HHS Public Access Author manuscript Author Manuscript
J Thromb Haemost. Author manuscript; available in PMC 2017 September 09. Published in final edited form as: J Thromb Haemost. 2016 September ; 14(9): 1773–1778. doi:10.1111/jth.13378.
Predictors of Active Cancer thromboembolic Outcomes: Validation of the Khorana Score among Patients with Lung Cancer Aaron Mansfield1, Alfonso J. Tafur2, Chihsiung E. Wang3, Taxiarchis V. Kourelis4, Ewa M. Wysokinska5, and Ping Yang6
Author Manuscript
1Department
of Oncology, Division of Medical Oncology, Mayo Clinic, 200 First Street SW, Rochester, MN 55905, U.S.
2Department
of Medicine, Division of Cardiology - Vascular Medicine Program, NorthShore University Health System, 2650 Ridge Avenue, Evanston IL, 60201; University of Chicago School of Medicine, Chicago, IL, U.S., [email protected] 3Department
of Surgery, NorthShore University Health System, 2650 Ridge Avenue, Evanston IL, 60201, [email protected]
4Department
of Medicine, Division of Hematology; Department of Oncology, Division of Medical Oncology, Mayo Clinic, 200 First Street SW, Rochester, MN 55905, U.S., [email protected]
5Willmar
Author Manuscript
Regional Cancer Center, 301 Becky Avenue SW, Willmar, MN 56201, U.S., [email protected] 6Department
of Health Sciences Research, Division of Epidemiology, Mayo Clinic, 200 First Street SW, Rochester, MN 55905, U.S., [email protected]
Abstract Objectives—Lung cancer is strongly associated with venous thromboembolism (VTE), but primary prevention against VTE is not a validated management strategy. Risk assessment models will be necessary for efficient implementation of preventative strategies.
Author Manuscript
Materials and methods—Utilizing a prospectively collected lung cancer database, we aimed to validate the Khorana Risk Score (KRS) in the prediction of VTE among patients with lung cancer. VTE events were retrospectively identified by reviewers unaware of the clinical prediction score calculation. The association between KRS and the risk of VTE was examined using cumulative
CORRESPONDING AUTHOR: Dr. Aaron S. Mansfield, Department of Oncology, Division of Medical Oncology, Mayo Clinic, 200 First Street SW, Rochester, MN 55905, United States, Phone: 507-266-9160, [email protected]. Addendum A. Tafur and A. Mansfield were responsible for the study conception and design; all authors were responsible for the acquisition of data, and analysis and interpretation of data. A. Mansfield and A. Tafur drafted the manuscript and A. Mansfield, A. Tafus and C. Wang critically revised the manuscript. Disclosure: A. Mansfield received honoraria from Celgene and Genetech for an educational activity and participation in an advisory board in 2014 and 2015 respectively, outside the submitted work. The other authors state that they have no conflict of interest.
Mansfield et al.
Page 2
Author Manuscript
incidence function with competing risks models. Mortality prediction was evaluated as secondary outcome. Results—We included 719 patients in our review. The patients were predominantly older males with NSCLC and 40% had metastatic disease at inception. The median follow up was 15.2 months. There were 83 VTEs (11.5%) and 568 (78.8%) patients died. A high KRS (Cumulative Incidence 12.4%, 95% Confidence Interval 6.4-20.5%) was not associated with VTE compared to an intermediate score (Cumulative Incidence 12.1%, 95% Confidence Interval 9.5-15.0%)) in both univariate and multivariable analyses. However, a high KRS was a predictor of mortality (HR 1.7 95% CI 1.4 - 2.2). Conclusions—Among patients with lung cancer the KRS did not stratify the patients at the highest risk of VTE. Improved risk stratification methods are needed for this group of patients prior to implementing a primary prevention strategy.
