SPARCtool Notes
SPARCtool Notes
back to SPARCtool
What’s new in Version 10.1?
Who created SPARCtool?
Peter Loewen B.Sc.(Pharm), ACPR, Pharm.D., FCSHP, R.Ph.
Associate Professor | Faculty of Pharmaceutical Sciences
David H MacDonald Professor in Clinical Pharmacy
Faculty, Collaboration for Outcomes Research & Evaluation (CORE)
Researcher, UBC Center for Cardiovascular Innovation (CCI)
The University of British Columbia | Vancouver Campus
2405 Wesbrook Mall | Vancouver, BC Canada V6T 1Z3
peter.loewen@ubc.ca | profile
Dr. Loewen has no financial or other conflicts of interest related to SPARCtool or the stroke prevention therapy options it does or does not include.
Where do the stroke risk estimates come from?
- CHADS2 & CHA2DS2-VASc estimates used in SPARCtool: Olesen et al. BMJ 2011;342:d124, 10 year followup estimates. This paper also standardizes the endpoint between CHADS2 & CHA2DS2-VASc, which is propagaged throughout SPARCtool. This endpoint is “stroke+thromboembolism”, or more specifically, “hospital admission or death due to thromboembolism, which includes peripheral artery embolism, ischaemic stroke, and pulmonary embolism”. SPARCtool uses “ischemic stroke+thromboembolism” as shorthand for this. Over 90% of thromboembolic events in AF are strokes. e.g., in ARISTOTLE, 8.9% of “ischemic stroke+systemic embolism” was systemic embolism.
- SPARCtool uses the linear predictor function from regression of Olesen et al. BMJ 2011;342:d124: risk=1.873 * score - 0.2076, with score = 0 fixed at 0.66 (the actual risk level in Olesen et al.) to prevent negative or zero risk estimates.
- CHADS2 vs. CHA2DS2-VASc: when using the individual risk scores, they have virtually identical predictive ability (c-statistic 0.781 vs. 0.777, respectively).
- Stroke+TIA risk without AF: Mitchell LB, et al. Heart. 2014;100(19):1524–1530. Added the Stroke & TIA rates from Table2. No risk stratification above CHA2DS2-VASc = 4 is provided. CHA2DS2-VASc preducts stroke in non-AF patients similarly well to AF patients. Of course, the risk estimates are much lower for non-AF patients. This was added to SPARCtool based on feedback that it is valuable for patients to see how much higher their stroke risk is because of AF vs. without it.
I read a different paper about CHADS2 or CHA2DS2-VASc and it shows different numbers than SPARCtool. Why?
Every validation study will produce different numbers for thromboembolism risk in every stratum of CHADS2 or CHA2DS2-VASc score. This is called the “calibration” of the prediciton model. The better the model, the more consistent the estimates will be between cohorts studied (i.e., better calibration). SPARCtool uses data from what is currently the largest and longest-term evaluation of the scores (see above).
HAS-BLED bleeding risk estimation
- validation studies: CHEST 2010; 138(5):1093–1100, J Am Coll Cardiol 2011;57:173–80.
- SPARCtool HAS-BLED risks are derived from pooling of these two validation studies, then using the linear predictor function from regression: risk=0.0175 * score+0.0064.
- although the maximum score is 9, no study has observed bleeding events in patients above score 6, but using the linear predictor function allows risk estimates anyway.
- the diagnostic performance of HAS-BLED is poor. Ann Hematol. 2011 Oct;90(10):1191–200
- using HAS-BLED (or any other bleeding risk estimation rule) is not superior to using “clinical judgement” to classify patients into low/intermediate/high risk for major bleeding. Am J of Med 2012;125(11):1095–110
Where do the efficacy and bleeding risks of the antithrombotic therapies come from?
- Warfarin: Cochrane Database of Systematic Reviews 2005, Issue 3. Art. No.: CD001927. DOI: 10.1002/14651858.CD001927.pub2, Ann Intern Med. 2007;146:857–867, and Cochrane Database Syst Rev 2005; CD001925.
Aspirin/ASA: Andersen LV, et al. Warfarin for the prevention of systemic embolism in patients with non-valvular atrial fibrillation: a meta-analysis. Heart. 2008;94(12):1607
Aspirin/ASA’s efficacy is imputed vs. warfarin’s (not vs. placebo) as much more reliable data for warfarin vs. aspirin/ASA and warfarin vs. no therapy is available than for aspirin/ASA vs. no therapy. SPARCtool uses the OR 0.33 for warfarin vs. antiplatelets (almost all of which was aspirin/ASA), converted to RR as 0.503. Aspirin/ASA’s major bleeding risk is imputed vs. warfarin’s for the same reasons.
