Article type
Year
Abstract
Introduction/Objective: To evaluate whether differences in the conduction and design of the studies may affect the performance of tumour marker TFA in the diagnosis of lung cancer.
Methods: A systematic search on MEDLINE, CancerLit and on references of relevant articles was performed in order to retrieve all the relevant published papers on the usefulness of TPA in the diagnosis of lung cancer. From each paper, details on type of gold standard, methods of performing the TPA test, criteria for patients selection, spectrum of patients included, time of administration of test relative to the reference standard, check for verification bias were gathered together with data of patients characteristics, the cut-off chosen, sensibility and specificity of the test.
Results: Overall, 25 articles, published between 1990 and 1997, have been evaluated: data of 979 healthy volunteers, 1156 patients with benign pulmonary disease, and 2234 lung cancer patients have been analysed. All the studies used the hystological confirmation of lung cancer as gold standard. The cut-off used was determined using an "ad hoc" sample of healthy controls or patients with benign pulmonary disease in 12 (48%) studies; details on determination of the cut-off value were provided in 21 (84%) studies. Twenty one studies (84%) presented clinical symptoms and radiological findings consistent with the diagnosis of lung cancer before administering the test; TPA concentration was determined after the hystological diagnosis in 23 (92%) studies, and only 1 study (4%) was performed on ambulatory patients. In most of the studies, the value of the cut-off was determined in order to get a specificity greater then 90%: therefore only the sensibility of the test could have been assessed. Sensibility was of 42% ranging from 30% to 83%, showing a great heterogeneity of results among studies. Using a multivariate logistic approach, the probability of cancer patients of being correctly classified as true positive was considered as function of study characteristics. The test administered in patients after the hystological diagnosis showed a reduced sensibility compared with test administered before the gold standard (odd ratio: 0.59, 95% CI: 0.40-0.86, p=0.006), and sensibility was higher if the criteria for determining the cut-off value were not determined using an "ad hoc" sample of healthy controls for choosing the value of cut-off at specificity of 90%-95% (odds ratio:1.45, 95% CI: 1.06-1.99, p=0.019).
Discussion: In the determination of the performance of diagnostic tests crucial aspects related to the design of the study, like the timing of test administration and criteria for determination of cut-off used, may have an impact on the outcome and should be carefully evaluated.
Methods: A systematic search on MEDLINE, CancerLit and on references of relevant articles was performed in order to retrieve all the relevant published papers on the usefulness of TPA in the diagnosis of lung cancer. From each paper, details on type of gold standard, methods of performing the TPA test, criteria for patients selection, spectrum of patients included, time of administration of test relative to the reference standard, check for verification bias were gathered together with data of patients characteristics, the cut-off chosen, sensibility and specificity of the test.
Results: Overall, 25 articles, published between 1990 and 1997, have been evaluated: data of 979 healthy volunteers, 1156 patients with benign pulmonary disease, and 2234 lung cancer patients have been analysed. All the studies used the hystological confirmation of lung cancer as gold standard. The cut-off used was determined using an "ad hoc" sample of healthy controls or patients with benign pulmonary disease in 12 (48%) studies; details on determination of the cut-off value were provided in 21 (84%) studies. Twenty one studies (84%) presented clinical symptoms and radiological findings consistent with the diagnosis of lung cancer before administering the test; TPA concentration was determined after the hystological diagnosis in 23 (92%) studies, and only 1 study (4%) was performed on ambulatory patients. In most of the studies, the value of the cut-off was determined in order to get a specificity greater then 90%: therefore only the sensibility of the test could have been assessed. Sensibility was of 42% ranging from 30% to 83%, showing a great heterogeneity of results among studies. Using a multivariate logistic approach, the probability of cancer patients of being correctly classified as true positive was considered as function of study characteristics. The test administered in patients after the hystological diagnosis showed a reduced sensibility compared with test administered before the gold standard (odd ratio: 0.59, 95% CI: 0.40-0.86, p=0.006), and sensibility was higher if the criteria for determining the cut-off value were not determined using an "ad hoc" sample of healthy controls for choosing the value of cut-off at specificity of 90%-95% (odds ratio:1.45, 95% CI: 1.06-1.99, p=0.019).
Discussion: In the determination of the performance of diagnostic tests crucial aspects related to the design of the study, like the timing of test administration and criteria for determination of cut-off used, may have an impact on the outcome and should be carefully evaluated.