Post on 27-May-2020
The Incidence of and Risk Factors for Thromboembolism in Older Patients with
Newly Diagnosed Multiple Myeloma
Ang Li, MD, MSActing Instructor
Division of HematologyUniversity of Washington
May 3rd 2019
Introduction
• Immunomodulatory drugs (IMID) have become the standard of care for multiple myeloma (MM) treatment
• Thrombosis remains high especially in older patients• Unclear if thrombotic rates differ between lenalidomide vs. thalidomide1
• Existing NCCN guideline to prevent IMID associated VTE:• 18 risk factors from expert consensus2
• Not validated, poor adherence3
1JTH 2011;9:653. 2NCCN VTE Guidelines 2018. 3J Oncol Pharm Pract 2016;22:248
Goals
• Aim 1: To assess differences in venous thromboembolism (VTE), arterial thromboembolism (ATE) for lenalidomide vs. thalidomide in older patients
• Aim 2: To assess the performance of existing NCCN VTE risk prediction consensus; to derive and validate a simplified clinical risk model
5/3/19 3
Myeloma – A Disease of Older Adults
Study Design
• Retrospective cohort study:• SEER-‐Medicare database• Followed from 1st IMID prescription until clinical event or death• Censored for disenrollment or end of study• ICD-‐9-‐CM codes for demographic, comorbidity, medication
• Outcome definitions:• VTE: 1 inpatient claim OR (≧2 outpatient claims >30d and <365d
and anticoagulant prescription within 90 day)• ATE: 1 inpatient claim of myocardial infarction OR stroke• ICD-‐9-‐CM codes have positive predictive values of 75-‐95%1
1Thromb Res 2010;126:61, Am Heart J 2004;148:99, Stroke 2014;45:3219
Patient Selection
Lenalidomide = 78% (n=1863)Thalidomide = 22% (n=534)
Aim 1: Incidence of Thrombosis in Lenalidomide vs. Thalidomide
Statistical Methods
• Inverse propensity score weighting for confounders:• Demographics: age, sex, race• Co-‐morbidities or risk factors• Concurrent chemotherapy, dexamethasone, anticoagulant• Year of diagnosis: surrogate for advances in treatment
• VTE/ATE: weighted cumulative incidence curves and Fine-‐Gray subdistribution hazards models (SHR) with death as competing risk
• Variance: estimated via 200 bootstraps1
1Stat Med 2016;35:5642
Before Propensity Score Inverse Weighting After Propensity Score Inverse WeightingLenalidomide Thalidomide Std Diff Lenalidomide Thalidomide Std Diff
Demographic No major differences <0.15 Refer to abstract <0.10Comorbidity No major differences <0.15 Refer to abstract <0.10Age, mean 75.0 75.9 .135 75.2 75.2 -‐.003Year of Diagnosis, %2007 4.6 21.9 .529 8.5 8.4 .0022008 8.7 27.9 .512 12.8 13.1 -‐.0082009 10.8 17.4 .191 12.3 12.6 -‐.0112010 14.3 14.0 -‐.007 14.2 14.4 -‐.0062011 16.9 9.0 -‐.236 15.2 14.8 .0102012 20.9 5.4 -‐.471 17.5 19.5 -‐.0532013 23.9 4.3 -‐.586 19.5 17.1 .063Chemotherapy, %None 73.4 80.0 .155 74.9 75.0 .003Bortezomib 22.6 12.2 -‐.278 20.3 20.4 .003Non-‐bortezomib 4.0 7.87 .165 4.8 4.5 -‐.012Dexamethasone, %None 22.9 40.8 .391 26.8 27.0 .001Low Dose (<120 mg) 14.7 9.9 -‐.146 13.7 14.4 .038Standard Dose (120-‐160 mg) 43.4 27.0 -‐.349 39.7 39.1 -‐.019High Dose (160-‐480) mg 13.9 13.9 -‐.001 13.9 13.6 -‐.012Very High Dose (480+ mg) 5.1 8.4 .133 5.9 6.0 -‐.001Anticoagulant, %None 89.2 81.5 -‐.220 87.4 87.7 -‐.021Warfarin 9.1 15.4 .191 10.6 11.2 -‐.018LMWH/DOAC 1.6 3.2 .099 2.0 2.1 .009
