Thrombophilia remains an important cause of morbidity and mortality in patients afflicted with brain tumors.1–7 Specifically, the incidence of deep vein thrombosis (DVT) or venous thromboembolism (VTE) has been noted to be as great as 72%1 or as small as 4.9%,6 depending on the anticoagulation regimen or clinical situation (e.g., postoperative versus ambulatory setting); however, as recently reviewed, an incidence of 20% to 30% of thrombophilia is generally observed.2–4 Primary brain tumors express tissue factor (TF) >80% of the time,7 and TF-containing microparticles (MP) shed from brain tumors are strongly procoagulant8–10 and are found in increased concentrations in vivo in brain tumor patients.9,10 Also, potentially contributing to the risk of thrombophilia, patients with brain tumors can also have increased circulating fibrinogen concentrations.11 While an increased prothrombin time, increased plasminogen activity, and decreased fibrinogen concentration correlated post hoc with occurrence of postoperative thrombosis in brain tumor patients,11 neither MP10 nor fibrinogen11 concentrations prospectively correlate with postoperative DVT or VTE. At best, increased MP concentrations demonstrated a 1.38 relative risk of postoperative VTE in brain cancer patients.9 Taken as a whole, important sources of increased intravascular thrombin generation (e.g., MP) and abnormally increased fibrinogen concentration have been identified as contributing etiologies of the frequent incidence of thrombophilia in the setting of brain tumors.
Of interest, brain tumors may either primarily produce or cause surrounding brain tissue to produce endogenous carbon monoxide (CO) via induction of hemeoxygenase-1 (HO-1) activity.12–15 HO-1 is solely responsible for endogenous production of CO during the catabolism of heme16 and is upregulated by either endogenous or exogenous nitric oxide (NO) exposure.12–14 The importance of CO in modulating coagulation has recently been demonstrated by increases in plasmatic coagulation kinetics by CO via interaction with a heme group(s) attached to fibrinogen in humans and rabbits in vitro and in vivo.17–21 CO also attenuates fibrinolysis via heme-based mechanisms.22–24 Thus, brain tumor-associated upregulation of HO-1 and CO production could potentially contribute to plasmatic hypercoagulation. In diagrammatic form, the key concepts we propose that are responsible for plasmatic hypercoagulability in the setting of brain tumors are depicted in Figure 1.
The goals of the present investigation were to determine whether patients with brain tumors had increased HO-1 upregulation/CO production, plasmatic hypercoagulability, and formation of carboxyhemefibrinogen (COHF). These goals were achieved by determination of carboxyhemoglobin (COHb), thrombelastographic analyses of patient plasma,17,20,22 and determination of the formation of COHF as previously described.21
Normal subject citrated plasma (George King Bio-Medical, Overland Park, KS) anticoagulated with sodium citrate was used for experimentation. A standard lot of 30 individuals (15 men, 15 women; mean age 34 years, with range of 19 years to 56 years) was used. All normal subjects were verified to be without blood-borne disease (e.g., hepatitis), not pregnant, and nonsmokers as per the vendor’s specifications. The rationale for using plasma of this nature is that standard, plasma-based tests of coagulation (e.g., prothrombin time, activated prothrombin time, fibrinogen concentration, and coagulation factor activities) commonly have 95% confidence interval values in clinical pathology laboratories within hospitals/ambulatory clinics established with such material. Indeed, that is the express commercial purpose of the manufacturing of such normal subject plasma lots because most facilities do not have the resources or time to insure that their normal reference population is actually disease/medication-free. Furthermore, by using the same lot of plasma for reference between hospitals, a greater standardization of “normality” is realized so that comparison of patient population sample values across institutions is more reasonable. As we plan/anticipate such multiinstitutional comparisons with the subsequently described thrombelastograph®-based assays, we thought it prudent to use standardized, normal subject plasma obtained from an internationally recognized commercial vendor.
