Factor VIII and Catheter-Related Thrombosis in Children
Factor VIII and Catheter-Related Thrombosis in Children
Our preliminary study is the first to evaluate biomarkers in predicting CVC-related DVT in critically ill children. Previous studies investigated the association of CVC-related DVT and inherited thrombophilias. We found that factor VIII activity is associated with incident CVC-related DVT, whereas G value is associated with prevalent DVT. The associations remained significant after adjusting for important predictors of CVC-related DVT. These findings are important because we may have found a test to stratify critically ill children based on their risk for CVC-related DVT. Currently, we do not have any tests to stratify these children. We may be able to use factor VIII activity, at a threshold of more than 100 IU/dL, to identify critically ill children who would or would not need pharmacologic thromboprophylaxis.
The relationship between factor VIII activity, platelet activity, and DVT is well described in adults with spinal cord injury. Factor VIII activity increases early in the course after injury while platelet activity does not increase until before DVT is detected. Consistent with these reports, we showed that factor VIII activity was elevated in critically ill children on the day of enrollment when they did not have a DVT and before the incident DVT developed. On the other hand, G value, which is a measure of platelet activity, is elevated in those with prevalent DVT. Based on our study design, we cannot determine whether the elevation in G value was a cause or a consequence of the DVT. However, in adults with spinal cord injury, increased platelet activity precedes the DVT. In contrast to our findings, some studies in adults and children report that factor VIII activity is elevated in those with prevalent DVT. These studies, however, measured factor VIII activity later in the course of illness. It is our conjecture that some form of consumptive coagulopathy might have occurred in our subjects with prevalent DVT that led to a reduction in factor VIII activity. D-dimer levels might have provided evidence of consumption, but we did not measure them because they did not predict DVT in critically ill adults. Kashuk et al reported in a retrospective study that elevated G value predicted DVT in critically ill adults admitted to the surgical ICU, which is contrary to our results. They analyzed the maximum G value at any time before DVT was diagnosed. This approach would not be useful when trying to predict DVT prospectively in clinical practice.
The most basic requirement for a biomarker is that it should be able to discriminate between patients with and without the outcome of interest. The significant associations between factor VIII activity and incident DVT and between G value and prevalent DVT support the discriminatory ability of these biomarkers. In addition, the AUC of both biomarkers was better than 0.50, which corresponds to random chance. Their performances are likely better than those of inherited thrombophilias. Except for factor V Leiden and prothrombin G20210A mutations, which may be associated with CVC-related DVT, none of the inherited thrombophilias have been shown to increase the risk for CVC-related DVT in children. Factor V Leiden and prothrombin G20210A mutations were not associated with DVT in our study partly because of inadequate statistical power. The performances of factor VIII activity and G value on their own are modest and slightly below the suggested minimum clinically acceptable AUC of 0.70. Combining the biomarkers did not improve their performance.
The optimal thresholds for factor VIII activity and G value identified using Youden index were both lower than previously reported thresholds. At the optimal threshold of more than 100 IU/dL, factor VIII activity was more sensitive (92.0% vs 33.3%) but less specific (41.3% vs 80.4%) than the previously reported threshold of more than 150 IU/dL. This suggests that factor VIII activity at the optimal threshold may be used as a screening test for incident DVT. Factor VIII up to 100 IU/dL may identify critically ill children who are at low risk for CVC-related DVT and would not need thromboprophylaxis. Because the normal values of factor VIII activity range from 50 to 150 IU/dL, our finding suggests that in critically ill children with CVC, factor VIII activity in the upper normal range may be sufficient to increase the risk for DVT. This is probably because these children have various degrees of inflammation, which is intricately intertwined with coagulation, that predispose them to DVT. In addition, factor VIII activity may need to be reduced to at most low normal values, such as with monoclonal antibody against factor VIII, to reduce the risk for CVC-related DVT. These hypotheses need to be tested in future studies. The sensitivity of G value using the optimal threshold of more than 11.0 kdyn/cm was better than the previously reported threshold of more than 12.4 kdyn/cm (50.0% vs 33.3%), although its specificity was lower (87.7% vs 94.5%). Because G value at either threshold is more specific than sensitive, it may be useful as a confirmatory, but not as a screening, test for prevalent DVT. The clinical application of this finding is limited because most cases of DVT can easily be confirmed with ultrasonography.
Another requirement for a biomarker is that it should be able to incrementally improve our ability to predict the outcome of interest above the predicted risk from established clinical predictors alone. This would justify the burden of measuring the biomarker. The appropriate method to evaluate this is to compare the AUC of a risk prediction model with and without the biomarker added to the model. Currently, there are no risk prediction models for CVC-related DVT in children. In its absence, we adjusted the associations between factor VIII activity and incident DVT for age, center effect, and recent surgery and between G value and prevalent DVT for age, center effect, recent surgery, and size of the CVC. The associations between the biomarkers and DVT remained significant after adjusting for these predictors, suggesting that the biomarkers may incrementally improve our ability to identify CVC-related DVT in critically ill children.
