Petterson, Tanya M.

Abstract Background: Most trials of standard heparin and warfarin therapy used the 3- to 6-month cumulative incidence of venous thromboembolism (VTE) recurrence as the primary efficacy outcome. Consequently, the effect of heparin anticoagulation intensity on short-term VTE recurrence after controlling for other predictors of recurrence (e.g., active cancer, neurological disease with extremity paresis) is uncertain. Objectives: To test standard heparin anticoagulation intensity, using activated partial prothrombin time (APTT) data, as an independent predictor of 14-day VTE recurrence. Measurements: Using the resources of the Rochester Epidemiology Project, we identified all Olmsted County, MN residents with an incident deep vein thrombosis or pulmonary embolism over the 14-year period, 1984–1997 (n=1165). We followed each case (conditional on surviving one day) forward in time through their complete medical records in the community for first VTE recurrence. We collected date of incident VTE onset, start and stop dates of heparin and warfarin, and date, time and result of all APTT and platelet count values. Using these and other measures of standard heparin therapy as time-dependent variables, we tested the intensity of heparin anticoagulation as an independent predictor of 14-day VTE recurrence while controlling for other baseline characteristics using Cox proportional hazards modeling. Results: Of the 1165 patients, 1026 (88%) and 989 (85%) received heparin and warfarin, respectively, and 254 (22%) developed recurrent VTE over 5461 person-years of follow-up; 23 recurred within 14 days. In the multivariate analysis, the independent predictors of 14-day recurrence included female gender (HR=3.06, 95%CI: 1.05, 8.91; p=0.04) and neurologic disease with extremity paresis (HR=6.01, 95%CI: 1.93, 18.66; p<0.01); the hazard of recurrence was marginally increased for active cancer (HR=2.21, p=0.13). Warfarin therapy (HR=0.44, p=0.11) and a therapeutic APTT within the first 24 hours of therapy (HR=0.41, p=0.08) were marginally protective, but age, body mass index, time interval between VTE onset and start of heparin, and other measures of heparin anticoagulation intensity (e.g., proportion of time in therapeutic APTT range, etc) were not predictors of recurrence. Three patients developed new and persistent thrombocytopenia beginning 2, 4 and 5 days after starting heparin, respectively and prior to VTE recurrence. Conclusions: VTE recurrence is uncommon within 14 days after starting heparin therapy. Female gender and neurologic disease with extremity paresis are independent predictors of 14-day recurrence, and active cancer may be an additional predictor. Warfarin therapy and a therapeutic APTT within the first 24 hours may predict a reduced recurrence risk. However, other measures of heparin anticoagulation intensity were non-predictive. Heparin-induced thrombocytopenia may partially account for early VTE recurrence.

Abstract Background: Most trials of standard heparin and warfarin therapy used the 3- to 6-month cumulative incidence of venous thromboembolism (VTE) recurrence as the primary efficacy outcome. However, the independent effect of warfarin anticoagulation intensity on VTE recurrence after controlling for other predictors of recurrence (e.g., active cancer, neurological disease with extremity paresis) is uncertain. Objectives: To test warfarin anticoagulation intensity, using international normalized ratio (INR) data, as an independent predictor of 6-month VTE recurrence. Measurements: Using the resources of the Rochester Epidemiology Project, we identified all Olmsted County, MN residents with an incident deep vein thrombosis or pulmonary embolism over the 14-year period, 1984–1997 (n=1165). We followed each case (conditional on surviving 14 days without VTE recurrence) forward in time through their complete medical records in the community for first VTE recurrence. We collected date of incident VTE onset, start and stop dates of heparin and warfarin, and date, time and result of all INR values. Using these and other measures of standard warfarin therapy as time-dependent variables, we tested