Aspirin/ASA + clopidogrel: ACTIVE-A
The efficacy of aspirin/ASA+clopidogrel compared to placebo is imputed based on ASA vs. placebo RR 0.78, ACTIVE-A RR 0.72 = RR 0.56, RRR 44%. Annual major bleeding vs. placebo can be imputed based on being 0.7% higher per year than with ASA alone, which is 0.45%/year higher than no aspirin Derry & Loke = absolute 1.15%/year excess risk = NNH 87 x 1 year. However, this tool describes it as “same as warfarin”, which ACTIVE W showed directly.
Efficacy & safety of dabigatran: RE-LY & RE-LY update
110 mg bid: non-inferior to warfarin for stroke [RR 0.91 (0.74–1.11)], less major bleeding than warfarin [RR 0.8 (0.69–0.93)]. NNT (chance of harm) for major bleeding compared to PLACEBO is imputed by augmenting the estimated ARR for warfarin (1.79%) by 20% (corresponding to the RR of 0.8).
150 mg bid: superior to warfarin for stroke [RR 0.65 (0.53–0.82)]. Efficacy vs. placebo is imputed by multiplyng warfarin’s RR vs. placebo (0.33) by this effect to yield a RR of 0.21. Major bleeding similar to warfarin [0.93 (0.81–1.07)]
Efficacy & safety of rivaroxaban: ROCKET-AF
ROCKET-AF showed rivaroxaban to be non-inferior to warfarin with respect to stroke (HR 0.79 [0.66=0.96]). Debate exists about whether it was superior to warfarin as NI, on-treatment, and ITT analyses differed somewhat with respect to the significance of the HR0.8. Major bleeding was similar to warfarin (HR 1.04 [0.96–1.11])
Efficacy and safety of apixaban: ARISTOTLE
ARITOTLE showed apixaban to be superior to warfarin for stroke prevention (HR 0.79 [0.66–0.95]), and reduced mortality (0.89 [0.8–0.99]). Major bleeding was significantly less with apixaban (0.69 [0.6–0.8]), as was hemorrhagic stroke.
Efficacy and safety of edoxaban: ENGAGE AF
Neither edoxaban dose had superior efficacy vs. warfarin [HR 0.87 (0.73–1.04), HR 1.13 (0.96–1.34) for high and low-dose, respectively] using the using the most appropriate analysis, intention-to-treat. Both doses showed less hemorrhagic stroke than warfarin and major bleeding was significantly less with both doses [HR 0.80 (0.71–0.91), HR 0.47 (0.41–0.55) for high and low-dose, respectively].
How is the risk of major bleeding on “no therapy” derived?
This is patient-specific based on sex and age from the “baseline cohort” in Selak et al. JAMA. 2018;319(24):2507–2520. SPARCtool uses age>60 vs. age<60 from this study.
What about paroxysmal (intermittent) atrial fibrillation?
No trials of efficacy of antithrombotic therapy in patients with paroxysmal atrial fibrillation have been conducted. All major consensus guidelines suggest treating similarly to permanent/persistent/chronic atrial fibrillation. In reality, evidence here and here indicates stroke risk is probably somewhat proportional to the amount of time the patient is actually in AF, up to the maximums estimated by CHADS2/CHA2DS2-VASc.
Can these ANNUAL risks of stroke (or bleeding) be added together to estimate longer-term risks?
Not directly, but they can be used to estimate longer-term risk. If the annual risk of stroke or bleeding were constant over time (which they are not… they generally go up as age and other risk diagnoses accumulate), the formula to estimate the risk of a an event after t years is:
cumulative probabiliy of event = 1-[(1-annual probability of event)^ t]
e.g. in a patient with 15% annual SSE risk, their probabiliy of having an SSE by the end of year 3 is 1-[(1-0.15)exp(3)] = 39%
Because of the above, in the absence of treatment changes, these will be underestimates of risk.
Renal dysfunction
Why does SPARCtool extend the HAS-BLED bleeding risk to dabigatran, rivaroxaban, apixaban, and edoxaban?
I believe this is reasonable for two reasons: (1) Head-to-head trials so far show major bleeding risk to be similar between warfarin and new oral anticoagulants (RE-LY, ROCKET-AF, ARISTOTLE) and (2) HAS-BLED was validated in a population which included both warfarin and ximelagatran patients in the SPORTIF trials.
Has this system been evaluated in clinical practice?