Author Manuscript
Keywords Lung Cancer; Venous Thromboembolism; Prediction Score; Mortality; Prevention
INTRODUCTION
Author Manuscript
Lung cancer is the second most common cancer in United States. Lung Cancer is also strongly associated with venous thromboembolism (VTE). As noted in a 13-year period study in the Olmsted County, Minnesota population, the standardized risk of VTE among patients with lung cancer is 13 times higher than expected [1]. VTE is a morbid, costly and potentially lethal complication among patients with cancer [2]. Indeed, cancer-associated thrombosis is a leading cause of death among patients with cancer [3]. Patients with lung cancer and early occurrence of VTE, have a worse survival than patients who do not develop VTE even after adjusting for stage, comorbidities, performance status [4]. Cancer-associated thrombosis is a potentially preventable disease, however primary prevention is not currently a validated strategy [5]. The implementation of risk assessment models for VTE risk stratification may improve the applicability of primary prevention strategies in this population.
Author Manuscript
Assessment of VTE risk is currently recommended before initiation of chemotherapy [6]. To date, the best validated cancer-associated thrombosis prediction score is the Khorana risk score (KRS) [7]. The score is heavily weighted on type of cancer, thus all patients with lung cancer have at least intermediate risk. Save-ONCO studied the use of semuloparin for primary VTE prevention in 3212 patients with cancer, and lung cancer was the most common type of malignancy that was included (36%). In this study the KRS failed to predict most of the thromboembolic events, and most of the cancer-associated thromboses occurred among patients with an intermediate KRS (64%) [8, 9]. If the KRS does not perform well among patients with one of the most common types of cancer, this group of patients will need a cancer-specific evaluation of VTE predictors. We aimed to evaluate the performance of the KRS for cancer-associated VTE prediction in a cohort of patients with lung cancer.
J Thromb Haemost. Author manuscript; available in PMC 2017 September 09.
Mansfield et al.
Page 3
Author Manuscript
POPULATION AND METHODS
Author Manuscript
We used a prospectively maintained database of consecutive patients with lung cancer treated at Mayo Clinic Rochester between January 1998 to December 2011. The primary outcome was VTE prediction and the secondary outcome was overall survival prediction. All cases were histologically examined by a Mayo Clinic pathologist. Demographic variables and cancer specific variables including performance status, stage, grade, cell type were prospectively collected. Patients on chronic anticoagulants were excluded from the analysis. For patients diagnosed after January 2010, the 7th edition of lung cancer staging system was used [10]. The data for the KRS correspond to clinical data obtained within 1 month of recruitment, including blood counts. The KRS was retrospectively quantified based on corresponding clinical-pathological variables (1 point each for lung cancer; BMI ≥35; Hemoglobin < 10g/dL or erythropoietin stimulating agent; white blood cells ≥ 11, 000; platelets ≥ 350,000). The primary outcome measure was VTE, including symptomatic or incidentally found deep vein thrombosis (DVT) of lower or upper limbs, pulmonary embolism (PE) and visceral venous thrombosis. Incidental VTE was defined as those were detected in the absence of a documented clinical symptom. VTE was retrospectively extracted from the electronic medical records by two of the coauthors (EW, TK) who were unaware of the clinical prediction score calculation. The VTE diagnosis was assigned based on pre-defined criteria: limb or visceral VTE confirmed by venogram, angiography, computed tomography, magnetic resonance, or compression ultrasound. Overall survival follow up was available for all patients. Time to death and time to VTE from recruitment date were determined for all cases. The Mayo Foundation Institutional Review Board approved this study.