the intensity of warfarin anticoagulation (expressed as the cumulative percent of time with an INR at or above 2.0 [or at or above 1.5] while on warfarin) as an independent predictor of 6-month VTE recurrence while controlling for other baseline characteristics using Cox proportional hazards modeling. Results: Of the 1165 patients, 1026 (88%) and 989 (85%) received heparin and warfarin, respectively, and 254 (22%) developed recurrent VTE over 5461 person-years of follow-up; 52 recurred within 15–180 days. In the multivariate analysis, the independent predictor of 6-month recurrence was active cancer (HR=3.98, 95% CI: 2.12, 7.45; p<0.01); age, gender, body mass index and neurologic disease with extremity paresis were not predictors of VTE recurrence. In models that included presence or absence of warfarin therapy along with proportion of time with an INR ≥2, this latter intensity variable was protective (HR=0.16 for 100% versus 0% of time, p=0.02) against recurrence, while mere presence of warfarin therapy (HR =0.72, p= 0.56) was not. Similar results were obtained when intensity was represented as proportion of time with an INR ≥1.5 (HR=0.17, p=0.02). Interestingly, a therapeutic APTT within the first 24 hours of heparin therapy also was protective against 6-month recurrence (HR=0.45, p=0.02). Conclusions: Active cancer is an independent predictor of 6-month VTE recurrence. Warfarin therapy with a higher proportion of time with an INR ≥1.5 or ≥2 offers similar protection against recurrence. A rapid heparin-response (as reflected by a therapeutic APTT within 24 hours of heparin therapy) also is a predictor of reduced 6-month recurrence, possibly reflecting lower factor VIII activity at the acute incident VTE event. Gender is not a predictor of long-term recurrence.

Abstract Background: While several uncommon disorders (e.g., myeloproliferative diseases, connective tissue disease, etc.) have been suggested as potential risk factors for VTE, the independence and magnitude of any risk is uncertain. Objective: To test uncommon characteristics as potential independent risk factors for VTE, and estimate the magnitude of risk for each. Methods: Using the resources of the Rochester Epidemiology Project, we identified all incident cases of deep vein thrombosis and pulmonary embolism over the 6-year period, 1992–1997. For each objectively confirmed case (n=562), we identified up to two Olmsted County residents without VTE who most closely matched a case on age, gender, and medical record number (n=755). Because medical record numbers are assigned sequentially, such matching assures a similar duration of medical follow-up. For all cases and controls, we reviewed the complete medical records in the community for over 125 baseline characteristics which were tested for an association with VTE using conditional logistic regression. Results: The mean ± SD (range) ages for cases and controls were 65.71 ± 18.85 (0–102) and 64.90 ± 18.66 (0–102) years, respectively, and 55.5% were female. Novel univariate risk factors for incident VTE included former tobacco smoker, ischemic heart disease, valvular heart disease, chronic lung disease, all-cause pulmonary hypertension, chronic renal disease, myeloproliferative disorders, any infection, femoral artery catheterization (angiography or percutaneous intervention), ICU admission and angiotensin converting enzyme (ACE) inhibitor therapy. Among infections, septicemia, pneumonia, and ENT, cardiovascular, GI, urinary tract and skin/soft tissue infections were univariately associated with VTE; reproductive tract infection was not. Diabetes mellitus, hyperlipidemia, lipid-lowering drugs (including statins), inflammatory bowel disease, asthma, connective tissue disease, and influenza were not univariately associated with VTE, and beta-blocker therapy was not protective. A multivariate analysis will test these characteristics as potential independent VTE risk factors. Conclusions: Several novel characteristics are univariately associated with VTE. Such novel characteristics may provide new insights into VTE mechanisms if, upon further analysis, these characteristics are independent VTE risk factors.

