ARTE - Dual versus single antiplatelet therapy after TAVI

in

Aspirin versus aspirin plus clopidogrel as antithrombotic treatment following transcatheter aortic valve replacement with a balloon-expandable valve: The ARTE (Aspirin Versus Aspirin + Clopidogrel Following Transcatheter Aortic Valve Implantation) randomized clinical trial. JACC: Cardiovascular Intervention 2017 [online]

Bottom line: In patients undergoing TAVI with no other indication for anticoagulation or DAPT, 3 months of DAPT increases the risk of major/life-threatening bleeding (number needed to harm 14) versus ASA alone without any apparent benefit.

 

Patients (n=222)

  • Included patients who underwent TAVI (aka TAVR) with Edwards SAPIEN XT or SAPIEN 3 valve (balloon-expandable valve)
  • Excluded patients requiring chronic anticoagulation or dual antiplatelet therapy (DAPT) for other indication, or those who had a major bleed within 3 months before TAVI or any history of intracranial hemorrhage
  • Baseline characteristics
    • Age 79 y
    • Male 58%
    • Indexed AVA 0.4 cm2/m2
    • LVEF 55%
    • Procedural characteristics
      • Femoral approach ~70%
      • SAPIEN XT ~92%, SAPIEN 3 8%
      • Post-TAVI indexed AVA 0.95-1.0 cm2/m2
      • New-onset AF 11%

Interventions

  • I: DAPT - clopidogrel (300 mg/d load <1 day before/after TAVI, then 75 mg/d) x90 days
  • C: ASA monotherapy
  • In both groups, ASA started >24h before TAVI & continued for 6+ months
  • PPI use not reported (though authors report that all patients with GI bleed were receiving a PPI prior to the event)

Results @ 90 days

  • Types of major/life-threatening bleeds in DAPT group (all occurred in first 30 days): GI bleed (5), peri-operative or access-site bleed (annular rupture, femoral hematoma; 7)

 

Generalizability

  • This trial represents elderly patients undergoing TAVI with moderate-high risk of peri-operative mortality
    • Although the results of this trial do not apply to those requiring anticoagulation, we would expect an even greater increase in serious bleeds when adding DAPT to an oral anticoagulant
  • These results with the SAPIEN valves are also likely generalizable to self-expanding valves (Medtronic's CoreValve line)

Internal validity

  • Unclear risk of allocation bias (allocation concealment not adequately described)
  • High risk of performance & detection bias (no blinding to treatment)
  • Stopped prematurely due to slow enrolment + loss of funding

 

Other studies

  • Results from our systematic review of 4 small studies (2 RCTs & 2 cohorts) were consistent with the ARTE trial; DAPT increased the risk of bleeding events without significantly reducing stroke or other ischemic events
  • Notably, none of the studies - alone or in combination - have sufficient power to evaluate the efficacy of DAPT in this setting, though it is likely that the severity and magnitude of harm from bleeds related to DAPT likely exceeds any possible benefit in this patient population
  • Therefore, the existing evidence does not support routine use of DAPT in patients undergoing TAVI

GAUSS-3 - Evolocumab vs ezetimibe in true muscle-related statin intolerance

in ,

Efficacy and tolerability of evolocumab vs ezetimibe in patients with muscle-related statin intolerance: The GAUSS-3 randomized clinical trial. JAMA 2016;315:1580-90

Bottom line:

  • ~43% of patients with perceived statin-related muscle symptoms had intolerance reproducible with a N-of-1 trial;

  • In those with muscle-related statin intolerance reproducible with a N-of-1 trial, evolocumab & ezetimibe were similarly tolerated;

  • LDL-C reductions with these agents were consistent with those from other trials with LDL-C reductions of 50-55% for evolocumab & 15-20% for ezetimibe.

