D-Dimer and Anticoagulation in Patients With Mechanical Prosthetic Heart Valves
A 2-Year Follow-up
Abstract The best anticoagulation level in patients with mechanical heart valve prostheses is still being debated. D-dimer, which detects the presence of cross-linked fibrin degradation products, has been demonstrated to be a useful marker of coagulation activation. This study was designed to verify whether heart valve prostheses in anticoagulated patients are associated with abnormalities in D-dimer plasma levels, and if so, whether such levels are related to the anticoagulation level and/or whether they could be predictive of acute vascular or hemorrhagic events. In 132 patients with single and 10 with double mechanical valve replacement, international normalized ratio (INR) and D-dimer plasma levels were determined. The INR levels of the previous 8 months were reviewed to assess the time that each patient spent in the therapeutic range. The D-dimer plasma levels were compared with those obtained from 102 matched control subjects. The patients were then followed up for 2 years to record acute vascular and hemorrhagic events. For the entire group, D-dimer plasma levels in patients were the same as those in the control group. Patients with double valve replacement had higher D-dimer plasma levels than either monovalvular implant patients or control subjects. Patients who had spent <75% of the time within the assigned anticoagulation range had higher values for D-dimer plasma levels (median, 270 vs 198 ng/mL, P=.02). The major determinants of D-dimer plasma levels were age (R2=.07, P=.009) and the percentage of time spent below the predetermined INR level (R2=.09, P=.001). During follow-up, 19 acute vascular and 16 hemorrhagic events occurred. High D-dimer tertile was the only parameter predicting the occurrence of thromboembolic events. In patients with mechanical heart valve prostheses, the D-dimer plasma level depended on the thoroughness of anticoagulation. Patients in the upper tertile of D-dimer values have an ≈5-fold risk of vascular thromboembolic events. D-dimer determination can therefore be useful in detecting patients who are at a higher risk of severe vascular events.
Reprint requests to Dr Carlo Giansante, Istituto di Clinica Medica Generale e Terapia Medica, Strada di Fiume 447, I-34149 Trieste, Italy.
- Received January 17, 1996.
- Accepted July 31, 1996.
Thromboembolism is a severe and common complication in patients who have undergone artificial heart valve implantation.1 2 Anticoagulation is indicated for these patients because it prevents this complication, although it may cause some side effects, mostly bleeding events.1 Strict monitoring of anticoagulation therapy can decrease the incidence of adverse hemorrhagic events but reduce that of thromboembolism by only 50%,3 so that thromboembolic events remain the major cause of morbidity and mortality in patients with mechanical heart valve prostheses.1 In fact, these events still occur at the rate of 2 to 5 events per 100 patient-years.1 2 4 5 6 Several factors may contribute to this failure, such as advancing underlying heart disease,6 but it probably reflects the complex control of the level of anticoagulation.3 Many studies have been undertaken to assess the correct level of oral anticoagulation, but the optimal level of anticoagulation for those with artificial heart valves still remains controversial.4 5 6 In addition, adjustment of the dose of oral anticoagulants presents several difficulties, so that a predetermined level of anticoagulation at all times is still difficult to achieve.3 Cannegieter et al7 have recently demonstrated that thrombotic and hemorrhagic events occur in patients whose INR in the preceding week was outside the range of 3 to 4. Although this study supports the importance of strict management of anticoagulation, it cannot predict the occurrence of acute vascular events more than 1 week in advance. Hence, laboratory tests would be helpful to identify patients at high risk of such events.
The plasma D-dimer level reflects fibrin turnover as well as the efficacy of fibrinolysis8 without being influenced by fibrinogen or non–cross-linked fibrin.7 8 9 High plasma levels of these cross-linked fibrin degradation products have been demonstrated in several pathological conditions as being associated with a high risk of thromboembolism, such as in patients with atrial fibrillation,10 11 those undergoing cardiopulmonary bypass surgery,8 or those who have received the Jarvik 7-70 total artificial heart.12 Furthermore, high plasma D-dimer levels have been detected in patients with acute ischemic stroke13 and peripheral vascular disease14 and have proved to be useful in the diagnosis of deep-vein thrombosis.15 Moreover, plasma D-dimer levels have been found to be predictive of left atrial thrombi and show a clear correlation with thrombi weights in those patients affected with mitral stenosis who are undergoing cardiac surgery.16 The US Physicians’ Study demonstrated a relationship between plasma D-dimer level >95th percentile and the occurrence of acute myocardial infarction.17
The present study was undertaken to verify whether plasma D-dimer levels depended on the intensity of oral anticoagulation and to verify whether plasma D-dimer level could be useful in predicting clinical outcomes in patients with mechanical heart valve prostheses.
