical outcomes from the Sloane Project9 38–49 are robust and generalisable to all women in the UK with screen-detected DCIS. This is supported by a recently published study of over 35 000 women with screen-detected DCIS treated within the UK NHSBSP from its start until 2014.50 Patient histopathology, treatment and outcomes reported reflect those curated within the more detailed Sloane cohort.

Limitations

Participation in data provision by NHSBSP units in the Sloane Project is voluntary, not specifically funded and 100% case capture was not achieved. However, comparison with the mandatory ABS/NHSBSP annual audits, which collect less detailed data, demonstrates that the Sloane dataset is representative of a cross-section of patients, geographical spread and of NHSBSP units’ practice at the time of accrual.

Engagement has been challenged by lack of resources, particularly NHS staff time, available for long-term data collection. However, staff participants in the project receive Continuing Professional Development (CPD) credits and certificates for their participation in primary and follow-up data capture which is useful for annual appraisal and revalidation. The Sloane Project DCIS database remains open for completion of data for those cases with missing information (and is currently over 94% complete).

Additional challenges include different health service structures and processes across England, Scotland, Wales and Northern Ireland. Data regarding the small proportion of patients treated outside the NHS are limited. Furthermore, coding of DCIS in some datasets may be inaccurate, resulting in a potential overestimation of the relative proportion of DCIS that is recurring as invasive disease compared with that recurring as DCIS.