 

Patients (Phase A n=491, Phase B n=218)

  • Included
    • Adults unable to tolerate atorvastatin 10 mg/d & any other statin (any dose) or 3+ statins
    • Baseline LDL-C
      • >2.6 mmol/L + CAD
      • >3.3 mmol/L + 2 CV risk factors
      • >4.1 mmol/L + 1 CV risk factor
      • >4.9 mmol/L (at least possible familial hypercholesterolemia [FH])
  • Baseline characteristics (of Phase B patients)
    • Age 59 y
    • Male 51%
    • CV hx: CAD 31%, cerebrovascular disease/PAD 20%
    • Hx of intolerance to at least 3 statins 82%
    • Worst muscle-related adverse effects: Myalgias 80%, myositis 14%, rhabdomyolysis 6%
    • Mean LDL-C 5.7 mmol/L

Interventions

  • Phase A (confirming statin-related muscle symptoms)
    • I: Atorvastatin 20 mg/d x10 weeks
    • C: Matching placebo x10 weeks
    • Note: Preceded by 4-week washout without any lipid-lowering therapy
  • Phase B (comparison of non-statin lipid-lowering monotherapy for those with reproducible statin-related muscle symptoms in Phase A)
    • I: Evolocumab 420 mg subcutaneously q1 month (+ ezetimibe placebo)
    • C: Ezetimibe 10 mg daily (+ evolocumab placebo)

Results

Phase A: Muscle symptoms with

  • Atorvastatin but not placebo (truly statin-related muscle symptoms) 43%
  • Placebo but not atorvastatin: 27%
  • Both atorvastatin & placebo 10%
  • Neither 18%

Phase B

  • Total muscle-related events: Evolocumab 20.7%, ezetimibe 28.8%, p=0.23
    • Myalgia: 13.8% vs 21.9%
    • Elevated CK: 2.8% vs 1.4%
  • LDL-C reduction
    • Evolocumab lowered by ~53% (-2.7 mmol/L) from baseline
    • Ezetimibe lowered by 17% (-0.8 mmol/L) from baseline
    • ~37% (1.9 mmol/L) difference between groups
    • Maximal LDL-C reduction achieved at ~4 weeks & maintained during 6-month follow-up

 

Considerations (generalizability, internal validity, etc)

  • Low risk of bias (allocation, performance, detection, attrition) in both phases due to computer-generated randomization sequence with allocation concealed by centralized allocation and blinding of patients and outcome assessors using matching placebos
  • Phase A of this trial is generalizable to our patients who have a history of perceived intolerance to numerous statins
    • The Phase A results indicate that many of these patients can tolerate a statin with rechallenge, particularly if bias is minimized by way of a N-of-1 double-blind trial design. However, up to 43% of these patients have true statin-related myalgia that is reproducible with a N-of-1 trial;
    • Given the high cost of PCSK9 inhibitors, this raises the question of whether it would be cost-effective to perform N-of-1 trials in patients with history of statin intolerance if it allowed us to get 53% of them back onto a statin rather than a more expensive (and in the case of ezetimibe monotherapy at least, inferior) lipid-lowering therapy?
  • Generalizability of Phase B is limited by the fact that most of us cannot perform N-of-1 trials routinely. Consistent with clinical practice however, it does indicate that some of these of these patients will go on to report muscle symptoms while receiving an alternate lipid-lowering agent and even discontinue these agents. Since Phase B of this trial did not have a placebo group, this cannot show that either of these drugs were truly responsible for the muscle symptoms.

TOPIC - Continuing ticagrelor/prasugrel vs switching to clopidogrel 1 month after ACS

in

Benefit of switching dual antiplatelet therapy after acute coronary syndrome: The TOPIC (timing of platelet inhibition after acute coronary syndrome) randomized study. Eur Heart J 2017 [online]

Bottom line: In patients with ACS who underwent PCI & tolerated prasugrel/ticagrelor x1 month, discontinuing prasugrel/ticagrelor & switching to a pill that combines ASA+clopidogrel reduced the risk of mainly minor bleeding over the subsequent 11 months.

