MR-Lymphography and Lymph Node Staging in Prostate Cancer

Overview

This proposal is targeted at all patients with prostate cancer who are candidates for either curative surgery or curative radiotherapy in whom lymph node staging is indicated. Recently, it has been shown, that in patients with PSA \<10 ng/ml and Gleason score \< 7 the risk of lymph node metastases is low. Therefore, unnecessary PLND and non-invasive imaging can be avoided safely in this group. PLND is nowadays performed only in patients with intermediate or high risk for nodal metastases. Thus the subgroup of patients targeted in this study consists of patients with prostate cancer with a PSA \>10 ng/ml and Gleason score \> 6. * If the high sensitivity (90%) and negative predictive value (96%) of MRL can be validated in the 8 participating centres, in patients with a negative MRL invasive PLND may be avoided. * In patients with a positive MRL with enlarged nodes (larger than 8 mm) histological diagnosis may be obtained by imaged guided biopsy, and thus also in these patients avoid PLND. A limitation of image guide biopsy, however, is the 30% false negative rate. \[Barentsz, Oyen, Wolf\] * In patients with positive small nodes (smaller than 8 mm) the urologist may, focussed by the MRL findings of a positive node outside his "surgical field-of-view", extend his dissection, and thus improve his accuracy. * Based on the expected higher sensitivity of MRL this technique will completely replace CT-scanning.

Objective(s) of the proposed study The first objective of this study is to evaluate whether MRL can replace currently used nodal staging techniques such as CT-scanning and PLND in patients with prostate cancer. The second objective is to determine the effects of implementing this new technique on diagnostic costs and the costs of therapeutic consequences. The third, methodological, objective is to develop a protocol for diagnostic test evaluation. Define research questions of the proposed study 1. Can the results -especially the high sensitivity (92%) and high negative predictive value (96%)- of a pilot study with a new nodal staging technique (MRL), be reproduced by other investigators? 2. Can MRL improve the detection rate of small nodal metastases in normal sized nodes in patients with prostate cancer, compared to CT and lymph node dissection? 3. Should MRL replace CT and PLND in the future, based on its high negative predictive value? 4. Will replacement result in a decrease of the diagnostic costs and the costs of therapeutic consequences with equal quality of life using a time horizon of 6 months? Study design (motivate) The design of the study to compare MRL versus CT and PLND to detect pelvic lymph node metastases in prostate cancer patients exists of two parts. First to study the accuracy of the diagnostic options, consecutive patients will be subjected to both MRL and CT and further diagnostic and therapeutic action will depend on the absence or presence of lymph node metastases, irrespective of the test which indicated the metastases. Given the fact that both tests are performed in every participating patient and that both investigations will be made separately and read independently, a (randomised) comparative study is not needed. Because negative MRL will lead to PLND (based on pilot in 72% of the cases) and a positive MRL (28%) to biopsy, this observational study makes it possible to compare the MRL and CT results of all patients to the gold standard (histological findings). In the patients who will undergo PLND, the results of PLND can be compared directly to those of CT and MRL. As the imaging findings will guide the PLND, in these patients also the additional value of the diagnostic imaging techniques, e.g. removal of extra -image positive- nodes will be evaluated. The second part of the study concerns the analysis of the economic aspects of both imaging strategies in combination with further diagnostic and therapeutic courses of action. The design for studying this aspect will be a decision analytic model. The first step in the decision analysis is the design of the structure of the model. The model will reflect the current clinical strategies compared to the alternative clinical strategy based on MRL. This step of the decision analysis will exist of reviewing the literature in order to be able to design a concept model structure. A meeting of radiologists and urologists of the participating hospitals will be organised in order to discuss and validate the model structure. The next phase in the modelling study will be the application of probabilities, costs and patient outcome data in the model. The empirical part of the study will provide the necessary data. Selection criteria for study population / data sources / search strategy All patients with a histologically proven prostate cancer who are scheduled for PLND are eligible for the study. Patients are only included if they are at intermediate or high risk for lymph node metastases (PSA \> 10 ng/ml, and Gleason score \> 6). Excluded will be patients below 18 years of age, patients with contraindications for MRI (such as claustrophobia, pacemaker, surgical clips in the brain), and patients with a history of severe allergy to contrast agents or drugs (anaphylactic shock, severe asthma attack). Description of the intervention(s) (including follow-up) / selection of information / data extraction In 396 consenting consecutive patients with prostate cancer, meeting the study criteria, a standardized diagnostic protocol is performed. The protocol includes both standard CT and PLND as well as the new MRL technique. CT scanning will be performed with state-of-the-art helical CT scanners using a slice thickness of maximal 5 mm, with contiguous slices, after admission of intravenous and oral contrast. On the CT scan lymph nodes will be rated positive based on size and shape criteria, described by Jager et al.