DOI: 10.1161/CIRCULATIONAHA.115.020338 Outcomes After Vena Cava Filter Use in Non-Cancer Patients with Acute Venous Thromboembolism: A Population-Based Study Running title: White et al.; Effect of vena cava filter use in patients with VTE Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 Richard H. White, MD, FACP1; Ann Brunson, MS2; Patrick S. Romano, MD, MPH1; Zhongmin Li, PhD1; Ted Wun, MD2 1 Division Di ivi visi sion si on ooff Ge G General nerral Internal Medicine, Department ne Departme mennt of Medicine,, UC D me Davis avis School of Medicine Medicine, Sacram Sacramento, men nto o, CA CA;; 2Di Division Div visi siionn ooff H Hematology ema mattologyy O ma Oncology, ncol nc ollog ogy, D Department eparrtm tmeent of M Medicine edic ed i in inee UC CD avis av is S ch hoo ool ooff M ed diccinee, S acraame ac mentto, CA A Davis School Medicine, Sacramento, Address for Correspondence: Richard H. White, MD, FACP Division of General Medicine UC Davis School of Medicine Suite 2400, PSSB 4150 V Street, Sacramento, CA 95817 Tel: 916-734-7005 Fax: 916-734-2732 Email: rhwhite@ucdavis.edu Journal Subject Terms: Embolism; Thrombosis; Epidemiology; Complications; Quality and Outcomes 1 DOI: 10.1161/CIRCULATIONAHA.115.020338 Abstract Background—Evidence that vena-cava filters (VCFs) are beneficial is limited. Methods and Results—We retrospectively analyzed all non-cancer patients admitted to nonfederal California hospitals for acute venous thromboembolism (VTE) from 2005-2010. Analysis was stratified by presence/absence of a contraindication-to-anticoagulation (active bleeding, major surgery). Outcomes were death within 30/90 days of admission, and the one-year incidence of recurrent-VTE manifested as pulmonary embolism (PE) or deep-vein thrombosis (DVT). Propensity-score methods were used to account for observed systematic differences in Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 baseline characteristics between patients treated versus not treated with a VCF. Among 80,697 patients with no contraindication-to-anticoagulation, VCF-use (N=7762, 9.6%) did not ignificantly reduce the 30-day risk of death (HR, 1.12; 95%CI, 0.98-1.28). Amongg 3, 33,017 0177 01 significantly patients with active bleeding, VCF-use (N=1095, 36.3%), reduced the 30-day riskk of ddeath eath ea th bby y 32% (HR, 0.68; 95%CI, 0.52-0.88) and the 90-day risk by 27% (HR, 0.73; 95%CI, 0.59-0.90). VCFF-us Fusee (N=489, us (N=4 (N = 899, 33.8%) did not reduce mortality mortali liity aamong mong 1445 patients ts w ho underwent major VCF-use who urg gery (HR, R 1.1 .1 1; 95 995%CI, %CI, %C I, 00.71-1.77). .711-1. 11.77 1. 77). 77 ). IIn n al ll su subgrou ups ffilter-use ilte il terte r-us ru e did did not not reduce r duce re ce the the riskk of surgery 1.1; all subgroups ubs bsequent PE. H owev ver, th he risk of su ubseque uent D ue VT iincreased ncre nc reassed by re y 50% % am monng VCF CF patients paatiientss subsequent However, the subsequent DVT among wi ith nno o co cont ntrain nt in ndi dica c ti ca tion on ((HR, HR R, 1. 11.53; 53;; 95 53 5%C %CI, I, 1.3 .34.3 4-1.74 474)) an 74 andd by 135 35% 35 % am amo ong VC ong VCF F pati ppatients a ien entss w ithh with contraindication 95%CI, 1.34-1.74) 135% among with acti ac tive ti vee bbleeding leed le edin ed ingg (H in (HR R, 22.35; .35; 35; 995%CI, 5%CI 5% CI, 11.56-3.52). CI .56 56-3 3.52) 52) 2). active (HR, Conclusions—VCF-use significantly reduced the short-term risk of death only among patients with acute VTE who had a contraindication-to-anticoagulation because of active bleeding. These results support the findings of a randomized clinical trial and current guidelines, which recommend VCF-use only in patients who cannot receive anticoagulation treatment. Key words: epidemiology; thrombosis; pulmonary embolism; venous thromboembolism 2 DOI: 10.1161/CIRCULATIONAHA.115.020338 Introduction Vena cava filters (VCFs) are frequently deployed in patients who are hospitalized for acute venous thromboembolism (VTE). If anticoagulant treatment cannot be given, placement of a VCF may be the only treatment option available to reduce the risk of pulmonary embolism (PE) or dying of fatal PE. However, there is no strong evidence that use of a VCF prevents either death or further pulmonary embolism (PE).1-3 Instead, observational studies of VTE patients treated with a VCF in conjunction with anticoagulant therapy have reported numerous Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 thrombotic and embolic complications.4-9 In a study of VCF use in California hospitals,10 we observed wide variation in the use of VCFs, from 0% to 40%, among patients admitted to different hospitals with a principal priinc ncip ipal ip al diagnosis dia iagn gnos gn o is of VTE, even after adjusting for important clinical parameters. These findings suggest that patients patien ents en ts with wit ithh similar simi mila mi l r clinical characteristics are being la bein be ing treated or not treated in treat ated at e with a VCF based largely practice arg gely on theirr hhospital’s ospi pittal’ pi l’ss prev l’ pprevalent rev eval alen al entt pr en raaccticee ppattern. attern n. relatively with In two wo rel e ative at elyy small smaall al randomized r nddom ra o iz izeed clinical clinnicaal trials, triaalss, VCF VCF use usse was w s studied wa sttuddied ied in patients pat atienntss wit th acute acut ac ute te DV DVT T 3 aand nd aacute cute cu te P PE E 11 who ho also als lsoo re rece received ceiv ce ived ed sstandard tand ta ndar nd ardd an ar anti anticoagulation tico ti coag co agul ag ulat lat atio ionn th io ther therapy. erap er apy. V ap VCF CF use se provided no significant survival benefit and did not reduce the frequency of subsequent PE in patients who presented with PE.11 In the study of patients who presented with acute DVT (randomized to a permanent or no VCF), the incidence of pulmonary embolism was decreased in patients randomized to VCF use, but recurrent VTE manifested as acute DVT was increased. 3 Current 2016 American College of Chest Physicians (ACCP) guidelines recommends use of a VCF only in patients with acute VTE who have a contraindication to anticoagulation, such as recent or planned major surgery or active bleeding. Conducting a randomized study of VCF use in patients with a contraindication to anticoagulation would be exceptionally difficult 3 DOI: 10.1161/CIRCULATIONAHA.115.020338 because most clinicians would feel compelled to insert a VCF if they were unable to administer anticoagulant treatment. However, it is possible to use administrative health data to identify patients with acute VTE patients who have a contraindication to anticoagulation, and determine the effect of VCF use by propensity-score methodology to analyze important outcomes, such as mortality and recurrent VTE. The aim of this study was to use a large observational dataset to analyze the effectiveness of VCF use in patients hospitalized for acute VTE who either had or did not have a Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 contraindication to anticoagulation therapy. The principal outcomes of interest were death within 30 or 90 days of admission, and recurrent VTE manifested as either PE or DVT within effect shortone year of hospital discharge. We hypothesizedd that VCF use would have no effe fect fe ct oon n sh shor ortor tterm erm survival, but would increase the incidence of recurrent VTE manifested as DVT. Methods Metthods Me Study Sttud udyy Design Desiign This using data. Wee de determined death This was as a rretrospective etro et rosp ro spec sp ecti ec tive ti vee sstudy tudy tu dy usi sing si ng oobservational bser bs erva er vati ati tion onal on al da data ta. W ta dete term te rmin rm ined in ed tthe he iincidence ncid nc iden id ence en ce ooff de deat athh at or recurrent VTE among adults without active cancer who were admitted to a non-federal California hospital for acute VTE during a 6-year period from 2005 through 2010 and who either received, or did not receive, a VCF during