TOPIC was underpowered to evaluate ischemic events, & is inadequate to inform practice or change routine care. Based on results from PLATO, employing this strategy would result in an increased risk of death/MI/stroke, particularly during the crossover period between stopping ticagrelor & waiting for clopidogrel to achieve its full antiplatelet effect (up to 7 days).

 

Patients (n=645)

  • Included: 
    • Adults with ACS, & PCI within 3 days of ACS admission initially treated with prasugrel/ticagrelor
    • No major adverse event in 1st month (including ischemia or bleed)
  • Key exclusion: Major bleed (BARC criteria) in last year, long-term use of anticoagulation
  • Baseline characteristics
    • Age ~60 y
    • Male ~83%
    • ACS type: STEMI ~40%, NSTE-ACS 60%
    • Previous CAD 30%
    • CV risk factors: Smoker ~45%, HTN ~50%, T2DM ~25%
    • Meds
      • P2Y12 inhibitor: Prasugrel ~55%, ticagrelor ~45%
      • ACE inhibitor or ARB ~75%
      • Beta-blocker ~70%
      • Statin ~95%
      • PPI 100%

Interventions

ASA 75 mg/d + ticagrelor/prasugrel (clinician selection) x1 month, then randomized to:

  • I ("switch"): Switch to clopidogrel 75 mg/d + ASA 75 mg/d (combined in 1 pill) x11 months (total 1 year on DAPT)
    • The published report does not specify what strategy was used to perform the switch to clopidogrel. This is critical as clopidogrel 75 mg/d takes 5-7 days to achieve its full antiplatelet effect
    • 86% still receiving at 1 year (~6% switched back to ticagrelor/prasugrel, ~6% on ASA only, <1% on no antiplatelet)
  • C ("continue"): Continue ticagrelor/prasugrel + ASA 75 mg/d (not combined into 1 pill) x11 months (total 1 year on same P2Y12 inhibitor)
    • 75% still receiving at 1 year (~17% switched to clopidogrel, ~7% on ASA only, <1% on no antiplatelet)

Results @ 1 year

 

Generalizability

  • P: This trial applies to ACS patients who tolerated ticagrelor/prasugrel for 1 month without issue, including recurrent coronary event or clinically-important bleeding
  • I/C:
    • The "switch" intervention was actually 2 distinct interventions:
      1. Consolidation of multiple pill administrations into a single tablet
        • For initial ticagrelor users, simplified from 3 tabs/day to 1
        • For initial prasugrel users, simplified from 2 tabs/day to 1
      2. Switch to lower-potency antiplatelet agent (clopidogrel)
      • Reduced bleeding rates in this trial are a combined result of decreased antiplatelet potency, as well as more consistent adherence due to decreased pill burden and simplification of administration regimen. This also reduces any apparent benefit on ischemic outcomes with the prasugrel/ticagrelor continuation, since non-adherence dulls their efficacy
    • Use of either prasugrel or ticagrelor in the control group complicates this further, as these agents have different efficacy/safety profiles (ticagrelor generally considered more efficacious + safer), & administration methods (ticagrelor is dosed BID)

Internal validity

  • Use of inappropriate composite primary outcome
    • Includes biologically disparate events (both ischemic/efficacy & bleeding/safety events)
    • Largely driven by the least "clinically important" outcome (bleeding requiring medical evaluation)
    • Inadequate power to evaluate more important components (CV/overall death, stroke, revascularization), & cannot rule in/out a difference nor determine consistency between different components
  • Enrolment occurred in a single center in France 2014-2016
    • Single-center trials systematically overestimate benefits compared to multi-center trials, likely related issues in minimizing typical bias domains (allocation, performance & detection)
  • Unclear risk of allocation bias
    • Allocation concealed by sequentially-numbered sealed envelopes: Application of this strategy varies, and is prone to manipulation
  • High risk of performance and detection bias
    • Patients, caretakers and clinicians not blinded to study intervention
    • "Switch" intervention simpler, easier to adhere to
    • Adjudicated by physician unaware of allocated intervention, though reporting of events to adjudicators may have been biased (particularly the subjective BARC 2 bleeds)
  • Low risk of attrition bias (intention-to-treat analysis with 2% loss-to-follow-up at 1 year)
  • Potential selective outcome reporting
    • Reporting of coronary events limited to those "requiring unplanned hospitalization for urgent coronary revascularization," but investigators did not report ACS events not requiring revascularization

