Development And Validation Of PICC Related Thrombosis Risk Prediction Model In Cancer Patients

Backgrounds:Peripherally inserted central catheter(PICC)cause catheter-related thrombosis(PICCRT),which is one of the most serious complications in patients receiving PICC placement.It is reported that cancer patients have overall higher incidence(2.22-19.50 folds)of PICCRT than non-cancer patients.PICCRT is the primary cause of the unplanned removal of PICCs,which increases the cost and prolongs the stay of hospitalized patients.Assessing cancer patients’ thrombotic risk factors in advance and moving forward thromboprophylaxis in high-risk patients can help reduce the incidence of PICCRT and improve its adverse outcomes.Therefore,it is required that clinicians change the previous clinical thinking of "treatment over prevention",and develop an effective risk predictive model of PICCRT in cancer patients.Understanding the clinical characteristics of PICCRT in cancer patients,the change of blood flow dynamics in the veins with catheter placement,and the effect of different thrombotic risk factors are the prerequisite and foundation for developing a PICCRT risk predictive model with excellent performance.Currently,the morbidity of PICCRT,such as time of onset,site and number of thrombosis,remains unclear.This is mainly due to the differences of PICCRT’s diagnostic criteria,detection methods,frequency and endpoints of patient follow-ups,and also due to the fact that the majority of the studies did not implement planned active patient follow-ups.The change of local blood flow dynamics of the cannulated vein is an important factor to trigger PICCRT after PICC placement.The few available relevant studies are based on the data acquired from static measurements at single time point among patients or from in vitro simulations.Hence,the effect of PICC placement on patients’ local blood flow dynamics needs to be further clarified.Many clinical scenarios can trigger PICCRT,and these scenarios can be divided into systemic risk factors such as age and hypercoagulable state,and local risk factors such as the diameter and blood flow velocity of the PICC placement vein.Clarifying the effects of these risk factors on PICCRT helps to screen the predictive factors with good positive predictive value.In this study,we chose cancer patients who received PICC placement,a high-risk group of PICCRT,as the study subjects.Scheduled colour Doppler imaging of the PICC route was performed every other day in 2 weeks after placement.The change of local blood flow dynamics and clinical characteristics of thrombosis were recorded and analysed.Based on the comprehensive evaluation of both local and systemic risk factors of PICCRT in this study,a risk predictive model of PICCRT in cancer patients were developed and internally validated according to the checklist of TRIPOD(Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis)Statement.Our established predictive model provides a tool for clinicians to accurately predict the risks of PICCRT in cancer patients,and guides the graded PICCRT prevention and treatment.Objectives:1.To reveal the clinical characteristics of PICCRT in cancer patients within 2weeks after PICC placement,including the incidence and time of onset,site of thrombosis,and clinical manifestations of thrombosis.2.To measure the changes of vascular diameter and blood flow velocity along the PICC route in cancer patients within 2 weeks after PICC placement,including insertion site,proximal segment of insertion site,axillary vein and subclavian vein;to determine the threshold vascular diameter and blood flow velocity that triggers PICCRT,as well as the cutoff point of CVR.3.To comprehensively evaluate and determine the effect of local risk factors such as number of insertions,threshold vascular diameter and blood flow velocity,and systemic risk factors such as age,hypercoagulable state,and body mass index,on triggering PICCRT in cancer patients within 2 weeks of PICC placement.4.To develop and perform internal validation of a comprehensive risk predictive model of PICCRT in cancer patients based on both local and systemic risk factors,to provide an efficient prediction tool for identifying patientsat high risk of PICCRT.Methods:1.This study is a single-center prospective cohort study.2.This cohort study consists of 173 cancer patients.Before PICC placement,patients’ general clinical information,results of blood test and DIC test were collected.Scheduled colour Doppler imaging of the PICC route was performed every other day within 2 weeks after PICC placement.The incidence,time of onset,site of thrombosis,patients’clinical manifestations,and special colour Doppler imaging observations were recorded.The clinical characteristics of PICCRT were presented by statistical analysis such as frequency,cumulative frequency,incidence,cumulative incidence,and histogram.3.The change trend and between-groups variance of patients’ vascular diameter and blood flow velocity before and after PICC placement between thrombosis group and non-thrombosis group were statistically analysed by using t-test and analysis of variance.The threshold vascular diameter and blood flow velocity triggering PICCRT and CVR cutoff point were determined by measuring the area under the receiver operating characteristic(ROC)curve.4.Both local and systemic risk predictive factors with good positive predictive value were screened based on the statistical analysis of blood flow parameters of the PICC placement route and the clinical characteristics of PICCRT,as well as clinical experience and literature research.These predictive factors are included in the development of a risk predictive model of PICCRT.5.Thrombotic predictive factors were further evaluated by LASSO regression analysis,and Logistic regression was used to develop the risk predictive model of PICCRT in cancer patients.C-index and calibration curve plot were used to evaluate the key performance of the predictive model,and a nomogram was formulated to illustrate the predictive model by using R software.Our predictive model was developed and reported in accordance with the TRIPOD Statement.Results:1.Clinical characteristics of PICCRT in cancer patients: Among 173 cancer patients in the cohort,thrombosis occurred in 126(72.83%)patients within 2 weeks of PICC placement,of which 75% of patients have asymptomatic thrombosis.All thrombosis occurred within 1 week after PICC placement,and the highest incidence(36.99%)was recorded at day 1.Thrombosis was detected at the insertion point in all PICCRT patients(N=126/173,72.83%),followed by the proximal segment of insertion site(N=120/173,69.36%),axillary vein(N=94/173,54.34%)and subclavian