the index hospitalization. We stratified the study population by the presence or absence of a contraindication to anticoagulation therapy. Database The study was conducted using the California Patient Discharge database (PDD) linked with the California Emergency Department (ED), and master death file databases.12 This data is collected and maintained by the California Office of Statewide Planning and Design (OSHPD). The PDD 4 DOI: 10.1161/CIRCULATIONAHA.115.020338 contains administrative hospital discharge data, as required (and audited) by the Center for Medicare & Medicaid Services (CMS). The database includes demographic information, a principal diagnosis for the hospitalization and up to 25 additional clinical diagnoses, and a list of up to 20 major procedures performed on every patient hospitalized in all non-federal acute-care hospitals in California (the PDD includes 95-97% of all discharges in the state). The ED records include similar data for cases evaluated, but not admitted, at all hospital-affiliated ED’s. Serial hospital/ED records can be linked using an encrypted form of the social-security number called Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 the record-linkage-number (RLN) that is generated by OSHPD for the 95% of patients who have a social security number. The PDD and ED data sets do not list the medications prescribed to the not reliable patient. Codes for the administration of parenteral anticoagulation exist but are no ot re reli liab li able ab le aass there here is no information about intensity of anticoagulation therapy, duration off therapy or overall adequacy events adeq qua uacy cy ooff an aanticoagulation tiico coagulation therapy. VTE event ntts di ddiagnosed agnosed and treatedd in in the out-patient arena are not ar not include includ de in either eitthe herr the the PDD PDD orr ED ED data ta bases. bases. Medical Medi diccal diagnoses di diag di a noses and an nd major m jo ma jorr procedures prooced durres (with (w h date datee performed) per errfo orm med e ) are arre coded code d d using usiing us International-Classification-of-Disease, V-9, Clinically-Modified Since nte tern rnat rn atio at iona io nall-Cl na Clas Cl assi as sifi si fica fi cati ca tion ti on-of of-Dis Dis isea ease ea se, V se -99, C lini li nica call lly-Mo ll Modi Mo difi di fied fi ed ((ICD-9-CM). ICD IC D-99-CM CM)). S CM ince in ce 11996 9966 al 99 alll medical diagnoses in the PDD require a present-on-admission (POA) indicator, which indicates if the condition was present-on-admission (POA=Y/W) or developed during the hospitalization (POA=N/U).13 Cause of death on death certificates is coded using ICD-10. This study was approved by the California Health and Human Services Agency’s Committee for the Protection of Human Subjects, and the University of California, Davis Human Research Protections Program. Study Population Figure 1 shows a schematic flow diagram of the criteria applied to assemble the study 5 DOI: 10.1161/CIRCULATIONAHA.115.020338 populations. We identified all cases age 18 years or older who had a RLN. Cases were excluded if they had a VCF previously placed, back to July 1, 1991, or if there was a diagnosis of active cancer (140.0-209.9, except for non-melanoma skin cancer, 173.0-173.9) either at the time of the index hospitalization or during any hospitalization within the prior 6 months. Acute VTE Cases admitted for acute VTE were identified by the presence of a specific principal diagnosis code (see Supplemental Table 1) for either PE or lower-extremity DVT. For each linked record, Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 only the first hospitalization for acute VTE was analyzed. Cases coded as having both DVT and PE were classified as having a PE. ICD-9-CM codes for a principal diagnosis of acute lower predictive extremity DVT or acute PE have been validated and shown to have a positive pred edic ed icti ic tive ti ve vvalue alue al ue of approximately 95%.14 The POA indicator has also been validated. 15 Vena na a Cava Cav avaa Filter F lter Fi er U se Use VCF VC F placemen placement nt wa w wass id iden identified enti en tiffied ti fied bby y th the pr ppresence reesen ncee off th the he IC ICDICD-9-CM D-9D9-CM 9CM ccode ode od de 38 338.7 8.7 ((interruption inte in terr te rrup rr uptiion ooff th up thee vena ve ena cava). VC V VCF F rremoval emo oval is IICD-9-CM=39.99 CD-9 CD 9-C -CM M=39. 9..99 9 or C CPT=037203, PT= T=00377203 T= 7203 3, bu but ut these t esee codes th co ode des were werre we encountered enco en coun co unte nte tere redd so iinfrequently re nfre nf requ re quen entl en tly (3 tl (314 14 ooff 93 9346 46 ccases, ases as es, 33.4%) es .44%) tthat hatt re ha retr retrieval trie tr ieva ie val al was waas no nott iincorporated ncor nc orpo or pora po rate ra tedd in te into to any analysis.16 Active bleeding To isolate the cases who were likely to have not received anticoagulation during all or part of the hospital stay, we identified all cases with active bleeding, which was ascertained using specific ICD-9-CM codes (Supplemental Table 2) that have been validated in other studies. 17,18 and the bleeding was classified as intracranial, gastrointestinal, and other. Hematuria and epistaxis were included as ‘other-bleed’ only if there was coding for one or more blood transfusions (99.00, 99.03, 99.04-99.07). Bleeding events were categorized as being “present-on-admission” 6 DOI: 10.1161/CIRCULATIONAHA.115.020338 (POA=Y/W) or “not present on admission” (POA=N/U) based on the POA indicator code, which accompanies each ICD-9-CM diagnosis codes. Surgery To isolate another subgroup of cases who were likely to have had anticoagulation withheld during all or part of the hospital stay, we identified cases that underwent a major surgical operation during the hospital stay. Major diagnostic or therapeutic operating room procedures were defined using a modification of the CMS reference codes,19 specifically excluding VCF Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 insertion (38.7) as a surgical procedure. Vascular procedures for venous thrombectomy or procedures used in conjunction with thrombolysis were not included in the definition of major urgery. (Supplemental Table 3) surgery. Outcomes Prin nci cipa pall ou pa outc tcom tc mess were death 30 days and 90 90 days, d ys, recurrent VTE da E manifested manifested as PE (+/Principal outcomes DVT DV T) or DVT al alon ne 1 year yeaar ye ar ooff ho hosp spit sp ital it a dis isch char ch a gee. A rrecurrent ecur ec urrre rent n V nt TE eevent vent aafter fter ft er tthe he iindex nddex DVT) alone 1 hospital discharge. VTE ad dmi miss s ion was wa defined d finedd by de by a hospital hos ospi p taal readmission reeaddmis issiion orr ann ED ED visit visiit with visi with h a pr princ cipa cipa pall di iagnosiss of iag admission principal diagnosis acut ac ute te DV DVT T or P E, oorr by by a ddiagnosis iagn ia gnos gn osis os is ooff ac acut ute te VT VTE E du dduring uri ring ri ng a ssubsequent ub ubse bse sequ quen entt ho en hosp spit sp ital it aliz al izat atio at ionn (P io (POA OA=N) N) acute PE, acute hospitalization (POA=N) that was within the specified follow-up time period. Covariates Covariates included demographic variables, comorbidities, prior surgery, severity-of- illness, PE versus DVT, use of thrombolysis day 1-2, and bleeding present on admission. Pre-existing comorbid conditions (n=23) were defined using the Elixhauser co-morbidity software.20,21 Proprietary software from 3MTM (APR-DRG grouper, V-24) was applied to every record to generate a severity-of-illness (SOI) level and a risk-of-mortality (ROM) level at the time of admission.22,23 Software categorized each of these parameters as mild, moderate, major or 7 DOI: 10.1161/CIRCULATIONAHA.115.020338 extreme. Hospital characteristics were specified based on size (< 100, 101-200, >200 beds), type (private, Kaiser-affiliated, academic), location (rural versus urban) and kind (profit, not-forprofit). Analytic and