Brain Tumor Patients
Our protocol (#12-0179-04) was approved by the University of Arizona IRB. Patients with brain tumors (N = 20) scheduled to undergo craniotomy, aged 18 to 80 years who were nonsmokers, were recruited for this investigation. The patients had no history of inherited bleeding disorder and were not being administered anticoagulant medications. After written consent was obtained, and after induction of anesthesia, but before operation, whole blood (15 mL) was obtained from the patients’ indwelling arterial catheters after removing crystalloid solution and 3 to 4 mL potentially diluted blood from the narrow-bore pressure tubing at the wrist. The blood sample was anticoagulated with sodium citrate (9 parts blood to 1 part 0.105 M sodium citrate) and subsequently centrifuged at 3000g for 15 minutes at room temperature, with plasma decanted, aliquoted, and stored at −80°C before experimentation. The COHb concentration was recorded from initial arterial blood gas analyses performed with our clinical laboratory instrument (GEM Premier 4000, Instrumentation Laboratory, Bedford, MA).
Coagulation Kinetics and COHF Assays
Plasma was rapidly thawed at 37ºC on the day of experimentation. The final volume for all subsequently described plasma sample mixtures was 359.4 μL. Sample composition consisted of 326 μL plasma; 10 μL TF reagent (0.1% final concentration in dH2O; Diagnostica Stago S.A.S., Asnieres sur Seine, France), 3.6 μL dH2O or CORM-2 (100 μM final concentration; Sigma-Aldrich, Saint Louis, MO,) or phenylhydroxylamine (PHA) (30 mM final concentration; Sigma-Aldrich) and 20 μL of 200 mM CaCl2 as per our previously described COHF assay.21 This concentration of CORM-2 reliably increases clot strength in the absence of CO, and the concentration of PHA used reliably converted heme groups associated with fibrinogen to a metheme state, abrogating the effects of CO already present.21 Plasma sample mixtures were placed in a disposable cup in a computer-controlled thrombelastograph® hemostasis system (Model 5000, Haemoscope Corp., Niles, IL), with addition of CaCl2 as the last step to initiate clotting. Data were collected at 37°C for 15 minutes. Elastic modulus-based parameters previously described were determined.17–21
Using the COHF assay methodology, we subsequently described hypercoagulability as an elastic modulus (G, dynes/cm2) value > 95% confidence interval value of the normal subjects’ dataset. Modifying our previous definition of COHF formation,20,21 we defined the presence of COHF as a sum of the % increase in G secondary to CORM-2 exposure and the % decrease in G secondary to PHA exposure that was significantly less than the average value of similar measurements in normal subject samples. This sum henceforth will be denoted as the fibrinogen redox gap (FRG), recognizing that changes in endogenous NO may affect the clot strength response to exogenous CO or PHA. This approach was taken after considering the variability in clot strength response in the presence of excess NO.25,26
Fibrinolytic Vulnerability Assay
Additional plasma samples from normal subjects and brain tumor patients were rapidly thawed at 37ºC on the day of experimentation and submitted to thrombelastographic analyses as previously described.27–29 The final volume for all subsequently described plasma sample mixtures was 360 μL. Sample composition consisted of 310 μL plasma; 10 μL TF/celite reagent (0.1% final concentration of TF in dH2O with 1% celite obtained from Sigma-Aldrich, Saint Louis, MO,); 10 μL tissue type plasminogen activator (tPA, 580 U/mg; Genentech, Inc., San Francisco, CA) diluted with 50 mmol/L potassium phosphate buffer (pH 7.4) for a final activity of 100 U/mL; 10 μL dH2O; and 20 μL of 200 mM CaCl2 as per our previously described assay.28,29 The rationale for strong thrombin generation (with TF and contact protein activator celite) in this assay is to strongly engage antifibrinolytic enzymes that are thrombin dependent (e.g., thrombin activatable fibrinolysis inhibitor, Factor XIII) and those that are not thrombin dependent (e.g., α2-antiplasmin). Data were collected at 37°C until clot lysis time (reduction of clot strength to 2 mm amplitude) was observed. Elastic modulus-based parameters previously described were determined.22–24,27–29
Statistical Analyses and Graphics
Data are presented as mean ± SD, with analyses conducted with a commercially available statistical program (Microsoft Excel 2007, Microsoft Corporation, Redmond, WA). All comparisons of normal subject and brain tumor patient coagulation kinetic assay values and fibrinolytic assay values were conducted with Student t tests. Graphics were generated with a commercially available program (Origin 7.5; OriginLab Corp., Northampton, MA; CorelDRAW12, Corel Corporation, Mountain View, CA). A P value of <0.05 was considered significant.
