Risk of colorectal cancer Risk of colorectal cancer in patients taking statins in patients taking statins
and NSAIDSand NSAIDSDr Yana Vinogradova, Prof Julia Hippisley-Cox,Dr Yana Vinogradova, Prof Julia Hippisley-Cox,Dr Carol Coupland and Prof Dr Carol Coupland and Prof Richard LoganRichard LoganDivisions of Primary Care, and Epidemiology & Divisions of Primary Care, and Epidemiology & Public Health, University of Nottingham, UKPublic Health, University of Nottingham, UK
UEGW Plenary sessionUEGW Plenary session
BerlinBerlin
2323rdrd October 2006 October 2006
Background to CRC statin Background to CRC statin analysis analysis
• Evidence from animal models and other lab Evidence from animal models and other lab data that statins might prevent cancerdata that statins might prevent cancer
• A meta-analysis of 26 large statin trials A meta-analysis of 26 large statin trials found no difference in incidence of CRC but found no difference in incidence of CRC but limited power as only 4 reported CRC datalimited power as only 4 reported CRC data
• Recent case control study Recent case control study (Poynter NEJM 2005) found a 47% reduction in CRC risk found a 47% reduction in CRC risk
MeanwhileMeanwhile
• Mass of observational and experimental Mass of observational and experimental evidence that aspirin and traditional NSAIDs evidence that aspirin and traditional NSAIDs lower CRC risklower CRC risk
• Trials of Colorectal Trials of Colorectal adenomaadenoma prevention with prevention with COX2 show a 40-50% reduction in recurrence COX2 show a 40-50% reduction in recurrence (Bertagnolli, Arber NEJM August 2006)
• Little observational data available on risk of Little observational data available on risk of colorectal colorectal cancercancer in COX2 users in COX2 users
Study population: Study population: QRESEARCH databaseQRESEARCH database
• Currently largest primary care database in the UKCurrently largest primary care database in the UK
• 537 general practices across the UK537 general practices across the UK
• > 9 million patients including those who have died > 9 million patients including those who have died
or left, as well as patients still registeredor left, as well as patients still registered
• > 30 million person-years of observation> 30 million person-years of observation
General Practice data General Practice data collection:collection:
Data source: Data source: QRESEARCH databaseQRESEARCH database
• Derived from GP clinical records Derived from GP clinical records
• Patient level consolidated databasePatient level consolidated database
• Anonymised dataAnonymised data
• Longitudinal data for 15+ yearsLongitudinal data for 15+ years
• Validated against external and internal Validated against external and internal
measuresmeasures
Study design & settingStudy design & setting• Nested case control studyNested case control study• Study period Jan 1995-July 2005 Study period Jan 1995-July 2005 • Cases were incident colorectal Cases were incident colorectal
cancer patientscancer patients• 5 controls matched by 5 controls matched by
• AgeAge• SexSex• PracticePractice• Calendar yearCalendar year
Exposure Exposure assessment :assessment :
• statins, NSAIDs, Cox2 Inhibitors and statins, NSAIDs, Cox2 Inhibitors and aspirinaspirin prescriptionsprescriptions
analysis restricted to subjects with +4yrsanalysis restricted to subjects with +4yrs
of prescribing dataof prescribing data • any use: any use:
• at least 1 script in 13-48 months prior to the at least 1 script in 13-48 months prior to the index index datedate (date of diagnosis in the case) (date of diagnosis in the case)