vein(N=41/173,23.70%).A total of 31 patients were diagnosed with symptomatic PICCRT with various clinical manifestations.The most common complaints included pain in the catheter arm during movement,feeling swollen and/or tightness in the catheter arm,and shoulder discomfort.Only 5 patients developed the typical symptom of PICCRT,"swollen upper arm",with an increase in arm circumference ranging from 0.8 to 2.8 cm compared with the circumference before PICC placement.2.Blood flow dynamics of PICC route in cancer patients: Before PICC placement,the vascular diameter and blood flow velocity were measured statistically different at the puncture point,proximal segment and axillary vein between thrombosis group and non-thrombosis group of patients.The vascular diameter of thrombosis group was significantly smaller than the diameter of non-thrombosis group,and the blood flow velocity of thrombosis group was significantly lower than that of non-thrombosis group.However,there was no statistical difference of the vascular diameter and blood flow velocity in subclavian vein between the two groups.After PICC placement,patients inthrombosis group and non-thrombosis group exhibit different changes of vascular diameter and blood flow velocity in different parts of PICC route: with prolonged catheter retention time,statistically significant decrease of the vascular diameter and blood flow velocity was measured at the insertion site,proximal segment of insertion site and axillary vein,and the decrease was more significant in thrombosis group than in non-thrombosis group.In the subclavian vein,only patients in the thrombosis group showed a decreasing trend in vascular diameter and flow velocity with prolonged catheter retention time,while no such trend was observed in the non-thrombosis group.ROC curve analysis showed that to trigger PICCRT,the threshold vascular diameter at insertion site,proximal segment of insertion site,and axillary vein are5.35 mm,5.45 mm,5.75 mm,respectively;the threshold blood flow velocity at insertion site,proximal segment of insertion site,and axillary vein are 16.65cm/s,16.60cm/s,22.90cm/s,respectively;and the CVR cutoff point at insertion site,proximal segment of insertion site,and axillary vein are 0.305,0.305,0.285,respectively.3.Effect of PICCRT risk factors in cancer patients: based on the structural equation modelling,except the path coefficient of catheter insertion times 0.06(P=0.369),the path coefficients of other local risk factors are statistically significant,ranging from-0.20 to-0.42(P<0.05).The path coefficients of systemic risk factors range from 0.14 to 0.22(P<0.05).It means that most local risk factors have larger path coefficients than that of systemic risk factors,indicating the local risk factors including vascular diameter and blood flow velocity at insertion site,proximal segment of insertion site,and axillary vein have greater effect on PICCRT than systemic risk factors such as age and hypercoagulable state.When thrombosis occurred in the subclavian vein,the path coefficients of vascular diameter and blood flow velocity were-0.04 and-0.08 respectively,while the path coefficients of systemic risk factors,such as age and hypercoagulable state,range from 0.15 to 0.31(P<0.05).This suggests that systemic risk factors play a more significant role in PICCRT than local risk factors when thrombi are formed in thesubclavian vein.4.Risk predictive model of PICCRT in cancer patients: Differences of PICCRT incidence were observed within 2 weeks of catheter placement at insertion site,proximal segment of insertion site,and axillary vein.Therefore,3 PICCRT risk predictive models were developed for each of the site respectively.Predictor variables of 3 models consisted of both local and systemic risk factors,including age,hypercoagulable state,catheter insertion times,vascular diameter,and blood flow velocity.However,every model only includes 4 of the above Predictor variables.Based on the nomogram of the predictor variables in the predictive models,patients can be graded to PICCRT low-risk and high-risk.The two-level risk model was measured by the value of C-index(0.830 – 0.952),calibration(0.862 – 0.897),and discrimination(0.722 – 0.936).The calibration plot for PICCRT risk showed optimal agreement between our model’s prediction and the actual observation,suggesting that this predictive model was able to accurately discriminate patients with high and low risks of PICCRT.It can be used by clinicians to predict the PICCRT incidence within2 weeks of catheter placement.Conclusions:1.Clinical characteristics of PICCRT in cancer patients: there is high incidence of thrombosis within 2 weeks of PICC placement,and these thromboses are mainly non-symptomatic.All thromboses occurred within 1 week after catheter placement,with the highest incidence on day 1.The incidence of thrombosis varies greatly at different venous sites.The further away the catheter is placed from central vein,the higher incidence and earlier onset of thrombosis.It is suggested that 1 week after PICC placement is the key intervention time for thrombosis prevention and control.2.PICC placement venous flow status in cancer patients: Small vascular diameter and slow blood flow velocity before PICC placement can increase the incidence of thrombosis.To prevent PICCRT,catheters should be placed in the veins with bigger vascular diameter and faster blood flow velocity,and smaller CVR.After PICC placement,regardless of whether the patient develops thrombosis,the vascular diameter of the placed vein becomes smaller and the blood flow slows down as the catheter retains for a prolonged period of time,and this downward tendency is more obvious in the thrombosis group of patients.It is suggested that clinicians should choose the veins with good vascular conditions to place the catheters in cancer patients,and interventions should be actively taken to prevent vasoconstriction and promote blood circulation during the period of catheter retention.3.The effect of PICCRT risk factors in cancer patients: In most cases,the local risk factors including vascular diameter and blood flow velocity have greater effect on PICCRT than systemic risk factors such as age and hypercoagulable state.It is suggested that local interventions,such as upper limb functional exercise,might be more effective than systemic interventions,such as anticoagulant drugs,in preventing PICCRT.4.Risk predictive model of PICCRT in cancer patients: Our predictive model evaluates both local and systemic risk factors,and is able to accurately discriminate patients with high and low risks of PICCRT.The use of this model can help clinicians predict the PICCRT incidence within 2 weeks of catheter placement.