Abstract Background: While currently-diagnosable hematological cancer is a potent VTE risk factor (RR range =12–32), whether undiagnosable hematological cancer is a risk factor is uncertain. Objective: To estimate the hazard of subsequent hematological cancer among incident VTE patients with no apparent active hematological cancer within ± 3 months of the VTE event. Methods: In a population-based cohort study, we followed Olmsted County, MN residents with incident VTE over the 27-year period, 1973–1999, and previously matched Olmsted County residents without VTE, who survived with no apparent active hematological cancer within ± 3 months of the VTE event or a routine medical visit date, respectively, forward in time through their complete medical records in the community for hematological cancer. Data were abstracted on baseline demographic and clinical characteristics, type and stage of hematological cancer, and any prior solid tumor treatments. For cases and controls, hematological cancer-free survival from the event date was estimated by the Kaplan Meier method, and the homogeneity of survival curves was examined by the log-rank test. Cox proportional hazards regression was used to adjust for covariates. Results: 1860 VTE cases (56% females) and 1890 controls (53% females) were followed for 16,497 and 15,320 person-years, respectively, conditional on surviving 3-months free of hematological cancer. 43 VTE cases and 24 controls developed hematological cancer at a median of 63.5 months (range 0.7 – 334 months). The median age for cases was 63.2 years compared to 67.8 years for the controls (p<0.001). Hematological cancer types included non-Hodgkin’s lymphoma (n=19), chronic lymphocytic leukemia (n=11), acute myeloid leukemia (n=1), multiple myeloma (n=9), myeloproliferative disorders (n=10), and myelodysplastic syndromes (n=14). Case status was not univariately associated with subsequent hematologic cancer (HR=1.42; 95% CI: 0.85, 2.39; p=0.18). After adjusting for age, also associated with cancer, case status was significantly associated with hematological cancer (HR=1.90; 95% CI: 1.13, 3.20; p=0.015). Conclusions: Incident VTE patients have nearly a two-fold increased risk of subsequent hematological cancer. We hypothesize that VTE can be the initial manifestation of an occult hematological cancer clone.

Abstract Background: Hematological cancer patients are at an increased risk for VTE (RR range = 12–32). However, whether VTE risk among such patients can be further stratified is uncertain. Objective: To test hematological cancer type, stage, stage progression and chemotherapy as potential risk factors for VTE among active hematological cancer patients after controlling for other previously-identified VTE risk factors. Methods: Using the resources of the Rochester Epidemiology Project, Mayo Clinic Master Diagnostic Index and Mayo Clinic Tumor Registry, we identified all Olmsted County, MN residents with active hematological cancer over the 28-year period, 1973–2000. From this prevalence cohort, we identified 86 patients with no prior VTE (controls) who were matched on age and date of hematological cancer diagnosis to 86 hematological cancer patients with incident VTE over the same time frame (cases). For all cases and controls, we reviewed the complete medical records in the community for baseline and hematological cancer-related characteristics. Hematological cancers were re-staged at the dates of the cancer and VTE diagnosis. We tested these characteristics as potential risk factors for VTE in active hematological cancer using conditional logistic regression. Results: In an initial multivariate analysis that included body mass index (BMI), hospitalization, and any infection or central venous catheter placement within 90 days prior to the VTE event, VTE was significantly associated with hospitalization (OR=6.70; p<0.001), and marginally associated with any infection (OR=2.18; p=0.09). After adjusting for the above variables, chemotherapy administered within the preceding 90 days was significantly associated with VTE (OR=4.25; p=0.02), while stage progression was marginally associated (OR=4.79; p=0.10). Compared to all other hematological cancer types, acute leukemia (OR=5.95; p=0.01) and non-Hodgkin’s lymphoma (OR=2.61; p=0.01) were associated with VTE. However, after adjusting for BMI, hospitalization, any infection and central venous catheter, only non-Hodgkin’s lymphoma was independently associated with VTE (OR=3.79, p=0.009), while acute leukemia was not (OR=2.66, p=0.35). Conclusions: Hematological cancer type (in particular, non-Hodgkin’s lymphoma and possibly acute leukemia), recent hospitalization, recent chemotherapy, and possibly stage progression and recent infection, are risk factors for VTE among patients with active hematological cancer.