LEADER - Liraglutide in diabetes with high CV risk

in ,

Liraglutide and cardiovascular outcomes in type 2 diabetes. NEJM 2016;375:311-22.

Bottom line: In patients with T2DM & high CV risk (~4%/year), liraglutide reduced the risk of death, MI or stroke (NNT ~50) over ~4 years versus placebo.

Notes:

  • Liraglutide's efficacy is not clearly due to its antihyperglycemic effect, but it should likely be considered as a 3rd-line agent (after metformin & empagliflozin) for patients with T2DM above their individualized A1c target;

  • Small trials raise concern for CV safety of liraglutide in HFrEF. I will personally avoid using this drug (or any GLP-1 agonist) in HFrEF until better data shows that it's safe.

 

Patients (n=9340)

  • Included
    • Type 2 diabetes (T2DM) with A1c 7% or more
    • Either
      • Age 50 y/o or greater +
        • CAD (no ACS <14 days of enrolment)
        • Cerebrovascular disease (no stroke/TIA <14 days of enrolment)
        • Peripheral vascular disease
        • HF NYHA functional class 2-3
        • CKD stage 3-5
      • Age 60 y/o or greater +
        • Micro- or macroalbuminuria
        • HTN + LVH
        • LV systolic or diastolic dysfunction
        • ABI <0.9
  • Baseline characteristics
    • Age 64 y
    • Male 64%
    • Diabetes duration ~13 y
    • PMHx
      • MI 30%
      • Stroke/TIA 16%
      • Prior revascularization 39%
      • Symptomatic CAD 9%, documented asymptomatic cardiac ischemia 26%
      • HF 14%
      • >60 y/o with micro/macroalbuminuria (11%), HTN+LVH (5%), ABI<0.9 (<3%)
    • Clinical variables
      • Wt 92 kg
      • BMI 32.5
      • BP 136/77
    • Labs
      • A1c 8.7%
      • eGFR 30-59 (21%), <30 (<3%)

 

Interventions

  • I: Liraglutide (started at 0.6 mg subcutaneously daily, uptitrated as tolerated to 1.8 mg)
    • Median dose 1.78 mg/d
  • C: Placebo
  • Patients in both groups received standard-of-care therapy to lower BP (<130/80), glucose (A1c <7%) & lipids (statin for all patients; LDL-C <2.6 mmol/L if no ASCVD, <1.8 mmol/L if ASCVD)
    • Patients in liraglutide generally had a negligibly fewer initiation medications for these risk factors during trial (e.g. statins initiated in 11% on placebo vs 9.4% on liraglutide), except for insulin initiation (started in 43% on placebo vs 29% on liraglutide)
  • Median duration of therapy 3.5 years

 