Twenty brain tumor patients were recruited because the statistical power to detect plasmatic hypercoagulability/COHF formation in a setting involving increased CO exposure was achieved with 20 subjects based on previous work.20 Specifically, this number of patients would be required to perform 2-tailed Student t test or analyses of variance statistical analyses. This previous investigation used similar biochemical/thrombelastographic analyses, and while not documented in the manuscript, the statistical power exceeded 0.9 with all analyses with P < 0.05 for all parameters tested.20
Twenty patients who underwent craniotomy for biopsy or resection of brain tumor were recruited into the study. The average age of the brain tumor patients was 58 ± 15 years, and the man:woman ratio was 11:9. The average COHb concentration measured in brain tumor patients was 1.5% ± 0.5%. While the normal range for COHb is 0% to 1.5% in our laboratory, values of 0.91% and above are significantly associated with increased cardiovascular events and death in never smokers.30 Sixteen of 19 COHb measurements among the 20 patients were >0.91%, with 10 COHb values >1.5%.
The plasma coagulation kinetic comparisons between the normal subject and brain tumor patients are displayed in Figure 2. Patients with brain tumors had thrombi that reached the maximum rate of formation significantly more quickly than normal subjects. Furthermore, brain tumor patients had maximum velocities of clot growth and final clot strengths significantly greater than normal subjects.
With regard to elastic modulus, as displayed in Table 1, patients with brain tumors had clot strengths that were significantly greater than normal subjects under control, carboxyheme, or metheme states. The 95% confidence interval value for clot strength in normal subjects was 2092 dynes/cm2; the normal subject FRG (based on individual, not group mean responses) was 112%. Thus, for a brain tumor patient to be considered hypercoagulable, the elastic modulus value had to exceed 2092 dynes/cm2; similarly, for COHF to be considered present, the FRG value had to be <112%.
Individual clinical characteristics and coagulation status data for the brain tumor cohort are displayed in Table 2. Ten patients had elastic modulus values considered hypercoagulable. Twelve patients had FRG values consistent with COHF formation. Furthermore, 5 patients were hypercoagulable and had COHF present. When patients had COHb values stratified by FRG, 5 patients without COHF formation had COHb values >0.91%, whereas 11 patients with COHF formation had COHb values >0.91%. With regard to other clinical data, no particular age or tumor type appeared to contribute to the incidence of hypercoagulability or COHF formation. Of the 10 brain tumor patients who were hypercoagulable, 5 had primary brain tumors, 4 had metastatic tumors, and 1 had an inflammatory process with neoplastic features.
Comparisons of fibrinolytic vulnerability between normal subject and brain tumor patient plasma samples are displayed in Table 3. Brain tumor patients had plasma thrombi that reached the maximum rate of formation significantly more quickly, had maximum velocities of clot growth, and had final clot strengths significantly greater than normal subjects. There were no significant differences in the time to onset of maximum rate of lysis, maximum rate of lysis, or clot lysis time values between normal subjects and brain tumor patients.
The primary findings of the present investigation were that brain tumors can be associated with upregulation of HO-1 activity as demonstrated by COHb concentrations associated with increased cardiovascular risk30 and/or greater than the normal range; can be associated with a hypercoagulable state evidenced by stronger than normal plasma thrombi; and can be associated with COHF formation. When integrated into the preexisting paradigm of plasmatic hypercoagulability associated with brain tumors involving increased TF-bearing MP and hyperfibrinogenemia,7–11 our findings of HO-1 upregulation and COHF formation may be a missing piece of the teleological puzzle of DVT and VTE in this patient population. Last, there appears to be no important brain tumor-associated decrease in fibrinolytic vulnerability; thus, CO-mediated inhibition of fibrinolysis, as seen in other systems,23,24 is unlikely to play an important clinical role in brain tumor-associated thrombophilia.
While encouraging, our results must be viewed as preliminary. Our investigation was intended as a “proof-of-concept,” and use of these analytical methods to correlate plasmatic hypercoagulability/COHF formation with clinical thrombotic events in longitudinal studies (e.g., surgical settings, chemotherapy/radiation therapy) involving greater numbers of patients is needed. It must be emphasized that only 5 of the brain tumor patients had hypercoagulability and COHF formation based on in vitro analyses; the in vivo correlation of hypercoagulability and COHF formation in the development of clinical thrombophilia remains to be determined. Then, if plasmatic hypercoagulability/COHF formation appears to be an important risk factor for pathological clotting, novel therapeutic anticoagulant strategies (e.g., HO-1 inhibition, interference with fibrin polymerization) could be rigorously tested to diminish brain tumor-related thrombotic morbidity and mortality.
