PREDICT: A prognostication and treatment benefit tool for early breast cancer - Caldas

Submitting Institution

University of Cambridge

Unit of Assessment

Clinical Medicine

Summary Impact Type

Health

Research Subject Area(s)

Medical and Health Sciences: Oncology and Carcinogenesis, Public Health and Health Services


Download original

PDF

Summary of the impact

PREDICT is a prognostication and treatment benefit decision aid aimed at aiding the breast cancer multi-disciplinary team in the management of women with early breast cancer. The user-friendly, web-based tool was developed in collaboration with the Cambridge Breast Unit multi-disciplinary team, the Eastern Cancer Registration and Information Centre. Implemented online, PREDICT is hosted on a NHS web-server. Since 2012 PREDICT has been used widely by clinicians throughout the UK and world-wide.

Underpinning research

The PREDICT model was developed in 2010 by a team jointly led by Professor Carlos Caldas (University Professor 2006-present, in the department of Oncology) and Professor Paul Pharoah (UOA 2: CR-UK Senior Clinical Research Fellow 1999-2009, University Reader 2009-13, University Professor 2013-present, in the department of Public Health and Primary Care).

Patients treated in the Cambridge Breast Unit are stratified for adjuvant chemotherapy according to a guideline developed in 2004. This takes into account the serious adverse events that occur with chemotherapy and that as a consequence, "many physicians consider a cut-off of an additional 3% or more added benefit sufficient to justify recommending treatment". Thus, for an absolute survival benefit of < 3%, chemotherapy is not recommended, for an absolute survival benefit of 3-5% the benefits and harms are considered equivalent and discussed with the patient, for an absolute benefit of >5% chemotherapy is recommended.

Until 2010 the absolute benefits of chemotherapy were estimated using Adjuvant! Online, an online prognostic model developed over a decade ago by an American oncologist (Peter Ravdin). This is based on US data and has not been validated with UK data. Furthermore, Adjuvant! Online does not include several important prognostic variables including mode of detection and molecular biomarkers such as tumour HER2 status. Thus there was a clinical need for an equivalent model, based on and validated using UK data that was flexible and able to incorporate additional prognostic variables.

Research carried out in Cambridge allowed the development of the PREDICT model, which was based on survival-time data on 5,700 women with early breast cancer treated between 1999 and 2003. These data were obtained through the Eastern Cancer Registration and Information Centre and used to determine the influence of key prognostic variables on survival [1]. The model was then validated using an independent data set from the West Midlands Cancer Intelligence Unit [1]. The PREDICT web interface was developed in 2010 and is hosted by the Eastern Cancer Registration and Information Centre.

In order to compare directly the performance of Adjuvant! Online and PREDICT, a second validation of PREDICT was carried out using the same data set. While both models performed well, the breast cancer specific survival calibration for PREDICT was significantly better than that of Adjuvant! Online with similar discrimination [2].

In parallel with the PREDICT model development work, Caldas and Pharoah have led a research programme investigating the molecular pathology of breast cancer in collaboration with other groups from the international Breast Cancer Association Consortium (BCAC). This research has enabled further development of PREDICT in response to feedback from clinicians and requests for additional features in the model. In particular, there were many requests to incorporate tumour HER2 and KI67 status into the model. Results from one of the BCAC projects were used to enable the incorporation of HER2 into the model [3], with the new PREDICT model being further validated using the British Columbia data set used to validate the original model [4]. More recently, the results from another of our molecular pathology studies has informed the incorporation of tumour KI67 status into the model [5].

References to the research

1. Wishart GC, Azzato EM, Greenberg DC, Rashbass J, Kearins O, Lawrence G, Caldas C, Pharoah PD. PREDICT: a new UK prognostic model that predicts survival following surgery for invasive breast cancer. Breast Cancer Res. 2010;12(1):R1.

 
 
 
 

2. Wishart GC, Bajdik CD, Azzato EM, Dicks E, Greenberg DC, Rashbass J, Caldas C, Pharoah PD. A population-based validation of the prognostic model PREDICT for early breast cancer. Eur. J. Surg. Oncol. 2011;37(5):411-7.

 
 
 
 

3. Blows FM, Driver KE, Schmidt MK, Broeks A, van Leeuwen FE, Wesseling J, Cheang MC, Gelmon K, Nielsen TO, Blomqvist C, Heikkila P, Heikkinen T, Nevanlinna H, Akslen LA, Begin LR, Foulkes WD, Couch FJ, Wang X, Cafourek V, Olson JE, Baglietto L, Giles GG, Severi G, McLean CA, Southey MC, Rakha E, Green AR, Ellis IO, Sherman ME, Lissowska J, Anderson WF, Cox A, Cross SS, Reed MW, Provenzano E, Dawson SJ, Dunning AM, Humphreys M, Easton DF, Garcia-Closas M, Caldas C, Pharoah PD, Huntsman D. Subtyping of breast cancer by immunohistochemistry to investigate a relationship between subtype and short and long term survival: a collaborative analysis of data for 10,159 cases from 12 studies. PLoS Med 2010;7(5):e1000279.