\[Jager96\] The MR-examination will be performed using state-of-the art equipment (that is at a magnetic field strength 1.5 or 1.0 T using a body phased-array coil) and sequences. Twenty-four to 36 hours after intravenous infusion of USPIO (particle size 35 nm; Sinerem, Guerbet, Paris, France) at a dose of 2.6 mg Fe/kg MR images will be acquired using high resolution 3 dimensional T1-weighted GRE, and 2 dimensional T2\*-weighted GRE sequences. The T2\*-weighted will be acquired in the axial and "obturator" planes. The 3D sequences will be evaluated by soft-copy reading using multiplanar reconstruction at a work-station. On the MRL images a node will be considered benign if it shows a reduction in signal intensity involving the entire node on the T2\*-weighted MR images, otherwise it will be considered metastatic. The CT and MR images will be assessed independently by two on-site readers and two centralized readers. The CT readers are not aware of each others results, and of the MR findings and vice versa. Prior to the start of the study the MRI protocol will be implemented and the image quality will be tested by the principal investigator. During the study regular quality control of both CT and MRL images will be performed, to ensure optimal results. Before entering patients, a learning course will be organized by the principal investigator to optimize the interpretation of the MRL images. For this purpose a set of images from the pilot study showing normal and abnormal nodes will be used. Both CT and MRL results will be used for further evaluation and treatment. If a positive enlarged node is found, image guided FNAB will be attempted. If metastases are confirmed no PLND will be done. When either CT or MRL show a small positive (\< 10 mm) node, the patient will have an image guided PLND. If CT and MRL are negative also PLND will follow. In case a metastasis is proven with either FNAB biopsy or histopathologic examination after PLND the patient will have hormonal therapy instead of prostatectomy or radiotherapy. If the nodes are negative patients will undergo either prostatectomy or radiotherapy. The PLND consists of either an open or a laparoscopic resection in which the obturator lymph nodes, external iliac and hypogastric nodes on both sides are resected. The imaging results are known prior to surgery by the urologist. He will be given a schematic drawing on which the radiologist will indicate all visible nodes of the MR images in relation to the iliac vessels, specifying both normal and suspicious ones. The urologist will be asked to dissect both normal-appearing lymph nodes and those that appear enlarged, with specific attention to pathological nodes on the CT and MR images. Also he will be asked whether the resected nodes are within or outside his "surgical field of view". During surgery the resected nodes will be placed on a grid which shows its location, which will thereafter be sent to the department of Pathology for further evaluation. The pathologist will indicate normal and metastatic nodes on an identical drawing to the one provided to the urologist. The histopathological examination of the nodes includes staining with hematoxylin and eosin, and if possible Prostate Specific Antigen (PSA) staining. Results of accuracy, sensitivity, specificity, negative and positive predictive value will be calculated for the different techniques. Because lymph node tissue will be available from all patients in the study, either by biopsy, laparoscopic or laparotomic PLND, histologic examination will be considered the gold standard for the presence or absence of metastases. Finally, end measurements (see below) will be performed and costs will be calculated. Also an attempt will be made to select a sub-group of patients, based on pre-imaging information, in which imaging is not cost-effective. Primary and secondary outcome parameters (theoretical and operational) Primary outcome parameters concern accuracy figures of both MRL and CT, being sensitivity, specificity, and positive and negative predictive values of both diagnostic tests. Secondary outcome measures are the costs per patient for each strategy. Besides this, quality of life after 6 months of follow up will be part of the analysis in order to explore the idea that both diagnostic and therapeutic strategies result in a comparable outcome at the patient level. Sample size calculation (motivate assumptions) and feasibility of recruitment Based on the pilot study in the prostate cancer patients (see above) the sensitivity of MRL is expected to be 90% with a NPV of 96%. The percentage of patients with positive nodes is expected to be 26%. In this study we intend to assess the accuracy of the new diagnostic test (MRL). We specifically do not intend to assess a statistically significant difference between the diagnostic performance of CT and MRL. The reason for this is that we believe that such a statistically significant difference per se is of no clinical relevance. It is merely the diagnostic performance of MRL itself which is the key factor whether the test will be implemented into clinical practice or not. The power analyses was therefore based on the reliability of the validity estimates. Following the paper by Buderer \[Buderer\] we evaluated the sample size needed to obtain a width of 6% of the 95% confidence interval