Yes, in various forms:
Release Notes
version 10.1 release notes (October 2022)
Updated efficacy estimate for edoxaban 60mg to 13% RRR vs. warfarin per ENGAGE-AF superiority analysis. Brings the efficacy estimate method into line with that of the other DOACs where the superiority analysis was statistically significant (dabigatran 150mg, apixaban).
version 10 release notes (October 2022)
SPARCtool & SPARCmobile
- updated stroke and bleeding risk estimates for aspirin based on Andersen LV, et al. Heart. 2008;94(12):1607
- aspirin’s efficacy is now imputed vs. warfarin’s (not vs. placebo) as much more reliable data for warfarin vs. aspirin and warfarin vs. no therapy is available than for aspirin vs. no therapy. SPARCtool uses the OR 0.33 for warfarin vs. antiplatelets (almost all of which was aspirin), converted to RR as 0.503
- aspirin’s major bleeding risk is imputed vs. warfarin’s for the same reasons. Previously aspirin vs. no therapy data was based on non-AF studies. Andersen et. found it to be indistinguishable from warfarin.
- cleaned up grammar and typos in SPARCnotes (Thank you, users, for pointing them out.)
- replaced version date with “current as of [current month & year]” to reflect that the estimates provided are not obsolete. SPARCtool is updated rapidly when relevant new evidence emerges.
- removed “population average” bleeding risks. All bleeding risks are now individualized based on HAS-BLED, except “no therapy”, which is individualized from a non-OAC cohort (explained above)
- replaced “ASPIRIN” with “ASA” throughout
- fixed labels in graphs
- fixed explanation for cumulative risk over time
- added 1 decimal place to Y-axis labels in graphs
- added “not recommended by most international AF guidelines due to questionable benefit/bleeding risk ratio” for aspirin and aspirin+clopidogrel
- decapitalized generic drug names throughout
- changed layout and formatting of therapy option panels
- added Net Clinical Benefit to therapy panels and top table
version 9 release notes (June 2019)
SPARCtool
- NEW annual risk of bleeding on NO THERAPY udpated to be patient-specific using sex and age from the “baseline cohort” (most applicable) in Selak et al. JAMA. 2018;319(24):2507–2520
- NEW Net Clinical Benefit estimates. Estimates strokes/systemic embolisms prevented per major bleed caused. These are time-independent (i.e. they apply to any unit of time). Values <1 mean more major bleeds are caused than SSE prevented. Values >1 mean more SSE are prevented than bleeds caused.
- NEW stroke and bleeding risks are now based on the linear predictor function of the respective prediction model derivation/validation study rather than on the actual risks in each stratum in those studies. This improves the reliability of the risk estimates by eliminating sporadic values (e.g. bleed risk less with HAS-BLED score 5 than score 4) and allows risk predictions for scores above those seen in actual studies (e.g. HAS-BLED >5) by extrapolating.
SPARCmobile
- [12JUN19] updated to v.9 as above
- [coming soon] “Other Attributes” of each antithrombotic therapy to support shared decision-making conversations
version 8.2 release notes (Sept 2017)
SPARCtool
- Removed the US-specific edoxaban renal warnings as monographs differ on this worldwide. Replaced it with the same renal function warning as other NOACs, which is consistent with all edoxaban monographs.
- Added “Text for EMR” box for be cut-and-paste into EMR. Thank you, Mark McConnell, for this suggestion.
SPARCmobile
- Removed the US-specific edoxaban renal warnings as monographs differ on this worldwide. Replaced it with the same renal function warning as other NOACs, which is consistent with all edoxaban monographs.
- added an estimate for “Patient’s annual risk of stroke or TIA with no therapy if they did NOT have AF (for comparison):” based on Mitchell LB, Southern DA, Galbraith D, et al. Prediction of stroke or TIA in patients without atrial fibrillation using CHADS2 and CHA2DS2-VASc scores. Heart. 2014;100(19):1524–1530. doi:10.1136/heartjnl–2013–305303. CHA2DS2-VASc preducts stroke in non-AF patients simiarly well to AF patients. The risk estimates are much lower for non-AF patients, of course. This is added to SPARCtool based on feedback that it is valuable for patients to see how much higher their stroke risk is because of AF vs. without it.
version 8.1 release notes (June 2016)
SPARCtool
- added an estimate for “Patient’s annual risk of stroke or TIA with no therapy if they did NOT have AF (for comparison):” based on Mitchell LB, Southern DA, Galbraith D, et al. Prediction of stroke or TIA in patients without atrial fibrillation using CHADS2 and CHA2DS2-VASc scores. Heart. 2014;100(19):1524–1530. doi:10.1136/heartjnl–2013–305303. CHA2DS2-VASc preducts stroke in non-AF patients simiarly well to AF patients. The risk estimates are much lower for non-AF patients, of course. This is added to SPARCtool based on feedback that it is valuable for patients to see how much higher their stroke risk is because of AF vs. without it.