Incidence of Thrombosis for Lenalidomide vs. Thalidomide
VTE ATE
Aim 2: Risk Model Assessment
NCCN guidelines 1.2018
Statistical Methods
• Model derivation:• Cohort: SEER-‐Medicare database• Variable selection: Cox regression backward stepwise selection
• Model validation:• Cohort: Veterans Health Administration database• Collaboration with investigators at Washington University in St. Louis• Discrimination: HR (p-‐value), Harrell’s c index (bootstrap)• Calibration: Kaplan-‐Meier failure curves
Derivation (SEER-‐Medicare) Validation (Veterans Health)Number 2397 1251Age in years 74 74Female sex 49% 2%RaceWhite 80% 76%Black 13% 22%Asian 7% 2%Year of Diagnosis<2007 0% 37%≥2007 100% 63%IMID TypeLenalidomide 78% 40%Thalidomide 22% 60%AnticoagulationNone 87% 79%Warfarin 11% 16%Low-‐molecular-‐weight heparin 1% 5%Direct oral anticoagulant 1% 1%
Patient Characteristics
SEER-‐MedicareRisk Factor Proposed HR p-‐valuePrior VTE 3.20 <0.01Obesity 0.86 0.63Central venous catheter 0.75 0.68Cardiac disease 1.00 0.98Chronic renal disease 1.14 0.38Diabetes 1.14 0.38Acute infection 1.42 0.20Immobilization 1.24 0.71General surgery or anesthesia 1.82 0.12Trauma 0.96 0.83Erythropoietin 0.89 0.61Blood clotting disorders 0.96 0.86High-‐dose dexamethasone ≥480 mg 1.09 0.55Doxorubicin no event n/aMultiagent (cytotoxic) chemotherapy 2.05 0.48Diagnosis of myeloma per se ? n/aHyperviscosity ? n/a
Univariate Analysis of Risk Factors in NCCN Guideline
Performance of Current NCCN Guideline
Low Risk High RiskSEER-‐
Medicare 57%
(n=1374)43%
(n=1023)
HR C IndexSEER-‐
Medicare 1.21
(p=0.17)0.51
(0.48-‐0.54)9%
8%*High risk defined by IMWG as 2 or more of risk factors on previous table or any therapy-‐related risks
Multivariable Analysis of New SAVED Score
Derivation Validation PointHR P-‐value HR P-‐valueSurgery (90 days) 1.74 0.15 2.30 0.05 +2Asian race 0.26 <0.01 0.43 0.40 -‐3VTE history 3.03 <0.01 4.65 <0.01 +3Eighty (≥80 years) 1.54 <0.01 0.75 0.26 +1Dexamethasone standard dose (120-‐160 mg) 1.36 0.05 1.18 0.50 +1
Dexamethasone high dose (>160 mg) 1.58 0.02 2.41 <0.01 +2
Risk stratification: High risk category if ≥2 points
Performance of New SAVED Score
5/3/19
Low Risk High Risk
Derivation 71%
(n=1711)29%
(n=686)67%
(n=837)33%
(n=414)Validation
HR C Index
Derivation 1.85(p<0.01)
0.61(0.57-‐0.64)
Validation 1.98(p<0.01)
0.60(0.54-‐0.64)
12%
7%
Strengths and Limitations
• Strengths:• Precision: large number (>3500)• External validity: observational data• Calibration: consistency in derivation and validation
• Limitations:• Lack of patient-‐level data• Derived in older patients >65 years on IMID• Possible misclassification of mild risk factors• Discrimination: no true “low” risk group
Conclusion
• Incidence of thrombosis was similarly high for lenalidomide and thalidomide in an older MM population
• Anticoagulant prophylaxis is under-‐utilized in the real-‐world• Consensus based NCCN VTE guideline does not perform well• Newly derived and validated SAVED score may identify the “highest” risk patients at the time of IMID initiation
• Most helpful as a backbone for prospective trials incorporating derived models + new biomarkers
Thank you!
