Ischaemia-Modified Albumin Test for Venous Thromboembolism
Ischaemia-Modified Albumin Test for Venous Thromboembolism
This was a prospective diagnostic cohort study. The study was approved by the local ethics committee (reference 08/H1014/56). The Thromboembolism Assessment and Diagnosis (THREAD) study was conducted in an inner city 900-bed university hospital, providing medical care for around 320 000 patients each year. The outpatient DVT investigation service for the entire population is run by the ED, which sees approximately 80 000 ED presentations per year.
Any patient investigated in the ED for DVT and any patient investigated either as an ambulant patient or as a hospital inpatient for PE was eligible to be included in the study. The hospital uses an integrated electronic pathway for diagnosing DVT and PE. This acts as a step-by-step guide for medical staff and forms part of the patient's medical records. Patients were identified for the study by three methods: electronically when a pathway was commenced, when a member of staff paged the lead researcher and by checking the diagnostic imaging requests. Patients gave written informed consent to participate in the study. Patients were contacted either in person in the hospital or by telephone if they had been discharged home from the ED. To ensure the study cohort reflected the full range of patients investigated for VTE, exclusion criteria were kept to a minimum. Exclusions were inability to consent, refusal to participate and age below 16 years.
The researcher identified the first blood sample taken at presentation of symptoms. This sample had already been used to analyse renal function, and the remainder stored at 4°C. The serum samples were stored in a secure −80°C freezer and analysed by a biochemist who was blinded to patient diagnoses. The albumin cobalt binding (ACB) test (Inverness Medical) was used to measure serum IMA concentrations. The assay was performed on the automated P Module clinical chemistry analyzer (Roche Diagnostics Limited, West Sussex, UK) and calibrated using a 5-point calibration curve. Assay performance was monitored daily using three commercially available quality control samples. Samples were analysed daily in batches of 50.
Each patient underwent standardised assessment for VTE. For PE, this commenced with pretest probability assessment using the Manchester modified Wells' score. The score adds injecting drug use as an additional 3.0 points to the standard Wells' score. Patients scoring low probability (<2.0 points) and a negative latex agglutination D-dimer test (IL Test D-dimer, cut-off <230 ng/ml) had PE excluded. Patients with an elevated D-dimer or higher clinical probability underwent diagnostic chest imaging. Women who were below 35 years of age or pregnant had a PIOPED-interpreted ventilation–perfusion (VQ) scan. All others underwent CT pulmonary angiography (CTPA). Every patient was followed clinically for 3 months through hospital record search and direct telephone contact. If the patient could not be contacted by telephone, the general practitioners' records were searched. Death certificates were obtained for all patients who died during follow-up. PE was excluded by a normal D-dimer in a low clinical probability patient, a normal VQ scan, a low-probability VQ in a patient with low clinical probability or a normal CTPA. In addition, these patients required a normal follow-up. PE was diagnosed by a high-probability VQ in a patient with high pretest probability or a positive CTPA.
Patients investigated for DVT were assessed with the Wells' score for DVT. Injecting drug users, pregnant women and patients with a history of VTE were considered automatically high risk. A normal D-dimer in a patient scoring <3 excluded DVT. All others underwent ultrasound examination of the femoral and popliteal veins. DVT was excluded by two normal ultrasounds, performed 1 week apart, or one normal ultrasound with another more obvious diagnosis (such as Baker's cyst). These patients also required a normal 3-month follow-up.
One researcher (KH) recorded risk factors, examination findings and diagnostic test results, at the time of consent, by patient interview and note review. A secure database was created using Microsoft Access.
Data were analysed using the SPSS V.16.0. ROC curves were constructed to assess the utility of IMA and the ratio of IMA:albumin, in the diagnosis of all VTE, for PE and for DVT separately. The area under the ROC curves (AUC) was calculated in each case. A sensitivity analysis systematically excluded groups of patients who might add bias to the results—inpatients, renal failure and ischaemic heart disease.
An interim analysis was planned after the first 380 patients were recruited. The investigators predetermined that if the AUC for the ROC curve was not greater than 0.5 (or if the 95% CIs crossed this point), further testing would be discontinued. This allowed resources to be focused on tests that had a statistically significant association with PE or DVT.
A sample-size estimate was performed during the planning stage of the study. The study aimed to look for an alternative test to the combined use of clinical probability and D-dimer. The null hypothesis was the research test is no better than conventional assessment for VTE (with clinical probability and D-dimer). The sample-size calculation assumed a 15% prevalence of disease, an AUC for conventional assessment as 0.87 (data from the MIOPED study), correlation of the tests within positive participants of 0.8 and within negative participants of 0.4. A sample size of 800 patients could detect a difference in the AUC of 6% with p<0.05 and 80% power.
