+
QCancer Scores –tools for earlier detection of cancer
Julia Hippisley-Cox,
GP, Professor Epidemiology & Director ClinRisk Ltd
GP Lincoln Refresher Course
18th May 2012
+
Acknowledgements

Co-author Dr Carol Coupland

QResearch database

University of Nottingham

ClinRisk (software)

EMIS & contributing practices & User Group

BJGP and BMJ for publishing the work

Oxford University (independent validation)

cancer teams, DH + RCGP+ other academics with whom we are
now working
+
QResearch Database

Over 700 general practices across the UK, 14 million patients

Joint not for profit venture University of Nottingham and EMIS
(supplier > 55% GP practices)

Validated database – used to develop many risk tools

Available for peer reviewed academic research where
outputs made publically available

Practices not paid for contribution but get integrated
QFeedback tool and utilities eg QRISK, QFracture.

Data linkage – deaths, deprivation, cancer, HES
+
Clinical Research Cycle
Clinical
practice &
benefit
Integration
clinical
system
Clinical
questions
Research +
innovation
+
QFeedback – integrated into EMIS
+
QScores – new family of Risk Prediction
tools



Individual assessment

Who is most at risk of preventable disease?

Who is likely to benefit from interventions?

What is the balance of risks and benefits for my patient?

Enable informed consent and shared decisions
Population level

Risk stratification

Identification of rank ordered list of patients for recall or reassurance
GP systems integration

Allow updates tool over time, audit of impact on services and
outcomes
+
Current published & validated
QScores
scores
outcome
Web link
QRISK
CVD
www.qrisk.org
QDiabetes
Type 2 diabetes
www.qdiabetes.org
QKidney
Moderate/severe renal failure
www.qkidney.org
QThrombosis
VTE
www.qthrombosis.org
QFracture
Osteoporotic fracture
www.qfracture.org
Qintervention
Risks benefits interventions to www.qintervention.org
lower CVD and diabetes risk
QCancer
Detection common cancers
www.qcancer.org
+
Early diagnosis of cancer: The
problem

UK has relatively poor track record when compared with
other European countries

Partly due to late diagnosis with estimated 7,500+ lives lost
annually

Later diagnosis due to mixture of

late presentation by patient (alack awareness)

Late recognition by GP

Delays in secondary care
+
Example of Colon cancer

This is one of the most common cancers

Half of patients never have a NICE qualifying sympton

Only one quarter diagnosed via 2 week clinic

One quarter present as emergencies

Earlier diagnosis my result in stage sift or prevent some
emergencies.
+
Example of pancreatic cancer

11th most common cancer

< 20% patients suitable for surgery

84% dead within a year of diagnosis

Chances of survival better if diagnosis made at early stage

Very few established risk factors (smoking, chronic
pancreatitis, alcohol) so screening programme unlikely

Challenge is to identify symptoms in primary care particularly hard for pancreatic cancer
+
Lung cancer

Commonest cause of death in UK

Very few diagnosed at operable stage

No screening tests currently but Chest xray useful

Vast majority present to GPs with symptoms.
+ Currently Qcancer predicts risk 6
cancers
Lung
Pancreas
Ovary
Colorectal
Kindey
Gastro-oesoph
+
QCancer scores – what they need
to do

Accurately predict level of risk for individual based on risk
factors and symptoms

Discriminate between patients with and without cancer

Help guide decision on who to investigate or refer and
degree of urgency.

Educational tool for sharing information with patient.
Sometimes will be reassurance.
+
QCancer scores – approach taken

Maximise strengths of routinely collected data electronic databases

Large representative samples including rare cancers

Algorithms can be applied to the same setting eg general practice

Account for multiple symptoms

Adjustment for family history

Better definition of smoking status (non, ex, light, moderate, heavy)

Age – absolutely key as PPV varies hugely by age

updated to meet changing requirements, populations, recorded data
+
Incidence of key symptoms vary
by age and sex
+
PPV of symptoms also vary by age
in men (Jones et al BMJ 2007).
20
18
16
14
12
10
8
6
4
2
0
45-54 yrs
55-64 yrs
65-74 yrs
75-84 yrs
haematuria haemoptysis dysphagia
rectal
bleeding
+
And PPV vary by age in women
(Jones et al BMJ 2007).
12
45-54 yrs
55-64 yrs
65-74 yrs
75-84 yrs
10
8
6
4
2
0
haematuria haemoptysis dysphagia
rectal
bleeding
+
Methods – development algorithm

Huge representative sample from primary care aged 30-84

Identify new alarm symptoms (eg rectal bleeding,
haemoptysis) and other risk factors (eg age, COPD, smoking,
family history)

Identify cancer outcome - all new diagnoses either on GP
record or linked ONS deaths record in next 2 years