• number of prescriptions: number of prescriptions: • 1 script only1 script only• 2-12 scripts2-12 scripts• 13-24 scripts13-24 scripts• 25 or more scripts25 or more scripts
Statistical analysisStatistical analysis
• Conditional logistic regressionConditional logistic regression
• Odds ratios + 95% CIOdds ratios + 95% CI
• Unadjusted & adjustedUnadjusted & adjusted
Study Sample:Study Sample:9,694 incident cases of colorectal cancer
1995/2005
5,686 cases with 4 years of medical records
24,982 controlswith 4 years ofmedical records
2,425 cases with 8 years of medical records
12,131 controls with 8 years of medical records
8.6 8.3
4.8 5.3
8.76.8
30.229.3
12.012.1
1.0 0.6 0.9 1.4
5.4 5.2
0
5
10
15
20
25
30
35
Pro
por
tion
of p
atie
nts
(%
)
IHD IHD+MI Diabetes HBP OA Colitis RA Stroke Adjusted OR: 1.01 0.90 1.26* 1.01 0.99 1.70* 0.65* 1.04 * - P value < 0.005 © QRESEARCH 2005
Proportions of patients with morbidities
cases controls
Comorbidity in CRC cases and controls:Comorbidity in CRC cases and controls:
Confounding factors :Confounding factors :
• Body mass indexBody mass index• Less than 25 kg/m2Less than 25 kg/m2
• 25 to 29.9 kg/m225 to 29.9 kg/m2
• 30 kg/m2 or more30 kg/m2 or more
• BMI not recordedBMI not recorded
• Smoking statusSmoking status• Non-smokerNon-smoker
• SmokerSmoker
• not recordednot recorded
• Socio-economic status Socio-economic status (Townsend score for post code)(Townsend score for post code)
• quintilesquintiles
• MorbiditiesMorbidities• IHD (IHD (±MI)±MI)
• DiabetesDiabetes
• High BPHigh BP
• OsteoarthritisOsteoarthritis
• ColitisColitis
• Rheumatoid arthritisRheumatoid arthritis
• StrokeStroke
9.5 9.7
4.6 4.4
32.933.9
21.6 21.5
0
5
10
15
20
25
30
35
40
Pro
por
tion
of p
atie
nts
(%
)
Statins COX-2 NSAIDs Aspirin Adjusted OR: 0.93 (0.83 to 1.04) 1.07 (0.92 to 1.24) 0.94 (0.88 to 1.00) 0.99 (0.90 to 1.08) © QRESEARCH 2005
Use of medications in 13-48 months prior to the index date
cases controls
Any use in 13-48 months prior to the Any use in 13-48 months prior to the index dateindex date
1
0
.25
.5
.75
1
1.25
1.5
1.75
1 2-12 13-24 25+ 1 2-12 13-24 25+ 1 2-12 13-24 25+ 1 2-12 13-24 25+
Statins COX-2 Inhibitors Traditional NSAIDs Aspirin
odds ratio compared to no prescriptions for a drug grouplower/upper limits of 95 percent confidence interval
Adj
uste
d O
dds
Rat
io
Number of scripts
© QRESEARCH 2005
Number of scripts in 13-48 months prior to the index date
1
0
.25
.5
.75
1
1.25
1.5
1.75
1 2- 13-25-37-49+ 1 2- 13-25-37-49+ 1 2- 13-25-37-49+ 1 2- 13-25-37-49+
Statins COX-2 Inhibitors Traditional NSAIDs Aspirin
odds ratio compared to no prescriptions for a drug grouplower/upper limits of 95 percent confidence interval
Adj
uste
d O
dds
Rat
io
Number of scripts
© QRESEARCH 2005
Number of scripts in 13-96 months prior to the index date
Risks with prolonged Risks with prolonged prescribing in 13-48 months prescribing in 13-48 months
prior to index dateprior to index date• 25 or more statin scripts:25 or more statin scripts: adj. OR adj. OR 1.13 1.13 (0.91 - 1.41) (0.91 - 1.41)
• 25 or more COX-2 scripts:25 or more COX-2 scripts: adj. OR adj. OR 0.34 0.34 (0.14 - 0.85) (0.14 - 0.85)
• 25 or more NSAID scripts:25 or more NSAID scripts: adj. OR adj. OR 0.760.76 (0.60 – 0.95) (0.60 – 0.95)
• 25 or more Aspirin scripts:25 or more Aspirin scripts: adj. OR adj. OR 0.880.88 (0.74 – 1.05) (0.74 – 1.05)
ConclusionsConclusions• Prolonged statin use is Prolonged statin use is notnot associated associated
with reduced risk of colorectal cancerwith reduced risk of colorectal cancer