Abstract Confinement to a nursing home (NH) is an independent risk factor for VTE and accounts for 13% of all VTE occurring in the community, but the question of which NH residents are at risk is largely unanswered. Objective: To determine VTE risk factors among NH residents. Methods: Using the Minnesota State Case Mix Review Program Public Research Files, we performed a case-control study of all Olmsted County, MN NH patients with an ICD-9 code-identified diagnosis of VTE from 1988–1994. From the same source, we identified one age-, sex-matched control and used conditional logistic regression to estimate odds ratios for an association of VTE with 46 characteristics assessed in the Minnesota Case Mix data. Results: Univariate associations between characteristics and odds of VTE are described below. Significantly increased odds were observed for residents assessed following hospitalization (a nearly 4-fold increase, p=0.002, Table) and residents considered incapable of self-preservation (i.e., unable to cope with potentially harmful situations) (> 2.0 fold increase, p=0.03). Non-significant 2–3 fold increased odds were observed for limitations in certain specific Activities of Daily Living (ADL) (including walking and dressing) and for limitations in more than 3 ADLs. Similar increased odds were observed for residents using a wheelchair, taking more than 3 oral medications, enrolled in professional activity programs, and needing behavior management. Conclusions: Need for assistance (especially problems with immobility), lack of self preservation, and recent hospitalization may identify a subset of NH residents that are at especially high VTE risk and who may warrant consideration of prophylaxis. Univariate VTE Risk Factors for Nursing Home Residents Characteristic Odds Ratio 95% CI P-value Oral medications > 3 3.00 0.97-9.30 0.06 Activities of Daily Living High/Medium (4-8) vs. low (0-3) 2.20 0.76-6.33 0.14 Help dressing 2.75 0.88-8.64 0.08 Help walking 3.00 0.61-14.86 0.18 Wheelchair use 2.50 0.97-6.44 0.58 Needs behavior management 3.00 0.62-14.86 0.18 Help toileting 2.00 0.68-5.85 0.20 Lack of self-preservation 2.33 1.07-5.10 0.03 Transfer from hospital 3.86 1.68-8.86 0.0015 Anti-depressant drugs 0.50 0.19-1.33 0.16 Professional activity program 2.25 0.69-7.3 0.18

Abstract Introduction: Hospitalization-associated VTE attack (incident plus recurrent VTE) rates over the 6-year period, 2005-2010, did not change significantly despite achieving near-universal in-hospital VTE prophylaxis; 75% of these VTE events occurred after hospital discharge with a median of 19.5 days from hospital discharge to VTE. The median durations of hospitalization and in-hospital prophylaxis were 3 days and 70 hours, respectively, suggesting that VTE prophylaxis was inadequate. However, the hospital population that would benefit most from post-hospitalization VTE prophylaxis is uncertain. Our study aim was to test hospital length of stay (LOS) as a potential predictor of post-hospital-discharge incident or recurrent VTE. Methods: Using the resources of the Rochester Epidemiology Project, we identified all unique Olmsted County, MN residents hospitalized at least once at a Rochester, MN Mayo Clinic hospital and all such residents with objectively-diagnosed occurrence of first or recurrent VTE during hospitalization or within 92 days after hospital discharge, 2005-2010. We tested LOS as a potential predictor of post-hospital-discharge VTE (conditional on surviving to hospital discharge free of VTE) using the Anderson-Gill version of Cox regression modeling and adjusting for age, sex, any surgery performed during that hospitalization, and Charlson Comorbidity Index calculated from conditions recorded from 1/1/2005 until the day of hospital discharge. Results: Over the period, 2005-2010, 31,588 unique Olmsted County residents were hospitalized at least once at a Rochester Mayo Clinic hospital. The distribution of LOS was: 1-2 (3.8%), 3 (24.2%), 4 (23.3%), 5 (16.4%), 6 (8.8%), 7-9 (12.4%), 10-19 (8.4%) and 20+ (2.3%). Of the 253 VTE events identified within 92 days post-hospital discharge, 6, 29, 31, 38, 38, 53, 42 and 16 had been hospitalized for 1-2, 3, 4, 5, 6, 7-9, 10-19 and 20+ days, respectively. The hazard ratios (HR) for post-hospital-discharge VTE were similar for hospital LOS ≤4 days; compared to LOS ≤3 days, the HRs were significantly increased for hospital LOS of 5 (HR=1.78), 6 (HR=2.86), 7-9 (HR=2.71), 10-19 (HR=3.16) and 20+ (HR=4.20) days (Table 1). Conclusions: The adjusted hazards of post-hospital discharge VTE for hospital LOS > 4 days were significantly increased 1.8 to 4.2-fold compared to LOS ≤3 days, suggesting that hospital LOS may be useful for identifying hospitalized patients suitable for longer duration VTE (including post-discharge) prophylaxis. Future studies should be aimed at identifying additional predictors of post-hospital-discharge VTE such that extended prophylaxis can be targeted to those patients who would benefit most. Disclosures No relevant conflicts of interest to declare.