A total of 145 patients who had undergone cardiac valve replacement with mechanical valves and who had been taking oral anticoagulants for >8 months were studied. All patients had undergone the replacement at least 3 months before the study was begun. Three patients suffering from recurrent deep-vein thrombosis were excluded from the study, because this condition heavily influences the plasma D-dimer level, so that 142 patients remained. One hundred thirty-two patients (60 and 72; mean±SD age, 60±10 years; range, 23 to 80 years) had received a single valve replacement (49 mitral and 83 aortic), and 10 patients had double valve replacements (7 mitral plus aortic and 3 mitral plus tricuspid). Atrial fibrillation coexisted in 45 patients with a mechanical single valve and in 8 with a double valve replacement. The types of monovalvular prostheses were 56 St Jude, 28 Sorin, 22 Bjork-Shiley, 12 Duromedics, 8 Starr-Edwards, 3 Carbomedics, 2 Omniscience, and 1 Lillehei-Kaster. The type of valve prostheses in polyvalvular replacements were 2 St Jude in 2 patients, 2 Bjork-Shiley in 2 patients, 2 Sorin in 1 patient, and 2 Duromedics in 1 patient; 2 patients had 1 Sorin plus 1 Duromedics, and 1 patient had 1 Sorin plus 1 St Jude. The baseline characteristics of our population (eg, number, age, sex, and presence of atrial fibrillation) are reported in Table 1⇓. The patients were seen at regular monthly intervals or more frequently, if necessary, by one of the cardiologists at the Cardiovascular Centre of Ospedale Maggiore of Trieste (northeastern Italy) to adjust their acenocoumarol dosage to maintain an INR between 2.4 and 4.4 (the target was 3.0).
One hundred two healthy subjects in normal sinus rhythm were recruited for the control group. They were sorted and matched for sex, age, and smoking habits from the general population of Trieste.18
Blood Sample Assays
The blood sample for determining plasma D-dimer level was taken during a regular INR check (namely, INR1) during March and April 1993. The plasma D-dimer level was determined by an ELISA (D-dimer Asserachrom, Boehringer Mannheim) according to the procedure previously published by our group.18 All determinations were carried out twice: the coefficients of variation (both interassay and intra-assay reproducibility) were within 5.5%. INR was immediately determined on a coagulometer ACL 300 (Instruments Laboratory) with the use of the ISI value provided by the manufacturer of thromboplastin (Thromborel S, Behring). This ISI value was periodically verified by a plasma calibration kit (a gift from Behring).
Analysis of D-Dimer, INR, and Their Relationship
We first compared the plasma D-dimer levels of patients with those of the control group; second, verified whether age, the coexistence of atrial fibrillation, and the site and type of valve prosthesis could influence D-dimer levels; and third, evaluated plasma D-dimer levels at different INR ranges. This last analysis was performed by dividing the patients into two subgroups: the first had their INR1 within the predetermined therapeutic interval, whereas the INR1 of the second group was below the range. A further comparison was carried out between patients whose INR1 was below the median INR1 at baseline versus those whose INR was above or below the mean INR value in the preceding 8 months (namely, INR2). In addition, we analyzed D-dimer levels and their relation to the percentage of time in the preceding 8 months during which the INR was within the therapeutic range. The percentage of time was based on the period between two consecutive determinations of INR: If it was within the therapeutic range, we considered the corresponding time as one of adequate anticoagulation; if otherwise, the time was considered one of inadequate anticoagulation. We assumed that the mean INR of a period between two determinations was related to the INR assessed at the end of each period. To estimate the time spent within the therapeutic range, we added all the days whose INR was estimated to be <2.4, and these were expressed as a percentage obtained by dividing the number of days of underanticoagulation by 8 months (243 days on average).