around the sensitivity and specificity. With the expected sensitivity, specificity and prevalence of patients with positive nodes 376 patients are needed for the study to assess the reliability of the new diagnostic test (MRL) independent of the other test. Given a percentage of 5% lost to follow up 396 patients will be included in the study. Data analysis and presentation / synthesis The CT, MRL and PLND will be performed and interpreted in a standardised way (see "description of intervention"). The evaluation of the image findings will be performed independently by 2 on-site readers and 2 centralized readers. To test the reproducibility, the degree of agreement between observers will be quantified with kappa statistics. The diagnostic accuracy sensitivity, specificity, negative predictive and positive predictive value of pre- CT and MRL will be determined in relation to PLND and histopathologic findings with 95% confidence interval. Statistical analysis will be carried out with SAS software. The additional nodes found by PLND based on the image findings will be recorded. The decision analytic model will indicate the probability that a prostate cancer patient is diagnosed correctly regarding lymph node metastases. The expected costs of both strategies, the one with MRL and the one without, will be calculated by combining probabilities of clinical events happening and the costs involved. Multi-variable uncertainty of the data used in the model will be explored using probabilistic sensitivity analysis. The distributions of the variables to be used in this uncertainty analysis will reflect the uncertainty of these variables as found in the empirical study. For probabilities used, a confidence interval will be determined and for the variables indicating costs the range of cost prices found in the participating hospitals will be basis for this probabilistic sensitivity analysis. In this way, the uncertainty of the difference in cost of both strategies can be determined. In case one of the strategies will be both more expensive and more effective in terms of patient outcome, an incremental cost-effectiveness ratio such as the additional diagnostic costs per extra patient diagnosed correctly will be calculated. The uncertainty surrounding this ratio will be presented in a cost-effectiveness acceptability curve. In a one way sensitivity analysis the important accuracy data much as the false positive rate and the false negative rate of MRL will be varied in order to study the impact of possible efficacy modifying after implementation of MRL from this clinical study in day to day health care practise. The impact of these parameters on the modelling results will be indicated by adjusting the cost-effectiveness acceptability curve. Economic evaluation: general considerations The economic evaluation will concentrate on the comparison of the two diagnostic strategies in combination with the therapeutic follow up to determine the presence or absence of pelvic lymph node metastases in prostate cancer patients either by MRL or CT in combination with PLND. The empirical study is a non-comparative observational study, therefore the economic evaluation will be based on a decision analytic model using the data from the observational study comparing two possible diagnostic strategies: MRL versus CT and PLND. The general considerations motivating the design of the economic evaluation are the following: although MRL compared to CT in combination with PLND is expected to lead to a more accurate and therefore more adequate treatment of the patient having prostate carcinoma, the patient outcome for the longer term is not expected to be different between the strategies. Therefore, the economic evaluation is based on the principles of a cost minimisation analysis. Using a differential approach patients are followed up for a period of 6 months because this period will cover the expected and most important differences in diagnostic and therapeutic care and thus the direct medical costs. Because of the specific patient population in this study non medical costs are not expected to be different between the patients groups, thus a health care perspective is basis for the analyses, indicating that only health care costs are subject to study. The research question of the economic evaluation is: will this result in a decrease -and how much- of the diagnostic costs and the costs of therapeutic consequences with equal quality of life using a time horizon of 6 months? Economic evaluation: cost analysis In the decision analytic model MRL or CT in combination with PLND will lead to different therapeutic courses of action. The cost analysis reflecting these courses of action will be based on the empirical data gathering. The cost analysis exists of two main parts. First, on the patient level volumes of diagnostic and therapeutic care will be measured prospectively using standardised Case Report Forms (CRF). In each participating hospital research nurses will register in the CRF the number of times a patient visits the out patient department, the number of days in hospital (normal care and intensive care distinguished), the number and kind of diagnostic procedures (MRL, CT, histological examination), the type and number of operations (biopsy, laparoscopic lymph node dissection, laparotomic lymph node dissection, the latter in combination with radical prostatectomy), the duration of the specific operations, the number and duration of radiotherapeutic sessions, the personnel involved, the amount and kind of pharmaceutical therapies (e.g. hormonal therapy, pain killers) and so on. An important aspect in the cost analysis as input for the decision analytic model is the cost related to the false positives and false negatives regarding MRL and CT in combination with PLND. This part of the CRF will be designed and pilot tested in the first months of the study. The second part of the cost analysis consists of determining the cost prices for each volume parameter to use these for multiplying the volumes registered for each participating patient. The Dutch guidelines for conducting pharmaco-economic studies. \[CVZ\] and the guidelines for cost analyses will be used.