Abstract Background: Active cancer without or with chemotherapy is associated with a 4- to 6.5-fold increased risk for VTE 1. However, cancer-associated VTE incidence may be under- or overestimated by reports that linked cancer registry data to an anticoagulation clinic database or to hospital discharge diagnosis codes 2–5. Objective: To estimate the incidence of VTE in active cancer in a well-defined geographic population that includes the full spectrum of VTE events (including rapidly fatal events) occurring in the community. Methods: Using the longitudinal resources of the Rochester Epidemiology Project and Mayo Clinic Tumor Registry, we identified all Olmsted County, MN residents with incident VTE by review of their complete medical records in the community 6, and all residents with active cancer (by cancer site), over the 28-year period, 1973–2000. Using residents with active cancer within the three months prior or subsequent to incident VTE as the numerator, and all residents with active cancer as the denominator, we estimated the incidence of VTE in active cancer, both overall and for the subsets of pancreatic cancer, brain cancer and lymphoma. Results: Over 78,236 active-cancer person-years, 595 Olmsted County residents developed incident VTE. The overall active cancer-associated VTE incidence was 760 per 100,000 person-years (95% CI: 701, 824), about 6.5-fold higher than the general population 7. Over 473, 983 and 4,258 pancreatic cancer, brain cancer and lymphoma person-years, respectively, 51, 18 and 46 residents developed incident VTE. The pancreatic cancer-, brain cancer- and lymphoma-associated VTE incidence rates were 10,784 (95% CI: 8029, 14,179), 1,831 (95% CI: 1085, 2,893) and 1,080 (95% CI: 791, 1,441) per 100,000, respectively. Overall, cancer-associated VTE incidence increased with age, and for all ages, men had a higher VTE rate than women. We observed a much higher pancreatic cancer-associated VTE incidence than previously reported. Conclusions: The incidence of VTE in individuals with active cancer (especially pancreatic cancer) is higher than the general population. The incidence of VTE in patients with pancreatic cancer is higher than previously estimated.

Abstract Abstract 709 Background: A recent analysis of merged genome-wide and candidate gene genotypes in VTE cases and controls identified multiple tag SNPs that were strongly associated with VTE. Objective: To identify rare and/or novel functional variants by sequencing the implicated genes. Methods: Cases (n=1488) were Mayo Clinic European-American patients of non-Hispanic ancestry with objectively-diagnosed VTE in the absence of active cancer, venous catheter or antiphospholipid antibodies. Controls (n=1439) were Mayo Clinic outpatients without VTE who were frequency-matched on case age, gender, race, MI/stroke status and state of residence. For this analysis, we selected a subset of these cases and controls for sequencing to take advantage of the joint configuration of two ABO SNPs of primary interest, rs8176719 (ABO exon 6 deletion determining type O blood group) and rs2519093 (ABO intron 1 tag SNP), which were previously shown to be strongly associated with VTE (p=5.7E-12 and p=3.0E-16, respectively). We randomly sampled 82 cases and 14 controls within 3 of the 9 potential allele frequency cells (Figure). The rs8176719 alleles are -−/−- (double deletion is the common allele), –/G, and G/G (the rare allele). The rs2519093 alleles are GG (G is the common allele), AG, and AA (A is the rare allele). For each SNP, the genotypes are represented as 0, 1, or 2 copies of the minor allele. We represented the joint allelic configuration of the two SNPs with the number of copies of the rs8176719 given first as 0/0 (both with 0 copies of the minor allele), 0/1, 0/2 (0 copies of the rs8176719 SNP), 1/0 (1 copy of the rare rs8176719 SNP), 1/1/, and 1/2, and 2/0 (2 copies of the rare allele for the rs8176719 SNP and 0 copies of the rare allele of the rs2519093 SNP), 2/1, and 2/2. From the Figure one observes discrepancies between cases and controls at