• Prolonged NSAID and Cox2 inhibitor Prolonged NSAID and Cox2 inhibitor use is associated with reduced risk of use is associated with reduced risk of colorectal cancercolorectal cancer
ImplicationsImplications• No increase in risk of colorectal No increase in risk of colorectal
cancer with cholesterol loweringcancer with cholesterol lowering
• Supports role for COX 2 inhibitors in Supports role for COX 2 inhibitors in secondary / tertiary prevention of secondary / tertiary prevention of colorectal cancercolorectal cancer
1
0
.25
.5
.75
1
1.25
1.5
1.75
2
1 2-12 13-24 25+ 1 2-12 13-24 25+
Lipophilic Hydrophilic
Adj
uste
d O
dds
Rat
io
Number of scriptsGraphs by Statins
Number of scripts
1
0
.25
.5
.75
1
1.25
1.5
1.75
2
Low Usual High Low Usual High
Lipophilic Hydrophilic
Adj
uste
d O
dds
Rat
io
Last doseGraphs by Statins
Last dose
Note: odds ratio (and 95% confidence interval) compared to no prescriptions for a statin© QRESEARCH 2005
Use of lipophilic and hydrophilic statins in 13-48 months prior to the index date
Methodological Methodological strengthsstrengths
• Large sample size and representative Large sample size and representative populationpopulation
• Data electronically collected – unlikely Data electronically collected – unlikely misclassification bias misclassification bias
• Data collected before the diagnosis – no Data collected before the diagnosis – no recall biasrecall bias
• Excluded prescriptions 12 months prior to Excluded prescriptions 12 months prior to cancer diagnosis cancer diagnosis
Use of individual statinsUse of individual statins
Casesno. (%)
Controlsno. (%)
Unadjusted odds ratio (95%CI)
Adjusted† odds ratio (95%CI)
Atorvastatin250 (4.4) 954 (3.8) 1.19 (1.02-1.38) 1.11 (0.95 - 1.30)
Simvastatin287 (5.0) 1420 (5.7) 0.88 (0.77-1.01) 0.83 (0.72 - 0.96)
Pravastatin 27 (0.5) 136 (0.5) 0.90 (0.59-1.37) 0.84 (0.55 - 1.28)
Fluvastatin 44 (0.8) 150 (0.6) 1.35 (0.95-1.91) 1.21 (0.85 - 1.74)
Cerivastatin 56 (1.0) 179 (0.7) 1.40 (1.03-1.91) 1.34 (0.97 - 1.86)
0
5
10
15
20
25
30
35
Pro
por
tion
of p
atie
nts
(%
)
IHD IHD+MI Diabetes HBP OA Colitis RA Stroke© QRESEARCH 2005
Proportions of patients with morbidities
cases with 4 years of records
cases with 8 years of records
controls with 4 years of records
controls with 8 years of records
Baseline characteristics:Baseline characteristics:CasesCases
n=5,686 n=5,686
ControlsControls
n=24,982 n=24,982
Males (number, Males (number, percent) percent) 3,181 (55.9) 3,181 (55.9) 14,014 (56.1) 14,014 (56.1)
Age in years (median, Age in years (median, IQR) IQR) 72 (64 to 79) 72 (64 to 79) 72 (64 to 79) 72 (64 to 79)
Months of records Months of records (median, IQR) (median, IQR) 88 (66 to 117) 88 (66 to 117) 88 (66 to 117) 88 (66 to 117)
Deprivation Deprivation Townsend score Townsend score -1.26 (-3.05 to 1.57) -1.26 (-3.05 to 1.57) -1.43 (-3.16 to 1.46) -1.43 (-3.16 to 1.46)
Body mass index Body mass index (median, IQR) (median, IQR) 26.0 (23.5 to 29.0) 26.0 (23.5 to 29.0) 26.1 (23.6 to 29.0) 26.1 (23.6 to 29.0)
Smokers (number, Smokers (number, percent) percent) 985 (17.3) 985 (17.3) 4,060 (16.3) 4,060 (16.3)
Study Study sample:sample: 9,694 incident cases
of colorectal cancer1995/2005
5,686 cases with 4 years of medical records
24,982 controlswith 4 years ofmedical records
3,460 cases withcolon cancer
2,226 cases withrectal cancer
Rel Risk of CRC with aspirin use by recency, duration and doseRel Risk of CRC with aspirin use by recency, duration and dose in UK GPRD (Garcia Rodriguez Epidemiology 2001)in UK GPRD (Garcia Rodriguez Epidemiology 2001)