Results @ median 3.8 years

  • Composite renal outcome: 15% vs 19%, HR 0.78 (0.67-0.92)
    • Death due to renal disease: 0.4% vs 0.3%, HR 1.59, (0.52-4.87)
    • Dialysis start: 3.1% vs 3.4%, HR 0.87 (0.61-1.24)
    • Persistent doubling of SCr: 4.9% vs 5.5%, HR 0.89 (0.67-1.19)
    • New-onset persistent macroalbuminuria: 0.9% vs 12.1%, HR 0.74 (0.60-0.91) - the least important component of the composite renal outcome, & the only one driving the significant difference
  • Safety
    • Serious adverse events: Liraglutide 49.7%, placebo 50.4%
    • Any adverse event: 62.3% vs 60.8% (p=0.12)
    • Severe hypoglycemia: 2.4% vs 3.3% (p=0.02)
    • Acute gallstone disease (cholecystitis or cholelithiasis): 3.1% vs 1.9% (p<0.001)
    • Acute pancreatitis: 0.4% vs 0.5%
    • Pancreatic carcinoma: 0.3% vs 0.1% (p=0.06)
  • Surrogates
    • @ 3 months: Liraglutide ~7.2%, placebo ~8.2%; difference decreased over time (A1c 0.4% lower with liraglutide on average through trial)
    • Wt 2.3 kg lower with liraglutide vs placebo
    • HR 3 bpm higher with liraglutide vs placebo

 

Generalizability

  • This trial represents a population of patients with long-standing diabetes (~13 years) with suboptimal glycemic control (baseline A1c 8.7%) at very high CV risk (primary outcome rate ~4%/year with placebo)
    • The relative risk reduction for the primary outcome components are likely generalizable to a lower-risk population, though this will ultimately lead to a lower absolute benefit (larger NNT) & perhaps no/minimal reduction in death
  • This trial did not evaluate a "more vs less intensive" glycemic control; it randomized patients with diabetes to liraglutide, which lowers glucose, among many other things, or placebo
    • The liraglutide had better glycemic control than placebo by ~1% at 3 months, but the trial's glycemic control protocol later resulted in much smaller differences in A1c between group
    • The benefit of liraglutide in this setting may be related to improved glycemic control, weight loss, other improvements in metabolic parameters, or to a yet-undefined mechanism
  • Heart failure
    • Only 14% of patients enrolled in this trial had HF, which was not limited to patients with reduced ejection fraction/LV dysfunction. Two small trials raise concerns about the use of liraglutide in individuals with HF:
      • FIGHT: This double-blind RCT of 300 patients evaluated the effect of liraglutide vs placebo in patients with HFrEF (HF with LVEF 40% or lower) hospitalized for HF in the previous 2 weeks on clinical outcomes & LVEF, with or without T2DM. After 6 months, liraglutide did not improve the primary outcome, LVEF, distance on a 6-minute walk test, quality of life, or NT-proBNP. There was, however, a concerning increase in the secondary outcome of death, re-hospitalization/ED visit for CV reasons (liraglutide 63%, placebo 55%, HR 1.34 [1.00-1.80])
      • LIVE: Similarly, this double-blind RCT of 241 patients evaluated the effect of liraglutide vs placebo in patients with stable HFrEF (<45%) NYHA class 1-3 on LVEF. Liraglutide did not significantly improve LVEF or any other echocardiographic measurement vs placebo, but it did increase the risk of cardiac serious adverse events (10% vs 3%, p<0.05). Notably, this was a heterogeneous collection of outcomes that included VT death, VT, AF requiring DC cardioversion, ACS, worsening HF.