 
 
 

4. Wishart GC, Bajdik CD, Dicks E, Provenzano E, Schmidt MK, Sherman M, Greenberg DC, Green AR, Gelmon KA, Kosma VM, Olson JE, Beckmann MW, Winqvist R, Cross SS, Severi G, Huntsman D, Pylkas K, Ellis I, Nielsen TO, Giles G, Blomqvist C, Fasching PA, Couch FJ, Rakha E, Foulkes WD, Blows FM, Begin LR, Van't Veer LJ, Southey M, Nevanlinna H, Mannermaa A, Cox A, Cheang M, Baglietto L, Caldas C, Garcia-Closas M, Pharoah PD. PREDICT Plus: development and validation of a prognostic model for early breast cancer that includes HER2. Br. J. Cancer 2012;107(5):800-7.

5. Ali HR, Dawson SJ, Blows FM, Provenzano E, Leung S, Nielsen T, Pharoah PD, Caldas C. A Ki67/BCL2 index based on immunohistochemistry is highly prognostic in ER-positive breast cancer. J. Pathol. 2012;226(1):97-107.

 
 
 
 

Details of the impact

PREDICT is implemented online as a national resource with the web interface being hosted on a NHS web-server at www.predict.nhs.uk. The monthly hits (> 4,000) on the website indicate clearly the impact of the tool.

One of the key decisions in the management of women with early breast cancer is whether or not to offer adjuvant chemotherapy in conjunction with primary surgery and radiotherapy. The key output of PREDICT is the expected absolute reduction in mortality at five and ten years associated with adjuvant chemotherapy.

Cambridge Breast Unit (CBU)

Until 2010 the CBU estimated absolute benefits of chemotherapy were estimated using Adjuvant! Online. During 2010 and 2011 both PREDICT and Adjuvant! Online were used in parallel.

Clinical audit

Demographically, the use of PREDICT could benefit 18% of women worldwide. The Cambridge Breast Unit carried out an audit of the first 200 patients discussed when both Adjuvant! Online and PREDICT were used by the multi-disciplinary team [1]. The chemotherapy recommendations that would have been made based on the output from each model were then compared. In 163 patients (82 per cent) the chemotherapy decision would have been the same whichever model was used. A different recommendation would have occurred for 37 patients (18 per cent), which would benefit them, some women avoiding unnecessary chemotherapy, others having effective treatment which would not otherwise have been given.

Change in practice

Since 2012 PREDICT has been the only model used routinely in Cambridge for all patients being discussed at the weekly multi-disciplinary team meeting. The absolute benefit of adjuvant chemotherapy estimated by PREDICT is used to guide the use of adjuvant chemotherapy according to the guideline described in section 2.

Other clinical departments in the UK

We have had multiple requests from clinicians for the incorporation of additional features indicating that the model is being widely used. PREDICT is used by the multi-disciplinary clinical teams in Belfast, Brighton, Derby, Dundee, Oxford and Sheffield, but the web usage statistics from 2011 suggest that PREDICT is also being used widely across the country. The impact of PREDICT on clinical practice is clearly demonstrated from the extensive use of the web interface (see below).

Public, Patient Partnership

PREDICT has been widely reported in regional and national media including ITV, The Times and The Daily Mail (ref 2). We have clear evidence that women with early breast cancer are accessing the interface in order to determine their own risk and to help them in their discussion with their oncologists about treatment options — this is very much in keeping with current thinking to empower patients through knowledge to play an important role in their own care.

Web usage data

PREDICT was designed to have a user-friendly interface to help clinicians in making clinical management decisions. Informal feedback from clinicians from both Cambridge and elsewhere has indicated that the interface is easy to use.

The number of visits to the web site each month has increased steadily since its launch in January 2011, with 3,266 visits in April 2013.

Monthly web usage statistics for the PREDICT website, Jan 2011-Jan
        2013
Monthly web usage statistics for the PREDICT website, Jan 2011-Jan 2013

There have been 43,870 visits to the web site with 70 per cent of the visits (30,563) being accessed from UK and ten per cent (4,468) from USA. The web site is visited from all over the UK, with London accounting for 17 per cent and Cambridge accounting for just 1 per cent of all traffic on the web site (ref 3).

Sources to corroborate the impact

  1. Loh S-W, Rodriguez-Miguelez M, Pharoah P, Wishart G. A comparison of chemotherapy recommendations using the Predict and Adjuvant models. Eur. J. Surg. Oncol. 2011;37(5):S21-S22.
  2. Press coverage: see http://www.predict.nhs.uk/press.shtml for details.
  3. Web usage statistics from Google Analytics at https://www.google.com/analytics