Risk of myocardial infarction on Cox 2 & conventional NSAIDs
Julia Hippisley-Cox Carol Coupland
Goals of presentation
• Overall
– Present analysis of study examining risk of MI in patients on NSAIDS
• Initially
– Overview of data source (QRESEARCH) used for the project to set context
Acknowledgements
Co-author
– Carol Coupland
QRESEARCH team
– Mike Pringle – Mike Heaps – Justin Fenty – Gavin Langford – David Stables – EMIS and EMIS practices
QRESEARCH: What is it?
• • • • • Patient level consolidated database Anonmyised data Longitudinal data for 15+ years Derived from GP clinical records Validated against external and internal measures • Industry independent
QRESEARCH: UK coverage
• Largest database in the world • 488 UK practices • > 6 practices in every Strategic Health Authority (administrative area) • > 8 million patients including those who died, left and still registered • > 29 million person years observation
QRESEARCH: Data contents
• • • • • • • Socio-demographics Diagnoses Clinical values Laboratory tests Prescription data Consultation data Category of clinician entering data
Study aims
• To determine risk of Myocardial Infarction in patients taking –use Cox 2 inhibitors –traditional NSAIDS –Compared with non use
Background
• Cox 2 drugs
– Type of painkiller – Supposed to cause less gastrointestinal problems than traditional NSAIDS
• But original RCT compared rofecoxib with naproxen
– Found 5 fold increased risk of MI – Adverse effect of Rofecoxib – Beneficial effect of naproxen
Study design & setting
• Nested case control study • 367 UK practices contributing to the new UK QRESEARCH • Registered population > 3 million patients [6% UK population] • Study period Aug 2000-July 2004
Study cohort
• All patients aged 25-100 • Registered with practices from 1st Aug 2000 to 31st July 2004 • Prior registration >12 months • > 8 million person years observation • 9218 patients with a first MI
Incidence of myocardial infarction per 1000 person years
In 367 practices in QRESEARCH (August 2000 to July 2004)
0.00
25-29
2.00
4.00
6.00
8.00
10.00
35-39
45-49
55-59
Ageband
65-69
75-79
85-89
males 95% CI
females
Case & controls
CASES • 1st ever MI during 4 year study period CONTROLS • 10 controls matched by
– age – Sex – Practice – Calendar year
• Cases & controls all with 3 years + prior data
Exposure
Main exposure • Use of any NSAID in 3 years prior to index date Extracted • Type • Date • Count of prescription
Types of NSAIDS
• Overall 27 various NSAIDS • Groups for analysis
– Celecoxib – Rofecoxib – Other Cox2 – Diclofenac – Ibuprofen – Naproxen – Other NSAIDs
Confounders
• • • • • • • • • • • Ischaemic heart disease Diabetes Hypertension OA Rheumatoid arthritis Smoking Obesity Deprivation Aspirin Statins Antidepressants
Statistical analysis
• • • • Conditional logistic regression Odds ratios + 95% CI Unadjusted & adjusted Investigations for interactions
Results: baseline
• Cases & controls well matched for • Age, sex, time, practice • Cases had higher prevalence
– Smoking – Obesity – Deprivation – comorbidity
Increased risk of MI (95% CI)
• Increased risk of MI current use of
– Rofecoxib 30% increase (10%-60%) – Other Cox 2 27% increase (0%-60%) – Diclofenac 55% increase (40%-70%) – Ibuprofen 20% increase (10%-40%)
• V significant duration response
– More scripts causing higher risk
Numbers needed to harm (aged 25 plus)
• No. patients treated for one year for one additional MI to result
– Diclofenac 1066 (95% CI 815 -1504) – Rofecoxib 1833 (955% CI 961-6517) – ibuprofen 2444 (95% CI 1504-5332)
Numbers needed to harm (patients 65 plus)
• No. patients 65 + treated for one year for one additional MI to result
– Diclofenac 521 (95% CI 355 to 866) – Rofecoxib 695 (95% CI 344 to 3841) – Ibuprofen 1005 (95% CI 569 to 3089)
Interactions, age, naproxen
• No convincing interactions
– Aspirin & NSAIDS – CHD and any NSAID
• So risk applies despite presence of these • No evidence that Naproxen lowers risk MI
Limitations study
• Observational study • Subject to bias
– Misclassification – Recording bias
• Confounding by indication
– But likely to have applied to all NSAIDS
• Residual confounding
Overcoming confounding by indication
• To address confounding by indication restricted analysis to those without
– Diabetes – Ischaemic heart disease
• Also restricted analysis aged > 65 years • This did not change results
Setting it in context
• Since this work was completed
– Rofecoxib withdrawn – Valdecoxib withdrawn – Celecoxib risk
• FDA memo on web
– Suggests class effect not just Cox 2 but also NSAIDs but serious lack of placebo controlled safety data
However
• Observational studies cannot prove anything • They are subject to residual confounding • They can only raise concerns • Sometimes observational studies are wrong eg HRT • Rapid access to very large representative samples & potential • Can help evaluate risks & benefits outside RCT setting
Conclusions
• Our study suggests risk of MI is a class effect as conventional NSAIDs also seem to increase risk • FDA memo also suggests this • ? Time to evaluate cardiovascular safety all NSAIDs