\[Oostenbrink\] Because of the health care perspective of the study, approximations of real integral costs will be used in the cost analysis. As a basis for the cost price calculations in each participating hospital based on standardised calculation methods a costing study will be performed resulting in several estimates of a cost price. For the baseline analysis a weighed average (using the number of patients included in the study by each participating hospital) will be used. The range of the resulting cost prices and, if available, the reference values as mentioned in the national guidelines for each volume item will be used in extensive sensitivity analyses. Economic evaluation: patient outcome analysis As a main outcome of the observational study, several parameters for the accuracy of MRL versus CT in combination with PLND will be determined. Although no difference is expected in final patient outcome (survival), correct and incorrect diagnosis of the pelvic lymph node status of the patient may lead to different short term effects on patient outcome. To explore these possible differences quality of life measurement is foreseen in the observational study. A description of the disease specific and generic quality of life of the prostate cancer patient will be given on the basis of the intake measurement using the UCLA prostate cancer index and MOS short form 36 (SF36), respectively. The UCLA Prostate Cancer Index can be used for men with prostate cancer and can be administered in about 20 minutes. It includes six disease-specific domains to measure function and dysfunction in urinary, sexual, and bowel domains. The SF36 studies the domains physical functioning, problems with role functioning due to emotional problems, pain, mental health, vitality, and general health perception. It reports each domain in a figure between 0 and 100 without generating a general score, however. This instrument is applied in a wide variation of patient populations and is considered to be reliable, valid, and sensitive. At the end of the follow up (6 months after intake), the same set of questionnaires will be completed by the participating patients. Analysis of these data will concentrate on exploring a possible difference in change of quality of life form intake to end of follow up between patients in whom the status of the pelvic lymph node metastases (present or absent) was diagnosed correctly immediately or missed initially. In case a difference does exist contrary to our expectations, these patient outcomes will be used in the decision analytic model and will lead to a cost-effectiveness analysis instead of a cost-minimization analysis. Time schedule Include information about the organisation and planning The study period is 3 years. In the first 3 months (month 1-3) the diagnostic protocols will be optimized, with monthly joined meetings of all participating centers. During the following inclusion period (month 4 to 29), data will be collected and plenary meetings once every 3 months will take place or more frequently if needed. Co-ordination between centres is further facilitated by initially weekly and later in the study monthly meetings of the principal investigator and research personnel (research fellows, research nurses, administrative personnel). In this period 396 consecutive consenting patients will be included, and the outcome of the diagnostic tests is evaluated after 3, 12 and 20 months (respectively month 7, 16 and 24). Inter and intra-observer variation studies are also performed in this period. All patients will complete the study within 29 months. Data will be collected at the participating hospitals in a uniform way (CRF) and sent to the UMC Nijmegen. In the final period (month 29-35) the analysis of the economic aspects (modelling study) will be performed by the Department of Medical Technology Assessment at the UMC Nijmegen. In the final month study findings will be reported and practice guidelines will be presented. An interim analysis of the results of CT and MRL (sensitivity, specificity, NPV. PPV , accuracy) will be performed after 150 patients. As the negative predictive value of MRL should be at least 90%. The NPV is expressed by the formula: NPV= spec (1-prevalence) spec (1-prevalence) + prevalence (1-sens) Especially the sensitivity and NPV of MRL are critical. Assuming a fixed specificity of 93% and a fixed prevalence of positive nodes of 26%, the sensitivity should at least be 70% to obtain a negative predictive value of 90%. The study will be stopped if the sensitivity is statistically significantly smaller than 70%. Assuming a binomial distribution the cut-off points for the sensitivity at n=150 is 62%. For the NPV the cut-off value which is significantly lower than 90% is 83.5%. Although a low specificity will not result in different diagnostic results and there will not be increased costs, the study will be stopped if the specificity is lower than 75% after 150 patients, because a lower specificity will have a negative effect on the acceptance (and thus implementation) of MRL.