- made the bleeding bars in the graph less alarmingly red
- made the legend title for bleeding in the graph shorter
- added a “Which therapy options to HIDE?” function. You can use this to tailor discussions with patients to the options you consider viable for whatever reason. It applies to the table and the detailed risk estimates. It does NOT affect the bars shown in the graph [I will fix this in the future].
- added individual stroke and bleeding risk graphs
- added ability to hide either or both graphs
- moved the renal function warnings into the drug name bars to save space
version 8 release notes (June 2016)
SPARCtool
- removed CHADS2 entirely. Although CHADS2 and CHA2DS2-VASc perform similarly as prediction tools (Olesen JB, et al. BMJ 2011;342:d124–4), it is no more difficult to use the CHA2DS2-VASc criteria and using one or the other simplifies the layout without losing any predictive performance.
- added a new “Hypertension (SBP>160 mmHg) for HAS-BLED” since the HTN criteria for CHA2DS2-VASc and HAS-BLED are different.
- fixed the units on SCr to “mcmol/L”
- removed color fills for a cleaner look
- fixed the “Age” logic. Now it’s a 3-position radio-button.
- improved the layout of the “1 in X” statements throughout
- added a “Reset” button to clear all the settings
- removed “population average” values from bleeding table and graph wherever a HAS-BLED estimate exists
- reformatted all the input areas
- explicit definition of major bleeding
- warnings for HAS-BLED >6 of no reliable risk data and explicit “>X%” notation for HAS-BLED bleed risks when score is >6.
SPARCmobile
- corresponding changes as made to SPARCtool above
version 7 release notes (January 2015)
SPARCtool
- added edoxaban (using ENGAGE AF trial data)
- bar chart is now HTML (no Flash required)
- made patient factor criteria more explicit for renal and hepatic dysfunction
- improved browser layout
SPARCmobile
- added edoxaban (using ENGAGE AF trial data)
- made patient factor criteria more explicit for renal and hepatic dysfunction
version 6.21 release notes (April 2013)
SPARCmobile
- updated the disclaimer
- fixed truncated renal dysfunction warning for new OACs on iPhone and increased visibility of warnings.
- changed text of the warnings to “RENAL FUNCTION WARNING: Consult product monograph.” - dosage adjustment thresholds and stragtegies are evolving, so no specific advice is offered in the SPARCtool.
- lightened most of the fill colors
- updated dabigatran 150mg thromboembolism risk with updated RE-LY data (updated RR 0.66 −-> 0.65, RRR vs. warfarin from 78% to 79%) [Connolly SJ, Ezekowitz MD, Yusuf S, Reilly PA, Wallentin L. Newly Identified Events in the RE-LY Trial. N Engl J Med 2010;363:1875–6.]
- corrected major bleeding risk of ASA+clopidogrel to be the same as warfarin based on ACTIVE-W.
- added tabular view of the risks
SPARCtool
- updated the disclaimer
- changed text of the warnings to “WARNING: may be contraindicated or require dosage reduction depending on degree of renal dysfunction. Consult your country’s monograph.”
- dosage adjustment thresholds and stragtegies are evolving, so no specific advice is offered in the SPARCtool.
- lightened most of the fill colors
- updated dabigatran 150mg thromboembolism risk with updated RE-LY data (updated RR 0.66 −-> 0.65, RRR vs. warfarin from 78% to 79%) [Connolly SJ, Ezekowitz MD, Yusuf S, Reilly PA, Wallentin L. Newly Identified Events in the RE-LY Trial. N Engl J Med 2010;363:1875–6.]
- corrected major bleeding risk of ASA+clopidogrel to be the same as warfarin based on ACTIVE-W.
- added tabular view of the risks
- added graphical representation of the risks (requires Adobe Flash to view)
version 6.1 release notes (March 2013)
- edited mobile/standard links to map to new site (www.sparctool.com)
- edited sparcnotes links to map to new site and completely rewrote and reorganized SPARCtool Notes
- fixed date field width on regular version
- linked all notes and warnings to HAS-BLED the sparcnotes page
- updated CHADS2 and CHA2DS2-VASc risk estimates to use BMJ 2011;342:d124, 10 year followup estimates.
- upated all the CHADS2 risk descriptors to match the CHADS-VASc one based on BMJ 2011;342:d124, which is paraphrased as “ischemic stroke+thromboembolism”
- updated warfarin population average major bleeding risk from 2.6 to 3.8%/year based on CMAJ November 26, 2012 cmaj.121218
- add a reference for bleeding risk CPR vs. clinician judgement
- added a warnings for dabigatran, rivaroxaban, apixaban if “abnormal renal function” is checked in HAS-BLED
- added a disclaimer to top of both tools