Author Manuscript
Statistical analysis
Author Manuscript
We calculated descriptive statistics to summarize the distribution of baseline parameters. Continuous measures were categorized into quartiles to explore their association with the outcomes of interest [11, 12]. We included height among the VTE predictors due to recent data that underscored an independent association with thrombotic events[13]. The association between KRS and the risk of VTE was analyzed using cumulative incidence function with competing risk models as proposed by Fine and Gray[14]; while the association between KRS and the hazard of death was examined using Kaplan-Meier survival curve and Cox proportional hazards regression modelling. A high score was defined as ≥ 3 [7]. We did not model for missing information, only available data were analyzed. All-cause mortality was studied as a secondary outcome. Multivariable analyses were done using forward modeling and limited given number of events. For all statistical analyses we used SAS 9.4 (SAS Institute Inc., Cary, NC, USA). Non-parametric continuous variables are expressed as median and inter quartile rage (IQR). A p value of 76yo
Demographics
Risk Factor
3.5
8.6
0.0
7.3
8.6
7.2
2.6
4.8
2.3
5.1
0.0
5.1
7.1
8.4
6.3
3.0
6 mo
5.2
10.6
0.0
7.3
11.5
8.9
3.0
5.8
2.3
7.1
0.0
5.1
8.6
10.1
8.7
3.6
12 mo
6.4
13.9
0.0
9.8
13.8
10.6
3.3
6.4
2.3
7.6
0.0
5.1
8.6
10.6
10.1
4.2
24 mo
9.3
14.6
2.3
14.6
17.1
13.3
8.6
8.0
2.3
12.7
0.0
10.3
11.4
11.7
12.8
6.6
Overall
5.7
4.4
5.5
4.8
2.6
3.5
5.9
5.9
5.0
5.1
4.9
4.9
4.7
4.2
4.7
5.3
3 mo
7.2
5.6
6.6
5.8
2.6
4.9
7.6
6.9
6.0
6.1
5.9
5.8
5.7
5.1
5.2
6.6
6 mo
8.4
6.9
8.1
7.0
2.9
5.8
9.7
8.7
7.5
7.4
7.2
7.3
7.1
6.1
5.8
8.4
12 mo
10.3
8.1
9.9
8.8
3.5
7.0
11.7
11.1
9.2
9.3
8.9
8.9
8.8
8.1
7.5
10.3
24 mo
No risk factor
Cumulative Incidence of VTE (%)* With risk factor
Accounting for competing risk of mortality
*
Author Manuscript Table 2
Author Manuscript
13.1
11.5
12.9
11.4
6.5
10.2
13.2
14.9
12.4
11.2
11.9
11.8
11.8
11.6
10.5
13.2
Overall
0.226
0.027
0.035
0.744
J Thromb Haemost. Author manuscript; available in PMC 2017 September 09. Published in final edited form as: J Thromb Haemost. 2016 September ; 14(9): 1773–1778. doi:10.1111/jth.13378.
Predictors of Active Cancer thromboembolic Outcomes: Validation of the Khorana Score among Patients with Lung Cancer Aaron Mansfield1, Alfonso J. Tafur2, Chihsiung E. Wang3, Taxiarchis V. Kourelis4, Ewa M. Wysokinska5, and Ping Yang6
Author Manuscript
1Department
of Oncology, Division of Medical Oncology, Mayo Clinic, 200 First Street SW, Rochester, MN 55905, U.S.
2Department
of Medicine, Division of Cardiology - Vascular Medicine Program, NorthShore University Health System, 2650 Ridge Avenue, Evanston IL, 60201; University of Chicago School of Medicine, Chicago, IL, U.S., [email protected] 3Department
of Surgery, NorthShore University Health System, 2650 Ridge Avenue, Evanston IL, 60201, [email protected]
4Department
of Medicine, Division of Hematology; Department of Oncology, Division of Medical Oncology, Mayo Clinic, 200 First Street SW, Rochester, MN 55905, U.S., [email protected]
5Willmar
Author Manuscript
Regional Cancer Center, 301 Becky Avenue SW, Willmar, MN 56201, U.S., [email protected] 6Department
of Health Sciences Research, Division of Epidemiology, Mayo Clinic, 200 First Street SW, Rochester, MN 55905, U.S., [email protected]
Abstract Objectives—Lung cancer is strongly associated with venous thromboembolism (VTE), but primary prevention against VTE is not a validated management strategy. Risk assessment models will be necessary for efficient implementation of preventative strategies.