the 0/0, 1/1 and 2/2 combinations. We randomly sampled from these three combinations, taking one third of the case series. For each SNP, we had 28 cases with 0/0 copies of the rare allele, 27 cases with 1/1 copies of the rare allele; and 27 cases with the combination of 2/2 copies of the rare allele. We compared these 82 cases with 14 controls that do not have any of these combinations. Sixteen genes were selected for deep sequencing, including 5 genes harboring SNPs significantly associated with VTE (F5, SLC19A2, ABO, NME7, ATP1B1), 10 genes with SNPs marginally associated with VTE (C1orf114, KLKB1, SELP, F11, SCUBE1, PRKCB1, CD44, ITPR1, GFRA1, BLZF1), and CYP4V2 which reportedly confounds F11 and KLKB1. Agilent SureSelect probes were designed to capture and enrich the ∼2 Mb genomic regions of these 16 genes. Samples were multiplexed (12-plex) and sequenced using Illumina HiSeq 2000. The sequence reads were aligned to the human genome build 36 using Burrows-Wheeler Aligner, and the single nucleotide variants (SNVs) and small INDELs were called using SNVMix and GATK, respectively. For this analysis, novel ABO SNVs were tested for an association with VTE using age-, sex-adjusted logistic regression and Fisher's Exact Test. Results: 98% of the targeted regions were sequenced with > 20X coverage. On average, ∼2500 SNVs and ∼200 INDELs were detected in each sample. Fifteen novel SNVs in intron 6 and 3' of the ABO gene were associated with VTE (p<E-06) and belonged to 3 distinctive LD blocks; none were in LD with the coding or tag ABO SNPs (rs8176719; rs2519093). SNVs inside the middle LD block at the 3' of ABO are located within an enhancer and promoter histone marked with putative transcription factor binding sites. In addition, strong evidence from both ENCODE and dbEST support the middle LD block as lying within a novel transcript, probably an extension of the 3' of ABO. In addition, we discovered a novel, significant, protective, frame-shifting single base (G) deletion at ABO chr9:135120877. Conclusion: Novel ABO functional variants that are associated with VTE were identified by deep sequencing. Disclosures: Heit: Daiichi Sankyo: Consultancy, Honoraria.

Abstract Introduction: Diabetes mellitus is often cited as a VTE risk factor. However, persons with diabetes are frequently hospitalized for medical illness or undergo surgery, both major VTE risk factors. Thus, the association of VTE with diabetes independent of surgery or hospitalization is uncertain. Methods: Using longitudinal, population-based Rochester Epidemiology Project resources, we identified all Olmsted County, MN residents who met objective criteria for incident VTE over the 25-year period, 1976–2000 (n=1922), and one to two controls per case, matched on age, gender, Olmsted County residency, and length of medical history (n=2115). For cases and controls, we reviewed their complete medical history in the community for baseline characteristics previously identified as independent VTE risk factors, including clinically-diagnosed diabetes mellitus. We tested diabetes as a potential VTE risk factor both alone and after adjusting for other baseline characteristics, and in the subset of cases with idiopathic VTE, using conditional logistic regression. Results: Among all cases and controls, 231 (12%) and 199 (9.4%) had diabetes, respectively. Univariately, diabetes was associated with overall VTE (odds ratio [OR]=1.32; 95% CI: 1.07, 1.62; p=0.009). However, after controlling for body mass index (OR=1.04, p<0.001), hospitalization with major surgery (OR=25.2, p<0.001) or hospitalization for medical illness (OR=7.3, p<0.001), active cancer (OR=8.3, p<0.001) and varicose veins (OR=1.3, p=0.003), diabetes was no longer associated with VTE (OR=1.09; 95% CI: 0.85, 1.41; p=0.49). Among 458 idiopathic VTE cases and 518 matched controls, 43 (9.4%) and 46 (8.9%) had diabetes, respectively. Diabetes was not associated with idiopathic VTE (OR=1.07, 95% CI: 0.69, 1.67; p=0.76). Conclusions: Diabetes mellitus is not an independent risk factor for overall or idiopathic incident VTE.