Our study has several strengths. We obtained consistent results when we evaluated the performance of the biomarkers using different measures. We actively surveilled all subjects ultrasonographically and detected both clinically apparent and asymptomatic CVC-related DVT. DVT is associated with adverse sequelae, such as pulmonary embolism, paradoxical embolic stroke, and loss of venous access, whether it is clinically apparent or not. We ascertained the diagnosis of incident DVT by screening for prevalent DVT on enrollment. We imaged the lower extremities to detect prevalent non-CVC-related DVT and excluded them in the analysis of incident DVT. All images were centrally adjudicated to minimize misclassification bias. We systematically tested for the most common inherited thrombophilias. We did not test for less common thrombophilias because they are not associated with CVC-related DVT. We also did not test for D-dimer, protein C, protein S, and antithrombin because they did not predict DVT in critically ill adults.
There were limitations to our study. We did not investigate the mechanisms underlying the association between the biomarkers and CVC-related DVT. The goal of prognostic research, such as our study, is to predict the outcome as accurately as possible and not to explain causality or pathophysiology, as is true for etiological research. Although elevated factor VIII may increase the risk for DVT, it may simply be a marker for other acute phase responses. We measured each biomarker once per subject. It would have been ideal to enroll children during the time of insertion of the CVC, obtain baseline study as soon as the CVC was inserted, and then measure the biomarkers serially within 24 hours after insertion of the CVC when the prevalence of CVC-related DVT is low and when the result might affect the decision to provide thromboprophylaxis. This was not feasible because CVCs were inserted emergently when the children were acutely ill and when parents were overwhelmed to provide consent. The average time from insertion of the CVC to obtain consent and draw blood was 19.8 hours. The performances of the biomarkers on their own and when combined are modest. Other biomarkers that are downstream to the activation of factor VIII (e.g., thrombin-antithrombin complex), biomarkers of platelet activation (e.g., platelet microparticles), or biomarkers of inflammation (e.g., interleukins) may be useful and should also be tested. Because this was the first study to determine the performance of factor VIII activity and G value in predicting CVC-related DVT, the study was not designed to assess the performance of these biomarkers at specific thresholds. We plan to validate these thresholds in future studies. Finally, we only enrolled children from two PICUs. Practices related to the care of the CVC and patient characteristics that may affect the risk for DVT may be different in other ICUs. Differences in patient characteristics may explain the difference in the frequency of DVT between our studies.
Discussion
Our preliminary study is the first to evaluate biomarkers in predicting CVC-related DVT in critically ill children. Previous studies investigated the association of CVC-related DVT and inherited thrombophilias. We found that factor VIII activity is associated with incident CVC-related DVT, whereas G value is associated with prevalent DVT. The associations remained significant after adjusting for important predictors of CVC-related DVT. These findings are important because we may have found a test to stratify critically ill children based on their risk for CVC-related DVT. Currently, we do not have any tests to stratify these children. We may be able to use factor VIII activity, at a threshold of more than 100 IU/dL, to identify critically ill children who would or would not need pharmacologic thromboprophylaxis.
The relationship between factor VIII activity, platelet activity, and DVT is well described in adults with spinal cord injury. Factor VIII activity increases early in the course after injury while platelet activity does not increase until before DVT is detected. Consistent with these reports, we showed that factor VIII activity was elevated in critically ill children on the day of enrollment when they did not have a DVT and before the incident DVT developed. On the other hand, G value, which is a measure of platelet activity, is elevated in those with prevalent DVT. Based on our study design, we cannot determine whether the elevation in G value was a cause or a consequence of the DVT. However, in adults with spinal cord injury, increased platelet activity precedes the DVT. In contrast to our findings, some studies in adults and children report that factor VIII activity is elevated in those with prevalent DVT. These studies, however, measured factor VIII activity later in the course of illness. It is our conjecture that some form of consumptive coagulopathy might have occurred in our subjects with prevalent DVT that led to a reduction in factor VIII activity. D-dimer levels might have provided evidence of consumption, but we did not measure them because they did not predict DVT in critically ill adults. Kashuk et al reported in a retrospective study that elevated G value predicted DVT in critically ill adults admitted to the surgical ICU, which is contrary to our results. They analyzed the maximum G value at any time before DVT was diagnosed. This approach would not be useful when trying to predict DVT prospectively in clinical practice.
The most basic requirement for a biomarker is that it should be able to discriminate between patients with and without the outcome of interest. The significant associations between factor VIII activity and incident DVT and between G value and prevalent DVT support the discriminatory ability of these biomarkers. In addition, the AUC of both biomarkers was better than 0.50, which corresponds to random chance. Their performances are likely better than those of inherited thrombophilias. Except for factor V Leiden and prothrombin G20210A mutations, which may be associated with CVC-related DVT, none of the inherited thrombophilias have been shown to increase the risk for CVC-related DVT in children. Factor V Leiden and prothrombin G20210A mutations were not associated with DVT in our study partly because of inadequate statistical power. The performances of factor VIII activity and G value on their own are modest and slightly below the suggested minimum clinically acceptable AUC of 0.70. Combining the biomarkers did not improve their performance.