Another potential limitation of our investigation is the choice of control plasma with which to compare plasma obtained from our brain tumor patients. The normal subjects had blood obtained via venous access, were not anesthetized, and did not have COHb determination from arterial blood gas analyses. Furthermore, the normal subjects were not precisely age-matched. In an effort to compensate for this situation in the case of COHb values, we included not just our laboratory normal range (determined periodically inhouse with volunteers assumed to not smoke) but also included a value determined from the study of thousands of individuals who had not smoked that was associated with increased cardiovascular events.30 We would posit that while these factors are of potential importance, they are most likely outweighed by the variability introduced by individual investigators attempting to recruit normal subjects without the same rigorous screening used by commercial vendors that is determined by governmental oversight. Also, as mentioned previously in the Methods section, the use of the same lot of 30 to 50 normal donors among different centers eliminates almost certain screening variablity of normals between laboratories. Thus, while our choice of control plasma could potentially influence our conclusions, we made the choice to minimize the major confounding variables that could affect our experimental system (e.g., smoking, liver disease, etc.).
Other important observations made in this investigation include the effects of different types of brain tumors (or inflammation) on both HO-1 upregulation and coagulation. As seen in Table 2, noncancerous processes, primary brain tumors or metastatic tumors all increased COHb and induced hypercoagulability. While not definitive, these data strongly support the concept that inflammation and/or ischemia of brain tissue surrounding the tumor/inflammatory lesion may play a role in endogenous CO production and hypercoagulability. Alternatively, it may also be possible that inflammed brain may release a substance(s) that upregulate HO-1 activity systemically. Persistent (or slowly resolving) brain inflammation may in part explain the prolonged period (weeks to months) of thrombophilia observed after resection of brain tumors.2,3
It is of interest that not all brain tumor patients with pathological30 COHb values were hypercoagulable (N = 4) or had COHF formation (N = 5); furthermore, there was 1 patient who was hypercoagulable and had COHF formation in the presence of a COHb value of 0.8%, while another patient was hypercoagulable without COHF formation with a COHb value of 0.3%. Also of interest, 5 patients in the hypercoagulable group had primary brain tumors, whereas the other 5 patients had metastatic/inflammatory lesions, suggestive that not only the tumor but also the impact of the tumor’s presence on surrounding brain may be hemostatically significant. The kinetics of CO binding on heme-bearing fibrinogen have not been fully characterized, and the increase in elastic modulus in the normal subject group to a fixed exposure of CO (100 μM CORM-2) resulted in increases in clot strength from as small as 42% to as large as 144%. Similarly, healthy smokers with average COHb values of 5% were only hypercoagulable 45% of the time, and 65% had COHF formation that was detectable with our thrombelastographic assay.20 It is unknown how much endogenous CO exposure circulating fibrinogen undergoes for any given COHb value observed, because the formation of COHb is indicative of upregulation of HO-1 activity in the setting of brain tumors and does not predict precise CO concentrations at the tumor blood-endothelial interface. In sum, patients with brain tumors are expected to have variable engagement of HO-1 and consequent CO formation, variable CO-independent and CO-dependent hypercoagulability, and variable COHF formation, which is consistent with clinical observations that patients with brain tumors are at increased risk of DVT/VTE but do not as a population uniformly suffer thrombophilia.
In conclusion, our investigation revealed that a preliminary sample of 20 patients undergoing craniotomy for brain tumor biopsy or removal had plasma that clotted significantly more quickly and was significantly stronger than normal subject plasma. Furthermore, 10 of these brain tumor patients had plasma clot strength that exceeded the 95% confidence interval value observed in normal subjects, and 5 patients in this particularly hypercoagulable subgroup had COHF formation. Importantly, we also documented that brain tumor presence engages the HO-1 system, with increased CO production resulting as evidenced by increased circulating COHb concentrations. Future investigations of the role played by HO-1 in the pathogenesis of brain tumor-related DVT and VTE are warranted.
Name: Vance G. Nielsen, MD.
Contribution: This author helped design and conduct this study, collected the data, and analyzed the data. He was primarily responsible for the preparation of the manuscript.