Statistical Methodology All cases were classified into one of three mutually exclusive groups: Group_1 had neither of the two identifiable contraindications to anticoagulant treatment; Group_2 had a transient contraindication to anticoagulant therapy because they underwent a major operating room Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 procedure, with or without active bleeding, during the index hospitalization or within 3 days prior to admission; Group_3 had had a transient or permanent contraindication to anticoagulant therapy during herapy because they had active bleeding, either present on admission or arising du duri ring ri ng tthe he hospitalization. In the absence of any reliable data (coding) to define ongoing therapeutic anticoagulation, likely antico coag co agul ag ulat ul atio at ion, this io thi h s grouping allowed us to distinguish disti ting ti ngui ng u sh patients who li ike kely received therapeutic anticoagulation treatment VTE patients who probably delayed, antiicoagulationn ffor an orr tr reatm tmen entt of tthe en he aacute cutte V cu TE from om m pat atie at ient ie ntss wh nt ho pr prob obab ob ably rreceived ecei eive ive vedd de del laye yed, ye d, interrupted, anticoagulation nte t rr rrup u ted, d,, iinadequate nade d quatte or noo an antico oaguulaatio on ttreatment. reatm tm men nt . Propensity Pr Propen sit si ity SScore core A logistic regression model was used to estimate each patient’s propensity of having a VCF inserted, based on factors likely associated with VCF use, 10 including hospital variables (type, location, size), demographic variables (age, sex, race/ethnicity) and clinical variables (e.g., PE, proximal DVT, distal DVT alone, comorbid conditions, severity-of-illness). Hospital-specific fixed effects were used to proxy for unobserved patient characteristics that may differ across hospitals. Outcomes were compared using models that used either inverse probability of treatment weighting, or matching based on the propensity score using the calipers method, with caliper = 0.001.24 The effect of propensity score adjustment on the standardized mean differences 8 DOI: 10.1161/CIRCULATIONAHA.115.020338 in baseline covariates between the cases who received or did not receive a VCF was determined for each subgroup.25 Immortal-time-bias Because the day of VCF insertion can vary from early to late during the index hospitalization, analysis of the effect of VCF use is subject to immortal time bias.26-28 Cases who did not receive a VCF might have died before a VCF could be inserted, whereas all of the VCF-treated cases were alive when the VCF was inserted (hence they were ‘immortal’ during the interval from the Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 date of admission to the date of insertion). To account for this bias, VCF use was entered as a time-dependent covariate in the inverse probability weighted proportional hazard models for death.25 In propensity-matched analyses, cases not treated with a VCF had to be alive ali live ve on on the the hospital day when the matched VCF case had the filter inserted.28 Outc Ou tcom tc o es were were modeled using proportional proportion onal on al hazard methodology gy after f testing for the Outcomes pproportionality rop portionalityy assumption. assump mpti mp tion ti on.. For on For cases case ca ses in Group_1 se G ou Gr oup_ p 1 and and Group_3, Grou Gr ouup_ p_3, 3 death dea eath ea th was was aanalyzed nallyze na lyze zedd att 30 30 days, da ays ys,, and 990 0 day da days ays fr from rom m thee da dayy of adm admission, dmissi dm siionn, w whereas heerea eaas for fo or Group_2, Grou Gr up_ p_2, 2 death death hw was a ana as analyzed nallyzedd from na om the he da day of ssurgery. urrge urge gery ry. The The demographic demo de mogr mo grap gr aphi ap hicc and hi and clinical clin cl inic in ical ic al risk risk ffactors acto ac tors to rs use used sedd fo se forr ad adju adjustment just stme st ment me nt iin n th thee mo mode models dels de ls for death (and recurrent VTE) in each group are shown on Forest plots (Figures 2-4), and included bleeding subtypes, PE (versus DVT) and risk-of-mortality. Categorical data were analyzed using Chi-square testing. Analyses were performed using SAS® V 9.3 and a two-sided p-value less than 0.05 was considered statistically significant. All models were tested for proportionality using log-negative log plots. Results The clinical characteristics of the cases in each subgroup are shown in Table 1. Across all three 9 DOI: 10.1161/CIRCULATIONAHA.115.020338 subgroups, cases treated with a VCF were older, had more chronic comorbidity, had a higher frequency of proximal DVT, and had greater severity-of-illness and risk-of-mortality at the time of admission. Figures 5-7 show the standardized mean differences of these parameters in each group before and after adjusting for propensity score using inverse probability weighting. As shown in these figures, the inverse probability of treatment weighting made the characteristics of the VCF and No-VCF groups more balanced compared to the observed differences. Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 A VCF was placed in 7762 of the 80697 (9.6%) cases who had no contraindication to anticoagulation, in 489 of the 1445 (33.8%) cases who underwent a major operating room procedure, and in 1095 of the 3017 (36.2%) cases who had active bleeding. Overall, Overral all, l, 19.4% 19. 9.4% 4% of of the he VCFs were inserted on hospital day 1, 29.2% on day 2, 17.8% on day 3 and 11.1% on day 4, with h 22.5% 22. 2.5% 5% inserted ins n errte ted after this day. effect outcomes summarized Table The ef ffe fecct off VC VCF F use use on on tthe he pprincipal rinncip ipall out ip tco omees is is sum umma um mari rize ize zedd in Ta abl blee 22.. As shown sho hown wn in in Figure use VCF significantly improve survival Group_1 Fiigu gure r 2, us se off a VC CF did nnot ot sig igni ig nifica ni cant ca n ly im mprov ve su surv r iv val aamong mong mo ng the he Gr roup_1 rou up ccases, ases, whoo had unadjusted mortality had no ccontraindication ontr on trai tr aind ai ndic nd icat ic atio at ionn to aanticoagulation. io ntic nt icoa ic oagu oa gula lati la tion ti on. Th on Thee unad un nad adju just sted st ed 330-day 0-da day mo da mort rtal rt alit al ity ra it rate te was as 44.5% .5% 5% in tthe hee VCF cases and 3.3% in the no-VCF cases. Using inverse probability weighting and adjusting for immortal time bias, the risk of death within 30 days was not significantly higher among VCFtreated cases (HR, 1.12; 95% CI, 0.98-1.28). There was no interaction between VCF use and a diagnosis of PE versus DVT (P>0.5). Similar findings were noted using matching methodology. The unadjusted incidence of PE 1 year was 2.8% among VCF cases and 3.2% among no VCF cases. There was no difference in the risk of recurrent PE within 1 year after discharge in the adjusted model (HR, 1.05; 95% CI, 0.89-1.24). The incidence of DVT 1 year was 5.4% among VCF cases and 3.7% among no-VCF groups. The adjusted risk of recurrent DVT within one 10 DOI: 10.1161/CIRCULATIONAHA.115.020338 year was significantly higher among VCF treated cases than among no-VCF cases (HR, 1.53; 95% CI, 1.34-1.74). Among the Group_2 cases who underwent surgery, VCF use did not affect: the 30-day risk of death (HR, 1.12; 95% CI, 0.71-1.77), the one-year risk of recurrent PE (HR, 0.85; 95% CI, 0.35-2.10), or the one-year risk of recurrent DVT (HR, 1.15; 95% CI, 0.57-2.32). Among Group_3 cases, all of whom had active bleeding, VCF use was associated with a crude 30 day mortality of 9.5% among VCF cases 11.5% among no-VCF cases. There was a Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 statistically significant 32% reduction in the adjusted risk of death within 30 days, after inverse probability weighting of the propensity and adjusting for immortal time bias (HR, 0.68; 95% CI, 0.52-0.88). The Forest plot is shown in Figure 3. Similarly, using propensity-matching propensity-mattch chin ingg an in andd adjusting for immortal time bias, the risk of death associated with VCF use was reduced by 39% (HR, CI, hazard HR, R, 00.61; .61; .6 1; 995% 5 C 5% I 0.39-0.95). Proportional haza I, zard za rd models for death 9900 days after admission showed VCF remained with similar how owed that VC CF uuse se re ema main ined in ed aassociated ssocia iateed wi ia w th a si imila larr significant la sign si gniffic gn ican nt reduction reedu duction on in in the the risk riisk of of death deaath de (HR, who HR, 0.73; 95% 95% 5 CI, CI, 0.59-0.90). 