Patients were followed up to determine the incidence of adverse events for the next 2 years. Follow-up information was obtained by periodic examinations, careful review of hospital records, and direct contact with the patients’ general practitioners. If a patient died during follow-up, the autopsy report was checked. For patients who experienced both hemorrhagic and acute vascular events, data analysis was performed as if each had happened to two different patients.
Values are expressed as mean±SD for INR data and as median and IQR for D-dimer. For entire groups of patients, D-dimer has also been described in terms of percentile distribution. Mann-Whitney U and Kruskall-Wallis tests were carried out to evaluate D-dimer levels. Correlation coefficients were calculated by using Spearman’s correlation rank. For INR, the parametric statistic is appropriate, and hence, we adopted ANOVA with Scheffé’s analysis and t test. Multiple regression analysis was also performed: D-dimer was assumed to be the dependent variable, and age, the percentage of time spent in or below the adequate INR level, and the mean of INRs during the previous 8 months as the independent variables. Relative risk was determined by evaluating the events in each tertile of D-dimer and INR parameters. Logistic regression analysis was performed by using D-dimer and all INR parameter tertiles and baseline characteristics as variables and thrombotic and hemorrhagic events as end points. The continuous variables (age, D-dimer, baseline INR, mean INR during the previous 8 months, and percentage of time spent in adequate anticoagulation during the same period) were classified in tertiles and analyzed together with the categorical variables (sex, atrial fibrillation, number, and type and site of valvular implants). For all tests, values of P<.05 were considered statistically significant.
The mean INR1 value in the determination contemporaneous with the D-dimer assay was 3.01±0.86. The mean D-dimer value was 348±393 ng/mL (percentile values were 25th, 171; 33rd, 187; 50th, 244; 66th, 297; and 75th, 360 ng/mL; the mode was 171 ng/mL).
Plasma D-dimer level in patients with heart valve prostheses (244 ng/mL; IQR, 171:360 ng/mL; range, 59 to 2106 ng/mL) was similar to that of the control group (corresponding values of 320, 212:460, and 126 to 660 ng/mL). Patients with double valve replacement showed higher D-dimer values (430, 216:759, and 173 to 1847 ng/mL, respectively) than the control or the monovalvular implant group, although this increase was not statistically significant (Figure⇓). No differences in plasma D-dimer level and INR were found in patients with versus those without atrial fibrillation and with different sites and/or types of valvular implant. No linear correlation could be demonstrated between baseline INR values and plasma D-dimer levels.
We recorded the INR achieved during the previous 8 months and evaluated the relationship with the plasma D-dimer level. The mean INR value during this period was 3.2±0.3 (low tertile, <3.05; medium tertile, 3.05 to 3.28; and high tertile, >3.28). On average, patients spent ≈75% (range, 21% to 100%) of the previous 8-month period in the predetermined anticoagulation range (32% of patients >80% of the time and 8%, <50% of the time; low tertile, <0.68; medium tertile, 0.68 to 0.80; and high tertile, >0.80). The plasma D-dimer level was significantly higher in the group that had spent <75% of the time below the lower INR reference value (270 versus 198 ng/mL; IQR, 171:449 versus 171:293 ng/mL; P=.02). No relation between INR2 and the plasma D-dimer level was demonstrated.
For the entire group, correlation analysis showed a relation between age and plasma D-dimer level (r=.25, P≤.001) and a negative correlation between time spent in the therapeutic range and plasma D-dimer (r=−.30, P≤.01). Multiple regression analysis indicated that the major determining factors of plasma D-dimer level were age (R2=.07, P=.0088) and the percentage of time spent below the adequate INR level (R2=.09, P=.0011). These two variables taken together had an R2 of .16.