Attestation: Vance G. Nielsen approved the final manuscript. Vance G. Nielsen attests to the integrity of the original data and the analysis and is the archival author.
Name: G. Michael Lemole Jr., MD.
Contribution: This author helped design the study, collect the data, and prepare the manuscript.
Attestation: G. Michael Lemole Jr. approved the final manuscript.
Name: Ryan W. Matika, MD.
Contribution: This author helped design the study, collect and analyze the data, and prepare the manuscript.
Attestation: Ryan W. Matika approved the final manuscript. Ryan W. Matika attests to the integrity of the original data and the analysis.
Name: Martin E. Weinand, MD.
Contribution: This author helped design the study, collect the data, and prepare the manuscript.
Attestation: Martin E. Weinand approved the final manuscript.
Name: Sana Hussaini, MD.
Contribution: This author helped collect data and prepare the manuscript.
Attestation: Sana Hussaini approved the final manuscript.
Name: Ali A. Baaj, MD.
Contribution: This author helped collect data and prepare the manuscript.
Attestation: Ali A. Baaj approved the final manuscript.
Name: Evangelina B. Steinbrenner, CCRC.
Contribution: This author helped design and conduct the study and collect the data.
Attestation: Evangelina B. Steinbrenner approved the final manuscript.
This manuscript was handled by: Jerrold H. Levy, MD, FAHA, FCCM.
1. Sawaya R, Zuccarello M, Elkalliny M, Nishiyama H. Postoperative venous thromboembolism and brain tumors: Part I. Clinical profile. J Neurooncol. 1992;14:119–25
2. Semrad TJ, O’Donnell R, Wun T, Chew H, Harvey D, Zhou H, White RH. Epidemiology of venous thromboembolism in 9489 patients with malignant glioma. J Neurosurg. 2007;106:601–8
3. Pan E, Tsai JS, Mitchell SB. Retrospective study of venous thromboembolic and intracerebral hemorrhagic events in glioblastoma patients. Anticancer Res. 2009;29:4309–13
4. Jenkins EO, Schiff D, Mackman N, Key NS. Venous thromboembolism in malignant gliomas. J Thromb Haemost. 2010;8:221–7
5. Aishima K, Yoshimoto. Screening strategy using sequential serum D-dimer assay for the detection and prevention of venous thromboembolism after elective brain tumor surgery. Br J Neurosurg. 2013;27:348–54
6. Gerber DE, Segal JB, Salhotra A, Olivi A, Grossman SA, Streiff MB. Venous thromboembolism occurs infrequently in meningioma patients receiving combined modality prophylaxis. Cancer. 2007;109:300–5
7. Thaler J, Preusser M, Ay C, Kaider A, Marosi C, Zielinski C, Pabinger I, Hainfellner JA. Intratumoral tissue factor expression and risk of venous thromboembolism in brain tumor patients. Thromb Res. 2013;131:162–5
8. Bastida E, Ordinas A, Escolar G, Jamieson GA. Tissue factor in microvesicles shed from U87MG human glioblastoma cells induces coagulation, platelet aggregation, and thrombogenesis. Blood. 1984;64:177–84
9. Sartori MT, Della Puppa A, Ballin A, Saggiorato G, Bernardi D, Padoan A, Scienza R, d’Avella D, Cella G. Prothrombotic state in glioblastoma multiforme: an evaluation of the procoagulant activity of circulating microparticles. J Neurooncol. 2011;104:225–31
10. Thaler J, Ay C, Mackman N, Bertina RM, Kaider A, Marosi C, Key NS, Barcel DA, Scheithauer W, Kornek G, Zielinski C, Pabinger I. Microparticle-associated tissue factor activity, venous thromboembolism and mortality in pancreatic, gastric, colorectal and brain cancer patients. J Thromb Haemost. 2012;10:1363–70
11. Sawaya R, Glas-Greenwalt P. Postoperative venous thromboembolism and brain tumors: Part II. Hemostatic profile. J Neurooncol. 1992;14:127–34
12. Takahashi K, Hara E, Suzuki H, Sasano H, Shibahara S. Expression of heme oxygenase isozyme mRNAs in the human brain and induction of heme oxygenase-1 by nitric oxide donors. J Neurochem. 1996;67:482–9