0..599-0.90 90). 90 ). The he distribution diist strribu utioon of of the thee day day off death deaath h among amo ong tthe he cases es wh ho received receive Figure ece ceiv ived ed oorr di didd no nott re rece ceiv ce ivee a VC VCF F is shown sho hown wnn in in Fi Figu gure re 88,, which whhic whic ichh demonstrates demo de mons mo nstr ns trat tr ates at es that tha hatt the the majority majo ma jori jo rity ri ty of of deaths in the first 8 days occurred in cases who did not receive a VCF, making adjustment for immortal time bias mandatory. In the Group_3 cases, who had active bleeding, the crude incidence of PE 1 year was 2.6% for VCF cases and 2.6% among the no-VCF cases, and there was no significant reduction in the adjusted risk of recurrent PE within 1 year (HR, 1.04; 95% CI, 0.67-1.61). Performing a sensitivity analysis to include recurrent PE events coded as occurring during the index hospitalization (POA=N/U), there was no change in the results. In this group, the crude incidence of recurrent DVT 1 year in the VCF cases was 6.1% versus 2.5% in the no-VCF cases, and in 11 DOI: 10.1161/CIRCULATIONAHA.115.020338 the adjusted risk model, recurrent DVT within 1 year was significantly higher among VCF cases (HR, 2.35; 95% CI, 1.56-3.52). Among the 1095 cases with active bleeding who received a VCF, 104 (9.5%) died within 30 days compared with 221 (11.5%) deaths among the 1922 cases who did not receive a VCF. Twenty-four of the 104 (23.1%) VCF treated cases who died had VTE listed as the cause of death, and 8 (33%) of these VTE deaths had a death-certificate diagnosis of PE (ICD-10= I26.9). In comparison, 57 (25.8%) of the deaths among the cases not treated with a VCF had VTE listed Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 as the cause of death, and 38 (66%) of these VTE deaths had a death-certificate diagnosis of PE. Discussion The principal findings of this study were that VCF use did not prevent either death or recurrent venous who veno ous thromboembolism thr hrom omboem om embolism manifest as PE in patients em pattie ient nts with acute VTE w nt ho had no contraindication anticoagulation, VCF reduce day death contraindicati con nt ion n too an ant tico coag co agul ag ulat ul atio at ion, io n, bbut ut V CF uuse se ddid idd red educ ed ucee th uc tthee 30 aand nd 90 da ay ri risk isk ooff de deat athh by at approximately therapy ap ppr prox o imatel elyy 30% el 300% inn patients patienntss with w th wi h active acctiv ve bleeding bleeeding ng who who likely lik kelly had hadd anticoagulation a tiico an oag gul ulat a ionn th ther rap py withheld admission index hospitalization. finding wiith with thhe held he ld eeither ithe it herr at tthe he he ttime imee of ad im admi miss mi ssio ss ionn or dduring io urrin urin ingg th thee in inde dex ho de hos spi pita tali ta liza li zati ati tion on. Th on Thee fi find ndin nd ingg th in that at tthe he lower risk of death persisted out to 90 days after admission suggests that VCF use did not simply delay the time of death. These findings are consistent with the recently published findings from the Prevention of Recurrent Pulmonary Embolism by Vena Cava Interruption 2 (PREPIC 2) clinical trial study,11 which reported that VCF use had no effect on morality or recurrent PE among patients who presented with PE who had no contraindication to anticoagulant therapy. Our findings support the recent 2016 American Society of Chest Physician guidelines,29 which recommends that a VCF be used for early treatment only in patients with an acute proximal lower extremity DVT or PE who cannot receive anticoagulant therapy. 12 DOI: 10.1161/CIRCULATIONAHA.115.020338 The rather striking reduction in short-term mortality associated with VCF use among the acute VTE patients who had active bleeding was not expected. In an analysis of a small number of patients in a European VTE registry who were at increased risk for bleeding during anticoagulation,30 VCF use was not associated with a significant reduction in death within 30 days. This study did report that VCF use was associated with a significant reduction in death attributed to PE, but no adjustment was made for immortal time bias.31 Similarly, several retrospective studies have reported a beneficial effect of VCF on mortality in highly specified Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 subgroups of patients with acute VTE who were at high risk for dying, but again, none of these studies used rigorous methodology to account for immortal time bias.30,32-34 A possible explanation for the observed reduction of mortality associated wi with th V VCF CF uuse se in n acute VTE patients who had active bleeding is treatment bias, i.e., ordering insertion of a VCF preferentially pref fer eren enti en tial ti ally al ly in le less s ill patients. However, the ov overall ver eral a l level of severityseverity-of-illness -of of-illness and risk-ofmortality treated m orrtality (ROM) (ROM M) at at the the time tim imee of admission adm dmis issi is s on was wass significantly sig gniifica caant ntly ly higher h gh hi gher err among amo mong ppatients atie at ient ie ntss tr nt trea ated ted wi with th a VCF patients who VCF. admission, VC F compared comp par ared e tto o the pa atien nts t w hoo ddid id d nnot ot rreceive ecceivee a V CF At tthe CF. hee ttime im me of of adm dmis dm i sion onn, 46.8% of the bleeding VCF placed he patients pati pa tien ti ents en ts with wit ithh major it majo ma jorr bl jo blee eedi ee ding di ng who ho hhad ad a V CF pl plac aced ac ed were wer eree classified er clas cl assi as sifi si fied fi ed as as having havi ha ving ing major maj ajor or or or extreme ROM , versus only 31.3% of the patients with major bleeding who did not receive a VCF. In fact, severity-of-illness and risk-of-mortality were significantly higher among the patients who had a VCF inserted in each of the three groups that were analyzed. Moreover, in the models for death, we adjusted for important risk factors for death including age, race, insurance status, and the number of comorbidities in addition to the risk-of-mortality. Although the acute VTE patients who required major surgery likely had anticoagulation discontinued. at least transiently, during the perioperative period, use of a VCF had no effect on the risk of death within 30 or 90 days. We had no direct measure of the duration or intensity of 13 DOI: 10.1161/CIRCULATIONAHA.115.020338 anticoagulant therapy, but we suspect that patients who underwent an unrelated major procedures during their hospitalization for VTE were probably highly selected, and these patients likely had therapeutic anticoagulation withheld for a relatively brief period. By contrast, patients with active bleeding were not selected (i.e., most of these events were probably unexpected complications of initial VTE therapy) and probably had a longer mean duration without receiving anticoagulation. The finding that VCF use reduced the short-term risk of death among patients with active Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 bleeding but had no long-term effect on the risk of recurrent PE may seem contradictory. If a VCF provides a physical barrier that prevents or retards the migration of embolus into the lungs, then PEs, hen VCF use should theoretically prevent both early deaths and symptomatic PEs Es, at lleast Es east ea st uuntil ntil nt the he time of VCF retrieval. Several factors may explain this paradox. First, many