Internal validity

  • Risk of bias
    • Low risk of allocation & detection bias
      • Allocation concealment maintained by computerized central allocation
      • Blinding with matching placebo
      • All macrovascular & microvascular outcomes adjudicated by company blind to assigned study intervention
    • Low risk of performance bias
      • Blinding with matching placebo
      • Standardized protocols to control CV risk factors
      • More patients in the placebo group received intensification of medications to reduce CV risk factors (likely because liraglutide produced mild improvements in metabolic parameters), particularly antihyperglycemic agents such as insulin, though this should produce a "conservative bias" towards the null in favor of the placebo group
    • Low risk of attrition bias
      • Analyzed by intention-to-treat
      • Low loss-to-follow-up (3.2% did not complete trial follow-up; 0.3% for mortality)
  • Use of composite primary outcome was completely appropriate
    • All components important (even silent MI is prognostically important)
    • Biologically rational to combine these macrovascular outcomes with overlap in pathophysiology
    • Similar contribution from each component of the composite outcome; MI and stroke both had estimates of effect nearly identical to that of the composite (HR ~0.86), making it appropriate to conclude that liraglutide reduced the risk of the composite, as well as each of the individual components
  • 2-week placebo run-in period to exclude early non-adherence to subcutaneous injections
    • Not a source of bias, but does exclude those individuals least likely to be adherent long-term
  • Per FDA guidance, designed to first prove non-inferiority trial of liraglutide vs placebo (to ensure CV safety to avoid issues as seen with other diabetes drugs rosiglitazone), then - if safe - to show superiority 
    • Not a source of bias; pre-defined & appropriate

CONCERN: Naproxen vs celecoxib in patients with recent upper GI bleed with an indication for low-dose ASA

in

Chan FKL, et al. Gastrointestinal safety of celecoxib versus naproxen in patients with cardiothrombotic diseases and arthritis after upper gastrointestinal bleeding (CONCERN): an industry-independent, double-blind, double-dummy, randomised trial. Lancet [epub ahead of print]

Bottom line:

  • Issues in reporting of this trial preclude a fully-informed interpretation of its results

  • In patients receiving a PPI & ASA, recurrent upper GI bleed occurred in ~6% of patients also receiving daily celecoxib at a dose of 100 mg BID compared to ~12% of those receiving naproxen 500 mg BID over 18 months.

 

Patients (n=514)

  • Included
    • Presenting with upper GI bleed while on NSAID + low-dose ASA
    • Follow-up endoscopy confirming ulcer healing 8 weeks after GI bleed
    • Negative for H pylori or successful H pylori eradication
    • Ongoing indication for low-dose ASA ("cardiothrombotic diseases" or "multiple CV risk factors")
    • Chronic arthritis pain anticipated to require regular NSAID use
  • Key exclusion criteria
    • Erosive esophagitis
    • Gastric outlet obstruction
    • Renal failure (SCr >200 umol/L)
    • Use of anticoagulants
  • Typical study patient
    • Age 72 y (>80 y in 24%)
    • Female 46%
    • Indication for NSAID: OA 70%, RA 3%, other 27%
    • Baseline upper GI bleed
      • Gastric origin 55%
      • Required endoscopic therapy ~30%
      • Required blood transfusions 55%
    • ASA indication: Secondary prevention (67%), primary prevention with 10-year CV risk >10% (33%)

Interventions x18 months

  • I: Celecoxib 100 mg PO BID
  • C: Naproxen 500 mg PO BID
  • ~90% in both groups took at least 70% of their doses; ~18% in each group discontinued the study drug before 18 months
  • Co-interventions:
    • Esomeprazole 20 mg PO once daily in all patients
    • ASA 80 mg/d (discontinued in ~30% of patients during trial in both groups)

Results @ median 18 months

  • Efficacy (not designed or powered to evaluate this)
    • Patient's global assessment of disease activity scale (range 1-5, lower=better):
      • Baseline: 3.0 in both groups
      • Month 18: Celecoxib 2.1 versus naproxen 2.3, p=0.386 for difference between groups
    • Quality of life, pain, use of other analgesics: ?
  • Safety
    • Primary outcome - recurrent upper GI bleed (hematemesis/melena or hemoglobin drop >20 g/L+ endoscopically-confirmed ulcers or bleeding erosions): Celecoxib 5.6%, naproxen 12.3% (NNT 15), hazard ratio (HR) 0.44 (0.23-0.82)
      • Fatal recurrent GI bleed: 0 in both groups
    • Major CV event (vascular death, MI or stroke): Celecoxib 4.4% vs naproxen 5.5%, HR 0.78 (0.36-1.73)
  • Discontinued treatment before 18 months: Celecoxib 18% vs naproxen 19%