13. Broholm H, Rubin I, Kruse A, Braendstrup O, Schmidt K, Skriver EB, Lauritzen M. Nitric oxide synthase expression and enzymatic activity in human brain tumors. Clin Neuropathol. 2003;22:273–81
14. Hara E, Takahashi K, Tominaga T, Kumabe T, Kayama T, Suzuki H, Fujita H, Yoshimoto T, Shirato K, Shibahara S. Expression of heme oxygenase and inducible nitric oxide synthase mRNA in human brain tumors. Biochem Biophys Res Commun. 1996;224:153–8
15. Deininger MH, Meyermann R, Trautmann K, Duffner F, Grote EH, Wickboldt J, Schluesener HJ. Heme oxygenase (HO)-1 expressing macrophages/microglial cells accumulate during oligodendroglioma progression. Brain Res. 2000;882:1–8
16. Landaw SA, Callahan EW Jr, Schmid R. Catabolism of heme in vivo: comparison of the simultaneous production of bilirubin and carbon monoxide. J Clin Invest. 1970;49:914–25
17. Nielsen VG, Kirklin JK, George JF. Carbon monoxide releasing molecule-2 increases the velocity of thrombus growth and strength in human plasma. Blood Coagul Fibrinolysis. 2009;20:377–80
18. Nielsen VG, Cohen JB, Malayaman SN, Nowak M, Vosseller K. Fibrinogen is a heme-associated, carbon monoxide sensing molecule: a preliminary report. Blood Coagul Fibrinolysis. 2011;22:443–7
19. Nielsen VG, Chawla N, Mangla D, Gomes SB, Arkebauer MR, Wasko KA, Sadacharam K, Vosseller K. Carbon monoxide-releasing molecule-2 enhances coagulation in rabbit plasma and decreases bleeding time in clopidogrel/aspirin-treated rabbits. Blood Coagul Fibrinolysis. 2011;22:756–9
20. Nielsen VG, Hafner DT, Steinbrenner EB. Tobacco smoke-induced hypercoagulation in human plasma: role of carbon monoxide. Blood Coagul Fibrinolysis. 2013;24:405–10
21. Nielsen VG, Arkebauer MR, Vosseller K. Redox-based thrombelastographic method to detect carboxyhemefibrinogen-mediated hypercoagulability. Blood Coagul Fibrinolysis. 2011;22:657–61
22. Nielsen VG, Kirklin JK, George JF. Carbon monoxide-releasing molecule-2 decreases fibrinolysis in human plasma. Blood Coagul Fibrinolysis. 2009;20:448–55
23. Arkebauer MR, Kanaparthy SS, Malayaman SN, Vosseller K, Nielsen VG. Carbon monoxide and nitric oxide modulate α2
-antiplasmin and plasmin activity: role of heme. Blood Coagul Fibrinolysis. 2011;22:712–9
24. Nielsen VG, Arkebauer MR, Wasko KA, Malayaman SN, Vosseller K. Carbon monoxide-releasing molecule-2 decreases fibrinolysis in vitro
and in vivo
in the rabbit. Blood Coagul Fibrinolysis. 2012;23:104–7
25. Nielsen VG. Nitric oxide decreases coagulation protein function in rabbits as assessed by thromboelastography. Anesth Analg. 2001;92:320–3
26. Olver CS, Nielsen VG. Thrombelastographic characterization of coagulation/fibrinolysis in horses: role of carboxyheme and metheme states. Blood Coagul Fibrinolysis. 2013;24:273–8
27. Nielsen VG, Cohen BM, Cohen E. Elastic modulus-based thrombelastographic quantification of plasma clot fibrinolysis with progressive plasminogen activation. Blood Coagul Fibrinolysis. 2006;17:75–81
28. Nielsen VG, Ellis TC. Thrombelastographic quantification of the contributions of thrombin activatable fibrinolysis inhibitor and alpha2-antiplasmin to antifibrinolytic activity in human plasma. Blood Coagul Fibrinolysis. 2007;18:29–33
29. Nielsen VG. Clot life span model analysis of clot growth and fibrinolysis in normal subjects: role of thrombin activatable fibrinolysis inhibitor. Blood Coagul Fibrinolysis. 2008;19:283–7
30. Hedblad B, Engström G, Janzon E, Berglund G, Janzon L. COHb% as a marker of cardiovascular risk in never smokers: results from a population-based cohort study. Scand J Public Health. 2006;34:609–15