patients with probably therapy active ve bbleeding leed le edin ed ing pr in prob o ably had anticoagulation ther erap er apy restarted as soon as the risk of further ap bleeding bl eding was deemed blee dee eem med too be be sufficiently suuff ffiicie icie ient n ly low. nt low w. Second, Secondd, a modest m de mo dest s proportion pro ropo port po rtio rt ionn off the io the VCFs VCF CFss might migh mi ghtt gh have removed within weeks insertion. ha ave bbeen een re emo m ved wi ith hin a few ew w e ks ooff in ee nseertio on.16 T Third, hird rd d, VCF VC us use se might mighht hhave mi ave iimproved mpro oved urv rviv ival al not not by by “catching” “cat “c atch at chin ch ing” in g” large lar arge ge clots clo lots ts that tha hatt embolized embo em boli bo lize li zed ed toward towa to ward ard the the lung, lung unng, but butt by by breaking brea br eaki ea king ki ng upp these thes th esee es survival embolizing thrombi into smaller pieces that were better tolerated hemodynamically. Although causes of death on death certificates are often not accurate,35 we did find that in the patients with active bleeding who died in the hospital, the proportion deaths of attributed to PE was higher in those who did not receive a VCF compared to those who had a VCF inserted. However, this finding could be due to attribution bias, as physicians may be less likely to blame a death on PE knowing that a VCF had been inserted. It is unlikely that a clinical trial will ever be conducted to assess the efficacy of VCF use among patients with a clear contraindication to anticoagulation therapy because of active 14 DOI: 10.1161/CIRCULATIONAHA.115.020338 bleeding. Such patients are relatively uncommon, they are quite ill, and the need to obtain prompt informed consent would make conducting such a study extremely challenging. The current retrospective analysis represents an attempt to determine whether VCF use may be beneficial among patients who have at least a temporary contraindication to anticoagulation. To minimize sources of bias inherent in retrospective observational studies, we used propensity score analytic methods to balance the distribution of all observed baseline characteristics 24, and we adjusted for immortal-time-bias.26-28 Immortal-time-bias arises in Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 observational studies when patients who appear to be eligible for an intervention, such as VCF insertion, die before they can receive the intervention yet they are included in the analysis and categorized as not receiving the intervention. Accepted methods off circumventingg this thi hiss bias bias to to ensure a fair comparison of VCF use and non-use were incorporatedd into our analysis.26,28 In our anallys ysis is ooff th the pa ati tients who had a contraindicati ion to to anticoagulation because beccause of active bleeding, be analysis patients contraindication the he reduction reduction inn the the risk ris iskk of of dying dyin dy ingg associated in asso as s ci c ated ed dw with ith VC VCF CF us use fe fell l ffrom ll r m 53 ro 53% % to 332% 2% aafter fter ft er aaccounting cccou ount ntin nt i g for foor im immortal immortal-time-bias. al-t al - im -t me--biaas. Limitations Limi Li mita mi tati ta tion ti onss of tthis on hiss an hi anal analysis alys al ysis sis include inc nclu lude de tthe he absence abse ab senc se ncee of any nc any ny information inf nfor orma or mati ma tion ti on about abo bout utt the the use, use se, se intensity, or duration of anticoagulation therapy. Except for patients with active bleeding and those who underwent major surgery, we assumed that anticoagulation therapy was administered to all patients. Active bleeding is likely an excellent surrogate for avoidance of anticoagulation. To the extent that patients classified as having major bleeding actually had only minor bleeding, and were either maintained on anticoagulation or taken off for only a short period of time, our findings may have underestimated the true effect of VCF-use on mortality among VTE patients who did not receive anticoagulation therapy. Because fewer than 5% of the VCF treated patients had a procedure code for VCF removal, we did not adjust for removal. The findings of this study 15 DOI: 10.1161/CIRCULATIONAHA.115.020338 do not apply to patients with cancer, who were excluded because of the difficulty of adjusting for their short-term risk of death. Finally, our data base did not allow us to identify any patients who had recurrent DVT diagnosed and treated only in the outpatient arena. However, because the proportion of patients diagnosed with recurrent DVT in the outpatient setting was likely similar among patients treated and not treated with a VCF, the results of our proportional hazard modeling are likely unaffected. In summary, the findings of this retrospective analysis of observational data suggest that Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 VCF use reduced the short-term risk of death among VTE patients without cancer who had a temporary contraindication to anticoagulation due to active bleeding. This lower risk of death was not observed among patients who had no contraindication to anticoagulationn ttreatment, reat re atme at ment me nt,, an nt and itt was not observed in patients who had a temporary contraindication to anticoagulation because they hey y underwent und nder erwe er went major we maj a or surgery. Among patients with with or without a contraindication contra raain indication to anticoagulation, associated with higher risk diagnosed antiicoagulation, an n iinsertion nser erttion er on ooff a VC VCF F wa was asso soociated ci d w itth a hi high g er ri iskk ooff bbeing ein ng di diag agnnose ag nose sedd wi with th acute within filter was ac cut utee DVT wi withi in onee yyear, ear, bbut ut filt ltter e uuse see w as not associated asssoociiate ia ed with with th h a reduction red e uc uction on in in the th he risk riskk of of recurrent ecu curr rren rr entt PE within en wit ithi it hinn one hi one year. yeear year ar. Funding Sources: This study was supported by the Hibbard E Williams Endowment at UC Davis. Conflict of Interest Disclosures: None. References: 1. Girard P, Meyer G, Parent F, Mismetti P. Medical literature, vena cava filters and evidence of efficacy. A descriptive review. Thromb Haemost. 2014;111:761-769. 2. Prasad V, Rho J, Cifu A. The inferior vena cava filter: How could a medical device be so well accepted without any evidence of efficacy? JAMA Intern Med. 2013;173:493-495. 16 DOI: 10.1161/CIRCULATIONAHA.115.020338 3. Group TPS. Eight-Year Follow-Up of Patients With Permanent Vena Cava Filters in the Prevention of Pulmonary Embolism. Circulation. 2005;112:416-422. 4. Nicholson W, Nicholson WJ, Tolerico P, Taylor B, Solomon S, Schryver T, McCullum K, Goldberg H, Mills J, Schuler B, Shears L, Siddoway L, Agarwal N, Tuohy C. Prevalence of fracture and fragment embolization of Bard retrievable vena cava filters and clinical implications including cardiac perforation and tamponade. Arch Intern Med. 2010;170:1827-1831. 5. White RH, Zhou H, Kim J, Romano PS. A population-based study of the effectiveness of inferior vena cava filter use among patients with venous thromboembolism. Arch Intern Med. 2000;160:2033-2041. Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 6. Wittenberg G, Kueppers V, Tschammler A, Scheppach W, Kenn W, Hahn D. Long-term results of vena cava filters: experiences with the LGM and the Titanium Greenfield devices. J Vasc Interv Radiol. 1998;21:225-229. 7. Rajasekhar A, Streiff MB. Vena cava filters for management of venous thromboembolism: a 2013;27:225-241. clinical review. Blood Rev. 2013;27:225 241. LaMuraglia GM. 8. Malek JY, Kwolek CJ, Conrad MF, Patel VI, Watkins MT, Lancaster RT, LaMu Mura Mu ragl ra glia gl ia G M. Presentation and treatment outcomes of patients with symptomatic inferior vena cava filters. Ann Vasc Surg. 2013;27:84-88. (IVC) Communication: 9.. U.S. U.S. S Food od d and Drug Administration. Inferior Vena Venna Cava (IVC C) Filters: Initial Communication Risk Adverse Events with Term Bethesda, MD; R iskk of Advers se E ven ents en ts w ithh Lo it Long ng T errm Us Use. Bet theesdda, M D; 22010. D; 0 0. 