• American Society of Hematology:• Clinical Research Training Institute
• Conquer Cancer Foundation:• Young Investigator Award
• Hemostasis and Thrombosis Research Society:• Mentored Research Award
• National Hemophilia Foundation:• NHF-‐Shire Clinical Fellow
• Washington State Medical Oncology Society
• University of Washington:• David Garcia• Gary Lyman• Edward Libby• Vicky Wu• Greg Warnick
• University of Vermont:• Neil Zakai
• Washington University in St. Louis:• Kristen Sanfilippo• Brian Gage• Suhong Luo
Lenalidomide N = 1863Thalidomide N = 534
Comparison with RCT Data on Overall Thrombotic Risk
ThalidomideLenalidomide
ASH Abstract. Blood 2017;130:553
• Median age 67• 12% thrombosis
55% high-‐risk patients45% low-‐risk patients
Compare and Contrast Myeloma VTE Risk ScoresSAVED score:• Patient Selection:
• Derived in older US patients• Limited to high-‐risk IMIDs
(same as NCCN guideline)• Index is initiation of IMID• Follow-‐up is censored off IMID
• Model Specification:• Cause-‐specific Cox model• Fewer risk predictors• Baseline only (prior to IMID)• Does not include anticoagulant
IMPEDE VTE score:• Patient Selection:
• Derived in US veteran males• Not limited to IMIDs (includes
lower risk patients)• Index is diagnosis of myeloma• Follow-‐up is not censored
• Model Specification:• Fine-‐Gray sub-‐hazard model• More risk predictors• Baseline + future treatment• Include anticoagulant
This is cool, but what should we do clinically?
• Caution:• There is no true low risk group in our study due to inclusion of IMIDs• The following are personal opinions only
• “Low”-‐risk (70%): VTE 7% at 6 months, 10% at 12 months• How long will patient remain on IMID therapy?• What is the patient’s risk tolerance profile?• Aspirin 81 mg at a minimum, prophylactic LMWH/DOAC if willing
• High-‐risk (30%): VTE 12% at 6 months, 20% at 12 months• Thrombotic risk unacceptably high• Prophylactic LMWH/DOAC if no severe renal failure
How is the presented model different than the abstract?
• Chemo:• Prior to external validation, we found a mistake in coding that led us identifying
more patients on concurrent chemotherapy than in reality. After fixing the error, chemotherapy was significantly reduced in number and was no longer a significant predictor. Modern therapy very rarely incorporates chemotherapy into IMID-‐containing regimen.
• Race:• Race other than Asian had a very weak contribution with borderline significance.
Since we wanted to make the final model as simple as possible we dropped that variable for simplicity.
• Validation including all original variables:• HR: 1.76 (P=0.0041)• C index: 0.574
NCCN Risk Factors HR P-‐valuePrior VTE 3.20 <0.01Obesity 0.86 0.63Central venous catheter or pacemaker 0.75 0.68
Cardiac disease (CHF, MI, arrhythmias) 1.00 0.98
Chronic renal disease 1.14 0.38Diabetes 1.14 0.38Acute infection 1.42 0.20Immobilization 1.24 0.71General surgery or anesthesia 1.82 0.12Trauma 0.96 0.83Erythropoietin 0.89 0.61Blood clotting disorders 0.96 0.86High-‐dose dexamethasone ≥480 mg 1.09 0.55
Doxorubicin* no event n/aMultiagent (cytotoxic) chemotherapy 2.05 0.48
Additional Risk Factor Considered HR P-‐valueAge (every 10-‐year increase) 1.23 0.04Sex (female versus male) 1.21 0.16Race__White 1__Black 1.33 0.13__Asian 0.25 <0.01Year of diagnosis (every 1-‐year increase) 0.98 0.66Geography (metropolitan versus rural) 0.92 0.67Liver disease 1.15 0.81Anemia 0.93 0.58Hypertension 1.28 0.10History of bleeding 1.22 0.41IMID type (lenalidomide versus thalidomide) 1.13 0.47Concurrent chemotherapy__Bortezomib 0.85 0.37__Melphalan 1.72 0.45__Cyclophosphamide 1.18 0.77Concurrent dexamethasone**__None or low dose (<120 mg / cycle) 1__Standard dose (120-‐160 mg / cycle) 1.27 0.13__High dose (>160 mg / cycle) 1.42 0.06Concurrent anticoagulation__None 1__Warfarin 1.09 0.69__Low-‐molecular-‐weight heparin 0.38 0.34__Direct oral anticoagulant* no event n/a
Univariate Analysis ofRisk Factors in Derivation
Sensitivity Analyses for SAVED Score
• Limit to patients without baseline anticoagulation in the derivation cohort (n=2097)• HR 1.68 (P<0.001), c index 0.60
• Limit to VTE diagnosis requiring hospitalization in the derivation cohort (n=2397)• HR 1.97 (P<0.001), c index 0.62
• Expand to all follow-‐up time instead of time on IMID + 90 day in the derivation cohort (n=2397)• HR 1.68 (P<0.001), c index 0.59
• Expand to patients of all ages in the validation cohort (n=2208)• HR 1.83 (P<0.001), c index 0.58