Thirty-five adverse events were recorded during the follow-up period: 19 acute vascular and 16 hemorrhagic events. The 19 acute vascular events were 2 fatal strokes, 2 fatal myocardial infarctions, 10 amaurosis fugax, and 5 cerebral transient ischemic attacks during the 2 years of follow-up. In 11 of the 16 hemorrhagic events, a condition potentially predisposing to bleeding was found: nasal sinus polyposis in 3, large-bowel polyposis in 2, gynecological malignancy in 2, periodontitis in 2, peptic ulcer in 1, and bleeding after surgery in 1. Two bleeding events (1 mild and 1 severe) occurred in the same patient. None of the bleeding events were fatal. One amaurosis fugax and 1 hemorrhage occurred in the same patient. Two amaurosis fugax were recorded in patients with two mechanical heart valve prostheses.
The incidence of adverse events was 1.5 for death, 7.2 for acute vascular events, and 6.0 for hemorrhage, all per 100 patient-years (Table 2⇓). Neither INR1 nor INR2 nor percentage of time spent was predictive of clinical outcome. The mean of plasma D-dimer level in those patients who subsequently died or experienced minor vascular events was higher than in event-free patients. Plasma D-dimer levels for event-free patients and control subjects was significantly different (Table 2⇓ and Fig 1⇑). Thrombotic and hemorrhagic events were equally distributed among the different categories of age, sex, atrial fibrillation, and site or type of valve replacement.
All deaths occurred in the upper tertile of D-dimer values, and the events in this category were more than four times those of the lower and middle tertiles (Table 3⇓). Hemorrhagic events were detected mainly in the middle (9 events) rather than the upper (5 events) or in lower (2 events) tertile. The relative risk for thrombosis in patients of the upper D-dimer tertile (>297 ng/mL) was 5.41 (95% confidence interval, 1.42 to 20.66) and 5.55 (95% confidence interval, 1.46 to 21.16) compared with middle (187 to 297 ng/mL) and lower (<187 ng/mL) tertiles, respectively (Table 3⇓). Logistic regression analysis demonstrated that the only predictor of thromboembolic events was D-dimer (R=.2827, P<.001), but age, INR1, INR2, sex, atrial fibrillation, number, and type and site of valvular implant were not predictive of the event. The same statistical approach did not enable us to predict hemorrhagic events.
Lifetime oral anticoagulation therapy is mandatory to reduce the risk of thromboembolic events in patients with prosthetic heart valves.1 Inadequate anticoagulation increases thromboembolism 2-fold to 6-fold. On the other hand, intense anticoagulation increases hemorrhagic risk in a dose-dependent manner.
Considerable effort has been made to define an optimal therapeutic range for INR in mechanical heart valve patients, in whom the risk of both thromboembolism and hemorrhage was low and acceptable. Altman et al,5 in a double-blind study, reported an INR of 2 to 3 to be safer and more protective against thromboembolism if antiplatelet drugs were used in tandem. Wilson et al,4 in a retrospective study, reviewed thromboembolic events in patients with low-intensity anticoagulation (prothrombin ratio 1.5×control) and high-intensity anticoagulation (prothrombin ratio >2) and concluded that low-intensity anticoagulation was adequate. Cannegieter et al,7 using a different approach, demonstrated that thrombotic and hemorrhagic events occurred outside the INR range of 2.9 to 4.9 and recommended an INR between 3 and 4. The strict monitoring of INR, together with the continuous education of patients and early identification of clinical conditions that can potentially lead to thrombosis or hemorrhage, can significantly decrease these adverse events, as was demonstrated by Cortellazzo et al.3 In that study, 70% of patients were maintained in the therapeutic range >50% of the time during the entire follow-up period. In our study, 92% of patients spent >50% of the time well anticoagulated. The lack of better results probably reflects the fact that maintaining anticoagulation status at all times can be a difficult task.6 With the INR range used and the percentage of time spent in the assigned range, our patient population can be considered properly treated and therefore representative of patients who usually attend the anticoagulation centers. Further research should take into account the fact that INR reflects the reduction in vitamin K–dependent clotting factors and is not necessarily related to the efficacy of warfarin in reducing thrombogenesis.19
Plasma D-dimer level, whose predictive ability has been demonstrated in this study and another previous one,17 rules out the clinical relevance of an increased resistance to fibrinolysis in low-thrombin/fibrinogen environments, such as those in anticoagulated patients.20 In this context, D-dimer, a valid and useful marker of thrombosis, is of particular significance. Until now, only the US Physicians’ Health Study reported higher plasma D-dimer levels in patients in a low-risk category, such as physicians, who subsequently developed cardiovascular events.21
Atrial fibrillation did not seem to influence plasma D-dimer levels in our patients. These data conflict with those of Kumagai et al,10 Gustafsson et al,11 and Lip et al.19 However, the fact that all of their patients were not anticoagulated and had not received a mechanical heart valve has to be taken into account. Both conditions could have masked the differences between patients with atrial fibrillation and control subjects or patients in normal sinus rhythm.