01 hhttp://www.fda.gov/safety/medwatch/safetyinformation/safetyalertsforhumanmedicalproducts/uc ht tp p://www.fda.ggov/saafeety/me medwatch/ssaffetyiinforma me nf maationn/s /saf afet af etyaleerttsforhu et humaanmedic nm iccalprod al ductss/u uc m221707.htm Accessed Mar 2016. m221 m2 21707.httm A 21 cccesssedd Ma ar 28, 28 20 2016 6. 10. White 10 Whit Wh itee RH, it RH Geraghty Gera Ge ragh ra ghty gh ty E, E, Brunson Brun Br unso nso sonn A, A, Murin Murin urrin S, S, Wun Wunn T, Wu T, Spencer Spe penc ncer nc er F, F, Romano Roma Ro mano ma no PS. PS. HIgh HIg Ighh variation vaari vari riat atio at ionn io between hospitals in vena cava filter use for venous thromboembolism. JAMA Intern Med. 2013:1-7. 11. Mismetti P, Laporte S, Pellerin O, Ennezat PV, Couturaud F, Elias A, Falvo N, Meneveau N, Quere I, Roy PM, Sanchez O, Schmidt J, Seinturier C, Sevestre MA, Beregi JP, Tardy B, Lacroix P, Presles E, Leizorovicz A, Decousus H, Barral FG, Meyer G. Effect of a retrievable inferior vena cava filter plus anticoagulation vs anticoagulation alone on risk of recurrent pulmonary embolism: a randomized clinical trial. JAMA. 2015;313:1627-1635. 12. Office of Statewide Health Planning and Development. Data Request Center. Available at: (http://www.oshpd.ca.gov/HID/Data_Request_Center/Types_of_Data.html) Accessed Mar 28, 2016. 13. Kassed C, Kowlessar N, Pfunter A, Parlato J, Andrews RM The Case for the POA Indicator: Update 2011. ONLINE November 1, 2011. U.S. Agency for Healthcare Research and Quality. Available: http://www.hcupus.ahrq.gov/reports/methods/methods.jsp. 17 DOI: 10.1161/CIRCULATIONAHA.115.020338 14. White RH, Garcia M, Sadeghi B, Tancredi DJ, Zrelak P, Cuny J, Sama P, Gammon H, Schmaltz S, Romano PS. Evaluation of the predictive value of ICD-9-CM coded administrative data for venous thromboembolism in the United States. Thromb Res. 2010,126:61-67. 15. Khanna RR, Kim SB, Jenkins I, El-Kareh R, Afsarmanesh N, Amin A, Sand H, Auerbach A, Chia CY, Maynard G, Romano PS, White RH. Predictive value of the present-on-admission indicator for hospital-acquired venous thromboembolism. Med Care. 2015;53:e31-36. 16. Peterson EA, Yenson PR, Liu D, Lee AYY. Predictors of attempted inferior vena cava filters retrieval in a tertiary care centre. Thrombs Res. 2014;134:300-304. 17. Fang MC, Go AS, Chang Y, Borowsky LH, Pomernacki NK, Udaltsova N, Singer DE. A new risk scheme to predict warfarin-associated hemorrhage: The ATRIA (Anticoagulation and Risk Factors in Atrial Fibrillation) Study. J Amer Col Cardio. 2011;58:395-401. Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 18. Arnason T, Wells PS, van Walraven C, Forster AJ. Accuracy of coding for possible warfarin complications in hospital discharge abstracts. Thromb Res. 2005;118:253-262. 19. CMS. Appendix E Operating Room Procedures and Procedure Code/MS-DRG Code/MS-DR RG Index. Inde In dex. de x. 2014 201 0144 ed. http://www.cms.gov/icd10manual/fullcode_cms/P0033.html. Accessed Mar 228, 8, 22016. 016. 01 6. 20. Healthcare Cost and Utilization Project (HCUP). HCUP Comorbidity Software. 2008. Agency Healthcare Rockville, www.hcupAgen en ncy ffor or H ealt lthc lt h are Research and Quality, Ro ock ckvi v lle, MD. www.hcu cu upus.ahrq.gov/toolssoftware/comorbidity/comorbidity.jsp. uss.a .ahhrq.gov/ hr v/to v/ toolssoftware/comorbidity/comorbidiity to y.jsp. Accessed ed Mar 28, 2016. 21. 2 . Schoenman JA, 21 JA, Sutton JA Suuttton JP, JP, Elixhauser Elixhau useer A, Love Lovee D. D. Understanding Und ndeersstan ndiing and nd Enhancing Enh nhancinng the nh thee Value Valuue of Hospital Discharge Data. Care Hosp Ho spit sp i al Dis issch c arrgee Da ataa. Med Meed Ca C re Res es Rev. Rev evv. 22007;64:449-468. 0077;64::44 449-46 468. 46 8 22. Health Information Systems. groups 22 3M H ealt ea lthh In lt Info form fo rmat rm atio at ionn Sy io S yst stem st emss. HCUP em HCU CUP P All All Patient Pati Pa tien ti entt Refined en Refi Re fine fi nedd Diagnosis ne Diag Di agno ag nosi no siss Related si Rela Re late la tedd gr te grou oups ps ((APR APR AP R DRGs), Version 24.0. Methodology Overview. 2007. https://www.hcupus.ahrq.gov/db/nation/nis/v24_aprdrg_meth_ovrview.pdf. Accessed Mar 28, 2016. 23. Baram D, Daroowalla F, Garcia R, Zhang G, Chen J.J., Healy E, Riaz A, P R. Use of the All Patient Refined-Diagnosis Related Group (APR-DRG) Risk of Mortality Score as a Severity Adjustor in the Medical ICU. Clin Med Insights Circ Respir Pulm Med. 2008;2:19-25. 24. Austin PC, Schuster T. The performance of different propensity score methods for estimating absolute effects of treatments on survival outcomes: A simulation study. Stat Methods Med Res. 2014. Feb 3. [Epub ahead of print] doi:10.1177/0962280213519716. 25. Curtis LH, Hammill BG, Eisenstein EL, Kramer JM, Anstrom KJ. Using inverse probabilityweighted estimators in comparative effectiveness analyses with observational databases. Med Care. 2007;45:S103-107. 26. Liu J, Weinhandl ED, Gilbertson DT, Collins AJ, St Peter WL. Issues regarding /`immortal time/' in the analysis of the treatment effects in observational studies. Kidney Int. 2012;81:341- 18 DOI: 10.1161/CIRCULATIONAHA.115.020338 350. 27. Giobbie-Hurder A, Gelber RD, Regan MM. Challenges of Guarantee-Time Bias. J Clin Onc. 2013;31:2963-2969. 28. Lévesque LE, Hanley JA, Kezouh A, Suissa S. Problem of immortal time bias in cohort studies: example using statins for preventing progression of diabetes. BMJ. 2010;340:340:b5087. 29. Kearon C, Akl EA, Ornelas J, Blaivas A, Jimenez D, Bounameaux H, Huisman, M, King CS, Morris T, Sood N, Stevens SM, Vintch JRE, Wells P, Woller SC, Moores CL, Antithrombotic Therapy for VTE Disease: CHEST Guideline, Chest 2016;149:315-352. Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 30. Muriel A, Jimenez D, Aujesky D, Bertoletti L, Decousus H, Laporte S, Mismetti P, Munoz FJ, Yusen R, Monreal M. Survival effects of inferior vena cava filter in patients with acute symptomatic venous thromboembolism and a significant bleeding risk. J Am Coll Cardiol. 2014;63:1675-1683. 31. Fernandes TM, White RH. Immortal time bias and the use of IVC filters. J Am C Coll o l Cardiol. ol 2014;64:955. 32. Stein PD, Matta F. Vena cava filters in unstable elderly patients with acute pulmonary embolism. Am J Med. 2014;127:222-225. 33. PD, vena 333. Stein St PD D, Dalen JE, Matta F. Underuse of ven na cava filters in n uunstable nstable patients with acute pulmonary embolism. pulm monary embo boliism bo m. Am J Med. Med ed. 2014;127:6. 201 014; 4 127: 7:6.. 7: 34. T,, Yasunaga 344. Isogai Isog Is o ai T Y su Ya sunagga H, Matsui Mats Mats tsui H, H, Tanaka Tannaka H, Ta H, Horiguchi Hooriggucchi H, H Fushimi Fushhimi m K. K. Effectiveness Effe Ef feectiven enesss of en of Inferior Therapy. Med. nferi rior ri or Vena Vena Cava Cava Filters Fil ilte terrs rs on Mortality Morrtality Mo rt y as an an Adjuvant Adju Ad juva vant ntt to to Antithrombotic Anttith An hro romb mbot mb otic ic T heera rapy py. Am py mJM ed. ed 2015;128:312.e23-31. 2015 20 15;1 15 ;128 ;1 28:3 28 :312 :3 12.e23 e23-31 3 1. 35. Tavora F, Crowder C, Kutys R, Burke A. Discrepancies in initial death certificate diagnoses in sudden unexpected out-of-hospital deaths: the role of cardiovascular autopsy. Cardiovasc Pathol. 2008;17:178-182. 