Author Manuscript
Materials and methods—Utilizing a prospectively collected lung cancer database, we aimed to validate the Khorana Risk Score (KRS) in the prediction of VTE among patients with lung cancer. VTE events were retrospectively identified by reviewers unaware of the clinical prediction score calculation. The association between KRS and the risk of VTE was examined using cumulative
CORRESPONDING AUTHOR: Dr. Aaron S. Mansfield, Department of Oncology, Division of Medical Oncology, Mayo Clinic, 200 First Street SW, Rochester, MN 55905, United States, Phone: 507-266-9160, [email protected]. Addendum A. Tafur and A. Mansfield were responsible for the study conception and design; all authors were responsible for the acquisition of data, and analysis and interpretation of data. A. Mansfield and A. Tafur drafted the manuscript and A. Mansfield, A. Tafus and C. Wang critically revised the manuscript. Disclosure: A. Mansfield received honoraria from Celgene and Genetech for an educational activity and participation in an advisory board in 2014 and 2015 respectively, outside the submitted work. The other authors state that they have no conflict of interest.
Mansfield et al.
Page 2
Author Manuscript
incidence function with competing risks models. Mortality prediction was evaluated as secondary outcome. Results—We included 719 patients in our review. The patients were predominantly older males with NSCLC and 40% had metastatic disease at inception. The median follow up was 15.2 months. There were 83 VTEs (11.5%) and 568 (78.8%) patients died. A high KRS (Cumulative Incidence 12.4%, 95% Confidence Interval 6.4-20.5%) was not associated with VTE compared to an intermediate score (Cumulative Incidence 12.1%, 95% Confidence Interval 9.5-15.0%)) in both univariate and multivariable analyses. However, a high KRS was a predictor of mortality (HR 1.7 95% CI 1.4 - 2.2). Conclusions—Among patients with lung cancer the KRS did not stratify the patients at the highest risk of VTE. Improved risk stratification methods are needed for this group of patients prior to implementing a primary prevention strategy.
Author Manuscript
Keywords Lung Cancer; Venous Thromboembolism; Prediction Score; Mortality; Prevention
INTRODUCTION
Author Manuscript
Lung cancer is the second most common cancer in United States. Lung Cancer is also strongly associated with venous thromboembolism (VTE). As noted in a 13-year period study in the Olmsted County, Minnesota population, the standardized risk of VTE among patients with lung cancer is 13 times higher than expected [1]. VTE is a morbid, costly and potentially lethal complication among patients with cancer [2]. Indeed, cancer-associated thrombosis is a leading cause of death among patients with cancer [3]. Patients with lung cancer and early occurrence of VTE, have a worse survival than patients who do not develop VTE even after adjusting for stage, comorbidities, performance status [4]. Cancer-associated thrombosis is a potentially preventable disease, however primary prevention is not currently a validated strategy [5]. The implementation of risk assessment models for VTE risk stratification may improve the applicability of primary prevention strategies in this population.
Author Manuscript
Assessment of VTE risk is currently recommended before initiation of chemotherapy [6]. To date, the best validated cancer-associated thrombosis prediction score is the Khorana risk score (KRS) [7]. The score is heavily weighted on type of cancer, thus all patients with lung cancer have at least intermediate risk. Save-ONCO studied the use of semuloparin for primary VTE prevention in 3212 patients with cancer, and lung cancer was the most common type of malignancy that was included (36%). In this study the KRS failed to predict most of the thromboembolic events, and most of the cancer-associated thromboses occurred among patients with an intermediate KRS (64%) [8, 9]. If the KRS does not perform well among patients with one of the most common types of cancer, this group of patients will need a cancer-specific evaluation of VTE predictors. We aimed to evaluate the performance of the KRS for cancer-associated VTE prediction in a cohort of patients with lung cancer.
J Thromb Haemost. Author manuscript; available in PMC 2017 September 09.
Mansfield et al.