Multivariate logistic regression analysis showed that, more than age, plasma D-dimer level increases as a function of the time spent outside the assigned INR range but not of the mean INR. It might be that a few days spent below the therapeutic range needs several days of adequate anticoagulation for D-dimer levels to recover.
In our study, the incidence of adverse events was rather high, but it must be considered that both minor and major events were detected. The sensitivity of the survey allowed us to detect several minor events that heavily influenced the quality of life of the patients and that large trials usually do not take into account. D-dimer determination selects a wide population sample at risk and maintains its predictive value for at least 2 years. We do not believe that D-dimer should replace INR for evaluating anticoagulation in these patients, but it should be taken into account when evaluating their risk profile. In our study, patients and control subjects had similar mean values but a completely different distribution of plasma D-dimer levels. Hence, we evaluated the risk attributable to D-dimer tertile and considered only those data and percentiles obtained from the category under study. The results of logistic regression showed that D-dimer tertile was the only parameter able to predict the occurrence of events even when compared with other determinants. When selected on the basis of D-dimer tertile, these patients therefore should theoretically require a different follow-up approach. However, patients who experience hemorrhage often have a definite underlying disease that can promote the event, which mainly occurs in the middle and not the lower tertile as expected.
The double valve replacements represent a more complex, unresolved therapeutic problem. The plasma D-dimer level and the incidence of events in these patients suggest that they ought to be treated different from those with monovalvular implants, but the low numbers did not allow us to draw any comparison and, hence, any conclusion.
In conclusion, our data support the idea that the shorter the period spent within the therapeutic range, the higher the plasma D-dimer level and the higher the incidence of adverse events. The acute vascular events that occur after D-dimer determinations are made closely depend on its level for at least the 2 previous years. This identifies a high-risk patient category who probably needs a different approach to anticoagulation. Plasma D-dimer levels are clinically useless, however, for detecting patients who will experience hemorrhage. Further studies are needed to answer how to prevent these thromboembolic events and to verify whether D-dimer is a reliable marker not only of increased risk but also of its potential reduction.
Selected Abbreviations and Acronyms
|ELISA||=||enzyme-linked immunosorbent assay|
|INR||=||international normalized ratio|
|ISI||=||international sensitivity index|
This study was supported by a grant from the Italian Ministry for University and Scientific Research. We are indebted to Irene Tretjak and Paola Pitacco for technical assistance. We are grateful to Dr Peter Brown for the accurate English overview.
Whitaker AN, Elms MJ, Masci PP, Bundesen PG, Rylatt DB, Webber AJ, Bunce IH. Measurement of cross linked fibrin derivatives in plasma: an immunoassay using monoclonal antibodies. J Clin Pathol. 1984;37:882-887.
Gustafsson C, Blomback M, Britton M, Hamsten A, Svensson J. Coagulation factors and the increased risk of stroke in non-valvular atrial fibrillation. Stroke. 1990;21:47-51.
Ridker PM, Hennekens CH, Cerskus A, Stampfer MJ. Plasma concentration of cross-linked fibrin degradation product (D-dimer) and the risk of future myocardial infarction among apparently healthy men. Circulation. 1994;90:2236-2240.
Lip GYH, Lowe GDO, Rumley A, Dunn FG. Increased markers of thrombogenesis in chronic atrial fibrillation: effects of warfarin treatment. Br Heart J. 1995;73:527-533.
Lee MH, Mankowsky M, McDonagh J. Plasminogen binding impairment to clots formed at high fibrinogen concentration: high level of fibrinogen as a risk factor for thrombosis. Circulation. 1992;86(suppl I):I-1735. Abstract.