19 DOI: 10.1161/CIRCULATIONAHA.115.020338 Clinical Perspective This observational study analyzed outcomes associated with inferior vena cava filter (VCF) use in 85159 patients who were hospitalized for acute venous thromboembolism (VTE). There was a statistically significant 30% reduction in the 30 & 90-day risk of death in the patient treated with a VCF vs. no-VCF, but this was observed only in the small subgroup (3.5%) of patients who had active bleeding. Among patients (1.8%) who underwent major surgery immediately before or Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 during the acute-VTE hospitalization, use of a VCF was not beneficial. In the remaining 80697 patients (95%) who had no-contraindication to anticoagulation therapy, use of a VCF (N= 7762) had no effect on survival or the risk of developing a pulmonary embolism (PE) w within ithi it hinn 1 ye hi year year, ar, bu ar bbutt VCF-use did increase the risk of recurrent deep-vein-thrombosis (DVT) by 50%. These findings were ree similar sim imil ilar il ar to to the th he results of two small randomized randomiize zedd studies, which reported repoort rted e that use of a VCF did nnot ot reduce the risk ris iskk off subsequent sub ubse sequ se quen qu entt PE, en PE but but did bu did lead lead d too ann increase inc ncre reas re a e inn the as the rrisk issk de dev developing velo velo lopi ping pi ng a nnew ew DVT. this study are alignment with the American College DVT DV T. The ffindings T. indinngs off th in his st tud udyy ar re in n al lign nme ment w ith h th he rrecent ecen ec nt 2016 2016 1 A merrica can Co ca Col lleg ge of Chest Ches Ch estt Ph es Phys Physician ysic sic icia iann An ia Anti Antithrombotic tith ti thro th romb ro mbot mb otic ot ic T Therapy hera he rapy ra py ffor or V VTE TE D Disease isea is ease ea se G Guidelines, uiide uide deli line li ness, whi ne which hich hi ch rrecommends ecom ec omme om mend me ndss th nd that at a VCF not be inserted in patients with acute VTE who can be treated with standard anticoagulation therapy. In light of the American Board of Internal Medicine’s “Choosing Wisely” initiative, the implications are straightforward: do not use a VCF if your patient can receive standard anticoagulation therapy. 20 DOI: 10.1161/CIRCULATIONAHA.115.020338 Table 1. Baseline Characteristics of Acute VTE Patients in each Subgroup. Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 Variable Age <50 50 50-59 0-59 60-69 0-69 70-79 0-79 80+ 0+ Gender er Male Female ema male ma l le Race-Ethnic Ethni nc Non-hispanic on-hi h spanic White African-American frican-American a Hispanic isp spanic ni Asian/Pacific sian/ n/Pa n/ Paci Pa cifi ci ficc Is fi Isla Islander landerr la Other/Unknown ther/Unknown Insurance ance Medicare Medi-Cal, government Private Self-Pay Other/Unknown Bleeding POA=Y POA=N No Bleed Group 1: No Contraindication to Anticoagulation N= 80697 (94.8%) Group 2: Surgery (+/- Bleed) N= 1445 (1.8%) Group 3: Active Bleeding N=3017 (3.5%) % No VCF (N=72935) % VCF (N=7762) p-value % No VCF (N=956) % VCF (N=489) p-value % No VCF (N=1922) % VCF (N=1095) p-value 26.0 16.4 16.5 18.3 22.9 16.1 12.9 15.9 21.1 34.1 <.001 <.001 0.15 <.001 <.001 27.0 16.9 16.3 20.5 19.2 17.8 18.0 18.4 21.3 24.5 0.001 0.62 0.32 0.73 0.02 22.7 14.5 14.55 15 15.3 .33 19.7 27.8 10. 0.7 7 10.7 12.0 12.0 15.9 15 .9 24.7 36.7 <.001 0.049 0.69 <.001 <.001 45.4 54.6 46.1 53.9 0.30 47.5 52.5 50.1 49.9 0.35 0. 35 49.0 51.0 43.2 56.8 0.002 6 69 .0 69.0 11.1 1 13.9 3 3.22 2.8 771.4 1.4 10. 00 10.0 12. 29 12.9 0 3.0 2.8 < .001 <.001 . 0.003 . 0.011 0.22 0.22 0.82 64.3 43 11.77 166.11 16.1 4.7 4.7 3.1 69.1 9.22 9.2 14.33 45 4. 4.5 2.9 0. 0 077 0.07 0.155 0.15 0.377 0.37 0 866 0. 0.86 0.77 68 .22 68.2 11.0 12.9 5.0 0 5.0 3.0 67. 7.7 7. 7 67.7 10. 06 10.6 14. 44 14.4 4.4 4 4.4 2.9 0.78 0.74 0.24 0.45 0.95 49.6 12.0 34.7 3.3 0.3 62.6 8.3 26.4 2.3 0.3 <.001 <.001 <.001 <.001 0.826 52.2 11.4 33.9 2.3 0.2 54.6 10.4 33.3 1.4 0.2 0.39 0.58 0.83 0.26 0.99 56.1 11.2 27.8 4.2 0.6 67.9 8.7 21.7 1.3 0.5 <.001 0.03 <.001 <.001 0.67 - 6.0 4.2 90.5 12.3 12.3 76.5 <.001 <.001 <.001 79.1 22.0 71.6 31.0 <.001 <.001 - 21 DOI: 10.1161/CIRCULATIONAHA.115.020338 Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 Type of Bleed Intracranial Gastrointestinal Other Bleed or Transfusion Major Surgery Yes No Surgery Vascular Procedure Prior Major Surgery <3d Prior Major Surgery <4-60 4-60 d Comorbidities orbidities 0 1 2+ + VTE Event Even nt PE E (+/(+/ + - DVT) D T) DV Proximal rox o imal m DVT Distal istall DVT Severity-of-Illness ity-of-Illness o Minor Mino in r Moderate Mode dera de r te Major Extreme xtr trem emee em THROMBOLYTIC OMBOLYTIC Yes Treatment No Risk of Mortality Minor Moderate Major Extreme 1.3 5.7 - 0.6 4.0 4.9 3.1 10.8 9.6 0.001 <.001 0.006 <.001 85.0 15.0 12.8 96.3 3.7 9.4 <.001 - 15.1 4.1 <.001 0.06 3.7 49.8 46.4 11.0 65.5 23.6 100.0 2.0 100.0 3.5 <.001 <.001 <.001 0.012 - 10.4 12.5 0.19 25.2 18.8 0.002 10.77 14.4 14.4 0.003 15.6 21.9 62.4 7.8 15.8 76.4 <.001 <.001 <.001 13.1 16.0 70.9 4.5 12.3 83.2 <.001 0.058 <.001 9.6 16.4 75.0 3.2 9.9 86.9 <.001 <.001 <.001 55.3 28 2 28.3 .3 16.4 1 .4 16 48.2 334.6 4.6 6 17.2 17. 72 <.001 < <. <.001 001 00.07 0. 07 45.6 41.4 .4 4 13. 13.0 30 52.4 330.3 0.3 0. 3 17 17.4 .44 0.015 0.01 015 <. <.001 .00 001 1 00.024 . 244 .0 68.8 21 21.3 .33 9.99 56.7 226.8 68 6. 16.5 16. 6.5 6. 5 <.001 <.001 <.001 27.8 7 43.6 3.6 25.1 33.5 .5 5 14.4 4 338.5 8.5 8. . 39.2 2 77.9 .9 9 < <.001 .001 <.00 <.001 001 00 < <. <.001 001 1 < <.001 .00 001 00 1 15.8 58 34 34.6 4.6 6 38 38.8 .8 8 10 10.8 .8 8 6.1 1 227.6 7.6 6 45.0 21 21.3 .3 3 <.00 <.001 001 00 0. 0.007 .00 0 7 0.024 < <.001 .00 001 00 1 10.2 336.6 36 .6 6 40.0 40 0 13 13.2 .2 2 44.2 4. 2 24. 24.4 44 51.6 19 19.8 .8 8 <.001 <.001 <.001 <.001 2.0 8.7 <.001 7.1 9.0 0.20 4.0 7.3 T<.001 98.0 91.3 92.9 91.0 96.0 92.7 52.8 30.8 14.0 2.4 30.9 37.4 26.2 5.4 21.7 36.7 20.7 4.9 22.5 38.2 28.4 10.8 34.8 33.9 23.4 7.9 16.8 36.4 35.0 11.8 <.001 <.001 <.001 <.001 22 <.001 0.57 <.001 <.001 <.001 0.15 <.001 <.001 DOI: 10.1161/CIRCULATIONAHA.115.020338 Table 2. Study Outcomes of Death and Recurrent Thromboembolism. Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 Study Sub-Group 1)) No Contraindication To Anticoagulation 2)) Ma Majo Major j r Surgery jo Surger eryy er 3)) Ac Acti Active tive ti ve B Bleeding leed le edin ed ingg in Outcome Death 30 days from admission Hazard Ratio (VCF Use Analytic Method vs. No VCF) 95% CIs Propensity-IPW, Adjusted for ITB 1.12 (0.98-1.28) Propensity-matched, Adjusted for ITB 1.03 (0.90-1.19) Pvalue 0.11 0.61 Death 90 days Propensity-IPW, Adjusted for ITB 1.15 (1.05-1.27) (1. 1 05 1. 05-1 1.2 .27) 7 0.004 PE 1 yr. of discharge DVT 1 yr. of discharge Propensity-IPW Propensity-IPW 1.05 5 1.53 (0 (0.89-1.24) 0.8 .899-1. 91.24 1. 24)) 24 (1.34-1.74) 0.56 <.0001 Death 30 days Propensity-IPW, Adjusted for ITB 1.12 (0.71-1.77) 0.63 after surgery Death 90 days PE 1 yr. r.. after aft fter er surgery sur u ge gery DVT 1 yr. DV r after surge surgery g ry y Death De ath th 30 days daays y ffrom fr om admission adm dmis issi is sion si Death Deat De athh 9900 days at days PE 1 yr yr. of discharge DVT 1 yr. of discharge Propen Propensity-matched, nsi sity ty-m ty - atched, Adjusted for ITB TB Propensity-IPW, Prop ensity n t -IPW, Adjusted ffor o ITB or Propensity-IPW Prop pen ensi sity si ty-I ty -IPW W Propensity-IPW Pr open pen e si sity ty-IPW I Propensity-IPW, Prop r ensity n t -IIPW PW, Ad A Adjusted djust u ed d ffor o IITB or TB B Propensity-matched, Prop Pr open op en nsi sity ty-m ty -mat atch tch ched ed, Ad ed Adju Adjusted justed ted ffor o IITB or TB Propensity-IPW, Prop Pr open op ensi en sity si ty-I ty -IIPW PW, Ad Adju Adjusted just ju sted st ed ffor or IITB TB Propensity-IPW Propensity IPW Propensity-IPW 1.63 1.10 0.85 0. 85 1.15 00.68 0. 68 0.61 0. 61 00.73 .73 73 11.04 04 2.35 (0.70-3.78) (0.76-1.60) (0.35-2.10) (0 .3355-2. 2.10 10) 0) (0.57-2.32) (0 0.57-2.32 3 ) (0.52-0.88) (0 0.52-0.88 8 ) (0.39-0.95) (0 .39.3 399-0. 0 95 0. 