Established methods to develop risk prediction algorithm

Identify independent factors adjusted for other factors

Measure of absolute risk of cancer. Eg 5% risk of colorectal
cancer
+
‘Red’ flag or alarm symptoms

Haemoptysis

Loss of appetite

Haematemesis

Weight loss

Dysphagia

Indigestion +/- heart burn

Rectal bleeding

Abdominal pain

Postmenopausal bleeding

Abdominal swelling

Haematuria

Family history

dysphagia

Anaemia

Constipation

cough
+ Results – the algorithms/predictors
Outcome
Risk factors
Symptoms
Lung
Age, sex, smoking,
deprivation, COPD,
prior cancers
Haemoptysis, appetite loss, weight loss,
cough, anaemia
Gastrooeso
Age, sex, smoking
status
Haematemsis, appetite loss, weight loss,
abdo pain, dysphagia
Colorectal Age, sex, alcohol,
family history
Rectal bleeding, appetite loss, weight loss,
abdo pain, change bowel habit, anaemia
Pancreas
Age, sex, type 2,
chronic pancreatitis
dysphagia, appetite loss, weight loss,
abdo pain, abdo distension, constipation
Ovarian
Age, family history
Rectal bleeding, appetite loss, weight loss,
abdo pain, abdo distension, PMB, anaemia
Renal
Age, sex, smoking
status, prior cancer
Haematuria, appetite loss, weight loss,
abdo pain, anaemia
+
Methods - validation

Previous QScores validation – similar or better performance
on external data

Once algorithms developed, tested performance

separate sample of QResearch practices

fully external dataset (Vision practices) at Oxford University

Measures of discrimination - identifying those who do and
don’t have cancer

Measures of calibration - closeness of predicted risk to
observed risk

Measure performance – PPV, sensitivity, ROC etc
+
Discrimination QCancer scores
ROC values for women
0.94
0.92
0.9
0.88
0.86
0.84
0.82
0.8
0.78
0.76
lung
renal
colorectal gastroes pancreas
ovary
+
Calibration - observed vs
predicted risk for ovarian cancer
+
Sensitivity for top 10% of
predicted cancer risk
Cut point
Threshold top
10%
Pick up rate for
10%
Colorectal
0.5
71
Gastrooesophageal
0.2
77
Ovary
0.2
63
Pancreas
0.2
62
Renal
0.1
87
Lung
0.4
77
+
Symptom recording in ovarian
cancer: cohort vs controls
QCancer BMJ (2012)
Hamilton BMJ (2009)
Abdominal pain
11.4%
8.7%
Abdominal distension
0.4%
0.6%
Loss appetite
0.5%
1.5%
Post menopausal
bleeding
1.6%
1.1%
Rectal bleeding
2.2%
1.5%
Weight loss
1.2%
Not reported
Note: different sample – QCancer national cohort 30-84 years
Hamilton local sample age matched controls 40+
+
Using QCancer in practice – v similar
to QRISK2


Standalone tools
a.
Web calculator www.qcancer.org
b.
Windows desk top calculator
c.
Iphone – simple calculator
Integrated into clinical system
a.
Within consultation: GP with patients with symptoms
b.
Batch: Run in batch mode to risk stratify entire practice or
PCT population
+
GP system integration:
Within consultation

Uses data already recorded (eg age, family history)

Stimulate better recording of positive and negative symptoms

Automatic risk calculation in real time

Display risk enables shared decision making between doctor
and patient

Information stored in patients record and transmitted on referral
letter/request for investigation

Allows automatic subsequent audit of process and clinical
outcomes

Improves data quality leading to refined future algorithms.
+
Iphone/iPad
+
GP systems integration
Batch processing

Similar to QRISK which is in 90% of GP practices– automatic
daily calculation of risk for all patients in practice based on
existing data.

Identify patients with symptoms/adverse risk profile without
follow up/diagnosis

Enables systematic recall or further investigation

Systematic approach - prioritise by level of risk.

Integration means software can be rigorously tested so ‘one
patient, one score, anywhere’

Cheaper to distribute updates
+
Clinical settings

Modelling done on primary care population

Intended for use in primary care setting ie GP consultation

Potential use in other clinical settings as with QRISK


Pharmacy

Supermarkets

‘health buses’

Secondary care
Potential use by patients - linked to inline access to health
records.
+
Summary key points

Individualised level of risk - including age, FH, multiple
symptoms

Electronic validated tool using proven methods which can be
implemented into clinical systems

Standalone or integrated.

If integrated into computer systems,




improve recording of symptoms and data quality
ensure accuracy calculations
help support decisions & shared decision making with patient
enable future audit and assessment of impact on services and
outcomes
+  Next steps - pilot work in clinical
practice supported by DH
+
Thank you for listening
Any questions (if time)