The optimal thresholds for factor VIII activity and G value identified using Youden index were both lower than previously reported thresholds. At the optimal threshold of more than 100 IU/dL, factor VIII activity was more sensitive (92.0% vs 33.3%) but less specific (41.3% vs 80.4%) than the previously reported threshold of more than 150 IU/dL. This suggests that factor VIII activity at the optimal threshold may be used as a screening test for incident DVT. Factor VIII up to 100 IU/dL may identify critically ill children who are at low risk for CVC-related DVT and would not need thromboprophylaxis. Because the normal values of factor VIII activity range from 50 to 150 IU/dL, our finding suggests that in critically ill children with CVC, factor VIII activity in the upper normal range may be sufficient to increase the risk for DVT. This is probably because these children have various degrees of inflammation, which is intricately intertwined with coagulation, that predispose them to DVT. In addition, factor VIII activity may need to be reduced to at most low normal values, such as with monoclonal antibody against factor VIII, to reduce the risk for CVC-related DVT. These hypotheses need to be tested in future studies. The sensitivity of G value using the optimal threshold of more than 11.0 kdyn/cm was better than the previously reported threshold of more than 12.4 kdyn/cm (50.0% vs 33.3%), although its specificity was lower (87.7% vs 94.5%). Because G value at either threshold is more specific than sensitive, it may be useful as a confirmatory, but not as a screening, test for prevalent DVT. The clinical application of this finding is limited because most cases of DVT can easily be confirmed with ultrasonography.
Another requirement for a biomarker is that it should be able to incrementally improve our ability to predict the outcome of interest above the predicted risk from established clinical predictors alone. This would justify the burden of measuring the biomarker. The appropriate method to evaluate this is to compare the AUC of a risk prediction model with and without the biomarker added to the model. Currently, there are no risk prediction models for CVC-related DVT in children. In its absence, we adjusted the associations between factor VIII activity and incident DVT for age, center effect, and recent surgery and between G value and prevalent DVT for age, center effect, recent surgery, and size of the CVC. The associations between the biomarkers and DVT remained significant after adjusting for these predictors, suggesting that the biomarkers may incrementally improve our ability to identify CVC-related DVT in critically ill children.
Our study has several strengths. We obtained consistent results when we evaluated the performance of the biomarkers using different measures. We actively surveilled all subjects ultrasonographically and detected both clinically apparent and asymptomatic CVC-related DVT. DVT is associated with adverse sequelae, such as pulmonary embolism, paradoxical embolic stroke, and loss of venous access, whether it is clinically apparent or not. We ascertained the diagnosis of incident DVT by screening for prevalent DVT on enrollment. We imaged the lower extremities to detect prevalent non-CVC-related DVT and excluded them in the analysis of incident DVT. All images were centrally adjudicated to minimize misclassification bias. We systematically tested for the most common inherited thrombophilias. We did not test for less common thrombophilias because they are not associated with CVC-related DVT. We also did not test for D-dimer, protein C, protein S, and antithrombin because they did not predict DVT in critically ill adults.
There were limitations to our study. We did not investigate the mechanisms underlying the association between the biomarkers and CVC-related DVT. The goal of prognostic research, such as our study, is to predict the outcome as accurately as possible and not to explain causality or pathophysiology, as is true for etiological research. Although elevated factor VIII may increase the risk for DVT, it may simply be a marker for other acute phase responses. We measured each biomarker once per subject. It would have been ideal to enroll children during the time of insertion of the CVC, obtain baseline study as soon as the CVC was inserted, and then measure the biomarkers serially within 24 hours after insertion of the CVC when the prevalence of CVC-related DVT is low and when the result might affect the decision to provide thromboprophylaxis. This was not feasible because CVCs were inserted emergently when the children were acutely ill and when parents were overwhelmed to provide consent. The average time from insertion of the CVC to obtain consent and draw blood was 19.8 hours. The performances of the biomarkers on their own and when combined are modest. Other biomarkers that are downstream to the activation of factor VIII (e.g., thrombin-antithrombin complex), biomarkers of platelet activation (e.g., platelet microparticles), or biomarkers of inflammation (e.g., interleukins) may be useful and should also be tested. Because this was the first study to determine the performance of factor VIII activity and G value in predicting CVC-related DVT, the study was not designed to assess the performance of these biomarkers at specific thresholds. We plan to validate these thresholds in future studies. Finally, we only enrolled children from two PICUs. Practices related to the care of the CVC and patient characteristics that may affect the risk for DVT may be different in other ICUs. Differences in patient characteristics may explain the difference in the frequency of DVT between our studies.
Source...