95)) (0.59-0.90) (0 .59599 0.90 90)) 90 (0.67-1.61) (0 67 1 61) (1.56-3.52) 0.25 0.61 0.73 0.70 0.003 0.027 0.003 0.88 <0.001 All hazard ratios reflect VCF use vs. No use. VCF = Vena cava filter IPW = Inverse probability weighted ITB = Immortal time bias 23 DOI: 10.1161/CIRCULATIONAHA.115.020338 Figure Legends: Figure 1. Study Flow Diagram. Figure 2. Forest Plot of Multivariable Model to Predict Death < 30 days in Patient with No Contraindication to Anticoagulation. Figure 3. Forest Plot of Multivariable Model to Predict Death < 30 days in Patients undergoing Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 Major Surgery. Figure 4. Forest Plot of Multivariable Model to Predict Death < 30 days in Patientss with witth Active Acti tive Bleeding. Bleeding g. Figure Figu Fi ure 5. Stan Standardized andaard an dizeed Mea Mean an Di D Differences fferren ncees be between etw ween Ca Cases ase sess wi w with th a V Vena ena Ca C Cava avaa F Filter ilter ve versu versus us Casess without a Vena Cava Filter who had No Contraindication to Anticoagulation. Anticoagu g lation. Figure 6. Standardized Mean Differences between Cases with a Vena Cava Filter versus Cases without a Vena Cave Filter who underwent Major Surgery. Figure 7. Standardized Mean Differences between Cases with a Vena Cava Filter versus Cases without a Vena Cave Filter who had Active Bleeding. Figure 8. Hospital Day of Death among Patients with Active Bleeding who Received or did not Receive a Vena Cava Filter. 24 Acute VTE Cases N=104,677 First admission with a principal diagnosis of pulmonary embolism or venous thromboembolism in the study period (POA Y): 2005 through 2010 Exclude cases with an active cancer: N=18,735 Exclude cases from Hospitals < 55 Acute VTE opportunities: N=783 Acute-VTE, No Cancer: N= 85,159 Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 No Contraindication Major Operating Room Procedure (+/- Bleeding*) N=80,697 N=1,445 1) NO operation in-hosp AND 2) No recent operation < 3 days prior to admit AND 2)) No bleeding, bleed din ing, either POA A = Y or N 1) Major operation during hospital stay OR 2) Recent operation < 3 days prior to admit Bleeding Coded POA = Y or N (without surgery) N=3,017 1) Bleed POA = Yes OR 2) Bleed POA = No *Bleed POA OA Y or N included if surgery VCFF VC N V No VCF CF VCFF VC N VCF No VCF VCFF VC N V No VCF CF N=7,762 N=72,935 N=489 N=956 N=1,095 N=1,922 VTE = Venous thromboembolism VCF = Vena Cava Filter 1 Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 No Filter 9% Filter 6% 3% Hospital Day Patient Died 29 25 21 17 13 9 5 1 Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 0% Percent Deaths Occuring within 30-days of Admission 12% Outcomes After Vena Cava Filter Use in Non-Cancer Patients with Acute Venous Thromboembolism: A Population-Based Study Richard H. White, Ann Brunson, Patrick S. Romano, Zhongmin Li and Ted Wun Circulation. published online April 5, 2016; Downloaded from http://circ.ahajournals.org/ by guest on October 1, 2016 Circulation is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231 Copyright © 2016 American Heart Association, Inc. All rights reserved. Print ISSN: 0009-7322. Online ISSN: 1524-4539 The online version of this article, along with updated information and services, is located on the World Wide Web at: http://circ.ahajournals.org/content/early/2016/04/05/CIRCULATIONAHA.115.020338 Data Supplement (unedited) at: http://circ.ahajournals.org/content/suppl/2016/04/05/CIRCULATIONAHA.115.020338.DC1.html Permissions: Requests for permissions to reproduce figures, tables, or portions of articles originally published in Circulation can be obtained via RightsLink, a service of the Copyright Clearance Center, not the Editorial Office. Once the online version of the published article for which permission is being requested is located, click Request Permissions in the middle column of the Web page under Services. Further information about this process is available in the Permissions and Rights Question and Answer document. Reprints: Information about reprints can be found online at: http://www.lww.com/reprints Subscriptions: Information about subscribing to Circulation is online at: http://circ.ahajournals.org//subscriptions/ SUPPLEMENTAL MATERIAL Supplemental Table 1. ICD-9-CM Codes to Identify Cases with PE or Lower Extremity DVT. CODE 673.2 415.11 415.13 415.19 VTE Type PE PE PE PE DESCRIPTION Obstetrical blood clot embolism (PE) (673.2) Iatrogenic pulmonary embolism and infarction (415.11) Saddle pulmonary embolus Other pulmonary embolism (415.19) 453.2 453.8 451.11 451.19 451.81 453.40 453.41 453.42 671.31 671.33 671.42 671.44 LE-DVT LE-DVT LE-DVT LE-DVT LE-DVT LE-DVT LE-DVT-PROXIMAL LE-DVT-DISTAL LE-DVT LE-DVT LE-DVT LE-DVT Venous thrombosis of inferior vena cava (453.2) (PRE Oct 1, 2009) eliminated Oct 1, 2009 Thrombophlebitis, deep vessels of lower extremities femoral vein (451.11) Thrombophlebitis of deep vessel of lower extremities, other (451.19) Thrombophlebitis of other sites iliac vein (451.81) Acute venous thrombosis of unspecified deep vessels of lower extremity (453.40) Acute venous thrombosis of deep vessels of proximal lower extremity (453.41) Acute venous thrombosis of deep vessels of distal lower extremity (453.42) DVT antepartum-Delivered (671.31) DVT-antepartum (671.33) Thrombosis postpartum-del W P/P (671.42) DVT-postpartum (671.44) Supplemental Table 2. ICD-9-CM Codes to Identify Cases with Acute Bleeding. Outcome Intracranial hemorrhage ICD-9-CM Codes 430- subarachnoid 431-intracerebral 432.0-other ICH 432.1-Subdural 432.9-Unspecified Note: all 5th digits 0-9 are searched for using the following traumatic ICH codes: 852.0-SAH after injury 852.2-subdural after injury 852.4-extradural after injury 853.0- other after injury Excluded Codes All codes for ICH with open intracranial wound (i.e., 852.1 852.3, 852.5, 853.1) Outcome GI hemorrhage Codes Included 455.2-internal hemorrhoids 455.5-external hemorrhoids 455.8- unspecified hemorrhoids 456.0-varices with bleeding 456.20-varices with bleeding 459.0-hemorrhage unspecified 530.7-laceration hemorrhage 530.82 esophogeal hemorrhage 531.00-.01-gastric ulcer bleed +/- obstruct 531.20-.21- gastric ulcer bleed+perf+/- obstr 531.40-.41-chronic gastric bleed 531.60-.61-chronic gastric bleed+perf 532.00-.01-duodenal ulcer bleed+/-perf 532.20-.21- duodenal ulcer bleed+perf+/- obstr 532.40-.41- chronic duodenal+ bleed 532.60-.61- chronic duodenal bleed+perf 533.00-.01- peptic ulcer bleed +/- obstruct 533.20-.21- peptic ulcer bleed+perf+/- obstr 533.40-.41- chronic peptic+ bleed 533.60-.61- chronic peptic bleed+perf 534.00-.01-gastrojejunal bleed 534.20-.21- gastrojejunal bleed + perf 534.40-.41- chronic gastrojejunal bleed 534.60-.61- chronic gastrojejunal + perf 535.01- acute gastritis with bleed 535.11-atrophic gastritis with bleed 535.21-gastric mucosal hypertrophy + bleed 535.31-alcoholic gastritis+bleed 535.41-other gastritis +bleed 535.51-unspecified gastritis with bleed Excluded Codes All ulcer codes without mention of hemorrhage (e.g., 534.31, 534.51) 535.61-duodenitis with bleed 537.83 angiodysplasia with bleed, stomach 562.02 small bowel diverticulosis with bleed 562.03-small bowel diverticulitis with bleed 562.12 diverticulosis of colon with bleed 562.13 diverticulitis of colon with bleed 568.81-hemoperitoneum 569.3-hemorrhage of anus and rectum 569.85-angiodyplasia of intestine with bleed 578.0-hematemesis 578.1-blood in stool 578.9-GI hemorrhage, not specified Outcome Other bleed Codes Included 423.0 (hemopericardium) 593.81 (vascular disorders of kidney) 599.7 (hematuria) 719.1x (including fifth digits: 0-9) 784.7 (epistaxis) 784.8 (hemorrhage from throat) (added 12/4/00) 786.3 (hemoptysis) Excluded Codes 623.8 (excessive vaginal bleeding) 626.2 (excessive vaginal bleeding 626.6 (menorrhagia) Supplemental Table 3. ICD-9-CM Codes to Identify Vascular Interventions the Might Reflect Procedures done to Remove or Thrombolyse a Lower Extremity Deep Vein Thrombosis. INCLUDED ICD-9-CM CODE 00.66 38.08 38.09 38.18 39.50 ICD-9-CM CODE DESCRIPTION PTCA OR CORONARY ATHER (Begin 2005) EMBOLECTOMY LEG VESSEL LOWER LIMB VEIN INCISION ENDARTERECTOMY LEG VESL ANGIOPLAST/ATHERECT (Begin 1995) EXCLUDED ICD-9-CM CODE 38.03 38.07 39.27 39.56 39.57 39.58 39.59 39.99 54.0 ICD-9-CM CODE DESCRIPTION UPPER LIMB VESSEL INCIS ABDOMINAL VEIN INCISION DIALYSIS ARTERIOVENOSTOM REPAIR VESS W TIS PATCH REP VESS W SYNTH PATCH REPAIR VESS W PATCH NOS REPAIR OF VESSEL NEC VESSEL OPERATION NEC (filer removal) ABDOMINAL WALL INCISION Use of aThromobolytic AGENT ICD-9-CM =99.10