Subgroup analysis (reported in supplement of authors' reply to letter to the editor)

  • Primary outcome
    • Continued ASA: Celecoxib 7.2%, naproxen 13.8% (p=.042)
    • Discontinued ASA: 1.4% vs 8.4% (p=.054)
  • CV events: 
    • Continued ASA: Celecoxib 5.5% vs naproxen 6.0% (p=.825)
    • Discontinued ASA: 1.5% vs 4.3% (p=.312)

Internal validity

  • Unclear/high risk of allocation bias
  • Unclear risk of performance/detection bias
    • Proper blinding dependent upon allocation concealment. If sealed opaque envelope method was subverted, then blinding would also be compromised
    • Blinding by "double-dummy" pills not prepared by manufacturer
    • GI & CV outcomes adjudicated by committee blinded to allocated treatment group
  • Attrition bias
    • Premature discontinuation occurred in ~18% of patients in each group
    • Modified intention-to-treat analysis for GI safety outcome only counted outcomes that occurred "during treatment," therefore excluding any outcomes occurring after premature discontinuation

Generalizability

  • Study conducted in single centre in Hong Kong, with enrolment taking ~10 years (2005-2015)
    • Many changes to GI bleed management (e.g. reduced transfusion threshold) & CV disease management (including ASA prescribing patterns in primary prevention). No reporting or comments on such trends in this publication, which complicates application of these trial results.
  • Many aspects of this trial remain unreported, which further complicates application of these results
    • CV risk & indication for ASA?
      • This is the key reporting issue in this trial. The authors do not clearly describe the indication for ASA beyond "cardiothrombotic diseases or multiple coronary risk factors", and do not describe these in their 'baseline characteristics' table. Notably, the authors report that ~30% of patients discontinued ASA during the trial, suggesting that some of these patients had no clear indication for ASA
      • Addendum: The authors reported in a response to letter to the editor that 67% received ASA for secondary prevention, with the remaining 33% receiving ASA for primary prevention in the context of a calculated 10-year risk of a CV event >10%. The benefit of ASA for primary prevention is minimal, & the harms of continuing ASA (especially in these patients who already developed a GI bleed on ASA) likely outweighed any benefit. Furthermore, reported subgroup analyses showed that those who discontinued ASA had a lower risk of CV events, suggesting that mostly the lower-risk primary CV prevention patients discontinued ASA during the trial. 
      • Use of dual antiplatelet therapy (DAPT) was not specifically prohibited during this trial, though its use was not reported
    • Time from initial upper GI bleed to initiating of study NSAID?
      • Addendum: The authors later reported that endoscopy was routinely performed to confirm ulcer healing 8 weeks after the GI bleed, at which point patients were enrolled into this trial.
    • Previous history of NSAID use and efficacy (i.e. previous treatment failures with either study NSAID)?
  • Effect of giving no NSAID or NSAID PRN?
    • The majority of patients with OA who use oral NSAIDs in practice do so on a PRN basis. The GI risk related to this kind of usage pattern is likely lower, and may be a reasonable alternative
    • In a previous trial by the same authors of patients with ASA-related upper GI bleed, recurrence using the same definition was <1% over 1 year in patients receiving ASA+PPI
  • Clinical importance of primary outcome?
    • The primary outcome of this trial was recurrent endoscopically-confirmed upper GI bleed that presented as either hematemesis/melena or a >20 g/L hemoglobin drop. None of the recurrent GI bleeds in this trial resulted in patient death, and most would be classified as "minor bleeds" under most commonly-used bleeding definitions. As such, impact of these recurrent events on quality of life would be a key domain to measure (but it wasn't).
    • If we accept that this is a clinically-important outcome, the risk remains far too high even with celecoxib, particularly given that only 90% took at least 70% of their doses and that 30% discontinued ASA during the trial.