Page 3
Author Manuscript
POPULATION AND METHODS
Author Manuscript
We used a prospectively maintained database of consecutive patients with lung cancer treated at Mayo Clinic Rochester between January 1998 to December 2011. The primary outcome was VTE prediction and the secondary outcome was overall survival prediction. All cases were histologically examined by a Mayo Clinic pathologist. Demographic variables and cancer specific variables including performance status, stage, grade, cell type were prospectively collected. Patients on chronic anticoagulants were excluded from the analysis. For patients diagnosed after January 2010, the 7th edition of lung cancer staging system was used [10]. The data for the KRS correspond to clinical data obtained within 1 month of recruitment, including blood counts. The KRS was retrospectively quantified based on corresponding clinical-pathological variables (1 point each for lung cancer; BMI ≥35; Hemoglobin < 10g/dL or erythropoietin stimulating agent; white blood cells ≥ 11, 000; platelets ≥ 350,000). The primary outcome measure was VTE, including symptomatic or incidentally found deep vein thrombosis (DVT) of lower or upper limbs, pulmonary embolism (PE) and visceral venous thrombosis. Incidental VTE was defined as those were detected in the absence of a documented clinical symptom. VTE was retrospectively extracted from the electronic medical records by two of the coauthors (EW, TK) who were unaware of the clinical prediction score calculation. The VTE diagnosis was assigned based on pre-defined criteria: limb or visceral VTE confirmed by venogram, angiography, computed tomography, magnetic resonance, or compression ultrasound. Overall survival follow up was available for all patients. Time to death and time to VTE from recruitment date were determined for all cases. The Mayo Foundation Institutional Review Board approved this study.
Author Manuscript
Statistical analysis
Author Manuscript
We calculated descriptive statistics to summarize the distribution of baseline parameters. Continuous measures were categorized into quartiles to explore their association with the outcomes of interest [11, 12]. We included height among the VTE predictors due to recent data that underscored an independent association with thrombotic events[13]. The association between KRS and the risk of VTE was analyzed using cumulative incidence function with competing risk models as proposed by Fine and Gray[14]; while the association between KRS and the hazard of death was examined using Kaplan-Meier survival curve and Cox proportional hazards regression modelling. A high score was defined as ≥ 3 [7]. We did not model for missing information, only available data were analyzed. All-cause mortality was studied as a secondary outcome. Multivariable analyses were done using forward modeling and limited given number of events. For all statistical analyses we used SAS 9.4 (SAS Institute Inc., Cary, NC, USA). Non-parametric continuous variables are expressed as median and inter quartile rage (IQR). A p value of 76yo
Demographics
Risk Factor
3.5
8.6
0.0
7.3
8.6
7.2
2.6
4.8
2.3
5.1
0.0
5.1
7.1
8.4
6.3
3.0
6 mo
5.2
10.6
0.0
7.3
11.5
8.9
3.0
5.8
2.3
7.1
0.0
5.1
8.6
10.1
8.7
3.6
12 mo
6.4
13.9
0.0
9.8
13.8
10.6
3.3
6.4
2.3
7.6
0.0
5.1
8.6
10.6
10.1
4.2
24 mo
9.3
14.6
2.3
14.6
17.1
13.3
8.6
8.0
2.3
12.7
0.0
10.3
11.4
11.7
12.8
6.6
Overall
5.7
4.4
5.5
4.8
2.6
3.5
5.9
5.9
5.0
5.1
4.9
4.9
4.7
4.2
4.7
5.3
3 mo
7.2
5.6
6.6
5.8
2.6
4.9
7.6
6.9
6.0
6.1
5.9
5.8
5.7
5.1
5.2
6.6
6 mo
8.4
6.9
8.1
7.0
2.9
5.8
9.7
8.7
7.5
7.4
7.2
7.3
7.1
6.1
5.8
8.4
12 mo
10.3
8.1
9.9
8.8
3.5
7.0
11.7
11.1
9.2
9.3
8.9
8.9
8.8
8.1
7.5
10.3
24 mo
No risk factor
Cumulative Incidence of VTE (%)* With risk factor
Accounting for competing risk of mortality
*
Author Manuscript Table 2
Author Manuscript
13.1
11.5
12.9
11.4
6.5
10.2
13.2
14.9
12.4
11.2
11.9
11.8
11.8
11.6
10.5
13.2
Overall
0.226
0.027
0.035
0.744