Methods
Study Design and Setting
This was a prospective diagnostic cohort study. The study was approved by the local ethics committee (reference 08/H1014/56). The Thromboembolism Assessment and Diagnosis (THREAD) study was conducted in an inner city 900-bed university hospital, providing medical care for around 320 000 patients each year. The outpatient DVT investigation service for the entire population is run by the ED, which sees approximately 80 000 ED presentations per year.
Selection of Participants
Any patient investigated in the ED for DVT and any patient investigated either as an ambulant patient or as a hospital inpatient for PE was eligible to be included in the study. The hospital uses an integrated electronic pathway for diagnosing DVT and PE. This acts as a step-by-step guide for medical staff and forms part of the patient's medical records. Patients were identified for the study by three methods: electronically when a pathway was commenced, when a member of staff paged the lead researcher and by checking the diagnostic imaging requests. Patients gave written informed consent to participate in the study. Patients were contacted either in person in the hospital or by telephone if they had been discharged home from the ED. To ensure the study cohort reflected the full range of patients investigated for VTE, exclusion criteria were kept to a minimum. Exclusions were inability to consent, refusal to participate and age below 16 years.
IMA Analysis
The researcher identified the first blood sample taken at presentation of symptoms. This sample had already been used to analyse renal function, and the remainder stored at 4°C. The serum samples were stored in a secure −80°C freezer and analysed by a biochemist who was blinded to patient diagnoses. The albumin cobalt binding (ACB) test (Inverness Medical) was used to measure serum IMA concentrations. The assay was performed on the automated P Module clinical chemistry analyzer (Roche Diagnostics Limited, West Sussex, UK) and calibrated using a 5-point calibration curve. Assay performance was monitored daily using three commercially available quality control samples. Samples were analysed daily in batches of 50.
Reference Standard
Each patient underwent standardised assessment for VTE. For PE, this commenced with pretest probability assessment using the Manchester modified Wells' score. The score adds injecting drug use as an additional 3.0 points to the standard Wells' score. Patients scoring low probability (<2.0 points) and a negative latex agglutination D-dimer test (IL Test D-dimer, cut-off <230 ng/ml) had PE excluded. Patients with an elevated D-dimer or higher clinical probability underwent diagnostic chest imaging. Women who were below 35 years of age or pregnant had a PIOPED-interpreted ventilation–perfusion (VQ) scan. All others underwent CT pulmonary angiography (CTPA). Every patient was followed clinically for 3 months through hospital record search and direct telephone contact. If the patient could not be contacted by telephone, the general practitioners' records were searched. Death certificates were obtained for all patients who died during follow-up. PE was excluded by a normal D-dimer in a low clinical probability patient, a normal VQ scan, a low-probability VQ in a patient with low clinical probability or a normal CTPA. In addition, these patients required a normal follow-up. PE was diagnosed by a high-probability VQ in a patient with high pretest probability or a positive CTPA.
Patients investigated for DVT were assessed with the Wells' score for DVT. Injecting drug users, pregnant women and patients with a history of VTE were considered automatically high risk. A normal D-dimer in a patient scoring <3 excluded DVT. All others underwent ultrasound examination of the femoral and popliteal veins. DVT was excluded by two normal ultrasounds, performed 1 week apart, or one normal ultrasound with another more obvious diagnosis (such as Baker's cyst). These patients also required a normal 3-month follow-up.
Data Collection and Processing
One researcher (KH) recorded risk factors, examination findings and diagnostic test results, at the time of consent, by patient interview and note review. A secure database was created using Microsoft Access.
Primary Data Analysis
Data were analysed using the SPSS V.16.0. ROC curves were constructed to assess the utility of IMA and the ratio of IMA:albumin, in the diagnosis of all VTE, for PE and for DVT separately. The area under the ROC curves (AUC) was calculated in each case. A sensitivity analysis systematically excluded groups of patients who might add bias to the results—inpatients, renal failure and ischaemic heart disease.
An interim analysis was planned after the first 380 patients were recruited. The investigators predetermined that if the AUC for the ROC curve was not greater than 0.5 (or if the 95% CIs crossed this point), further testing would be discontinued. This allowed resources to be focused on tests that had a statistically significant association with PE or DVT.
A sample-size estimate was performed during the planning stage of the study. The study aimed to look for an alternative test to the combined use of clinical probability and D-dimer. The null hypothesis was the research test is no better than conventional assessment for VTE (with clinical probability and D-dimer). The sample-size calculation assumed a 15% prevalence of disease, an AUC for conventional assessment as 0.87 (data from the MIOPED study), correlation of the tests within positive participants of 0.8 and within negative participants of 0.4. A sample size of 800 patients could detect a difference in the AUC of 6% with p<0.05 and 80% power.