Activated protein C resistance

Abstract: We compare results of factor V DNA analysis with three different clotting-based assays designed to detect activated protein C (APC) resistance (APCR), using samples from 958 patients undergoing assessment for thrombophilia. The original and most commonly used APTT-based procedure (generating an APTT ratio in presence versus absence of APC), showed the least correlation with DNA findings, with a large overlap between normals and heterozygotes. Using this procedure, over 40% of patients with a normal DNA pattern gave APTT ratio results within the heterozygotes' ratio range, and thus is a poor predictor for factor V DNA Leiden mutation (sensitivity 94.3%, specificity 47.0% [APC ratio cut-off: 3.1]; sensitivity 52.1%, specificity 92.9% [APC ratio cut-off: 2.0]). Two commercially available procedures (protein C impedance [PCI] test and protein C pathway [PCP] test), using modified Russell's viper venom time (RVVT) assays, showed less overlap between normals and heterozygotes than did the APTT-based method. Fewer than 10% of normal individuals gave PCI or PCP test ratio results that fell within the respective heterozygotes' ratio range (PCI: sensitivity 95.3%, specificity 96.0%; PCP: sensitivity 97.3%, specificity 82.4% [APC ratio cut-off: 1.6 and 1.9 respectively]). Use of previously described normalisation procedures (patient's APTT ratio over pooled normal plasma [PNP] APTT ratio) showed little improvement in discriminatory power (sensitivity 96.4%, specificity 44.8% [normalised APC ratio cut-off value: 0; sensitivity 58.8%, specificity 90.1% [normalised APC ratio cut-off: 0.68]). Use of factor V-deficient plasma as sample diluent improved discrimination for all assays, but added considerable time and cost to the testing process. Furthermore, use of factor V-deficient plasma dilutions in the APTT-based test (sensitivity 97.1%, specificity 93.8% [APC ratio cut-off: 2.0]) did not substantially improve discrimination compared with either PCI or PCP performed without factor V-deficient plasma. Overall, a combination of RVVT- and APTT-based tests was found to provide excellent discrimination, particularly negative prediction, with respect to the likely factor V DNA result. Of 567 patients co-tested, all factor V DNA-normal patients (n = 299) gave both PCP-RVVT and APCR-APTT (not prediluted with factor V-deficient plasma) test ratio values *2.2. In conclusion, it is important to recognise the limitation of plasma-based assays, in particular the APTT procedure, to discriminate the factor V mutation.

Key words: Diagnosis. Factor V Leiden. Laboratory techniques and procedures. Partial thromboplastin time. Protein C. Prothrombin time.

Introduction

Activated protein C (APC) resistance (APCR) is a recently described risk factor for venous thrombosis,' reported in up to 40% of patients suffering familial thrombophilia.2-5 Some 90% of APCR cases were believed originally to arise from a single-point mutation in the gene for factor V.6" Heterozygosity for the gene defect is associated with a seven-fold increased risk of thrombosis, whereas homozygosity is associated with an 80-fold increase."' The presence of a DNA mutation (factor V Leiden) is evaluated by specific factor V DNA analysis, whereas APCR is assessed using functional clotting-based assays. The two different test procedures are often (incorrectly) linked synonymously, and there is common confusion regarding the best testing approach. DNA analysis is only available at specialist molecular biology departments, whereas the APCR procedure can be performed by any laboratory capable of doing coagulation assays. APCR is easy, quick, relatively cheap, and a commercial APTT-based kit (which uses an APTT ratio in presence versus absence of APC) is now in popular use. Unfortunately, the original observations which suggested general concordance of factor V gene mutation with the presence of APCR has not held true in many studies, with considerable overlap observed in `APC sensitivity ratios' between patients with normal factor V and those with the factor V Leiden mutation.7")" '3

Strategies to overcome the lack of sensitivity of the APTT-based method include the use of factor V-deficient plasma as a patient plasma diluent. Although this procedure adds another costly and time-consuming step to the APTT-based APCR assay, this modification increases the method's sensitivity to detect a factor V DNA defect, and may be of use in screening patients on oral anticoagulant therapy.ll 19 Another strategy involves the use of normalisation procedures potentially to improve discrimination between normal and heterozygous factor V patients.2' Unfortunately, this process works best when the pooled plasma (as used in the normalisation procedure) is a selected pool of factor V DNA-tested normals; thus, the preparatory process is timeconsuming and only available to testing sites where factor V DNA analysis is also available, confining it to specialised laboratories.

In the current study, we compare three different, commercially available clotting-based assays and undertake co-analysis of factor V DNA, using samples from 958 thrombophilia patients. The study assesses the effectiveness of the most commonly used APTTbased APCR clotting assay, compared to two new commercially available RVVT-based clotting procedures (protein C impedance [PCI] and protein C pathway [PCP] tests) to discriminate samples from those individuals with a normal factor V DNA pattern, and those showing a factor V mutation. We assess the effectiveness of these assays: (i) with and without patient plasma predilution in factor V-deficient plasma; and (ii) with and without normalisation procedures.

Materials and methods

Thrombophilia patients and blood samples

Samples for both factor V DNA analysis and APCR were obtained during the past three years from 958 patients undergoing routine diagnostic clinical assessment of thrombophilia. Samples were received both from within our institution (Westmead Hospital, a tertiary level referral hospital) and externally as referred specimens. As it is now well known that specimens derived from patients on anticoagulant therapy or those suffering coagulation defects such as lupus inhibitors, cause difficulties in APCR testing,'F'2 and are also generally considered unsuitable for such testing,23 we wished to assess the relative ability of each APCR procedure to 'cope' with such samples. Accordingly, for the purpose of comparison, patient samples were separated into test groups as follows: Group A: all thrombophilia patient samples; Group B: normal baseline coagulation/nonanticoagulated thrombophilia patient sample group samples derive from thrombophilia patients who gave normal baseline coagulation results (i.e. normal APTT and normal international normalised ratio [INR]) or who were not identified as on anticoagulant therapy at time of testing; and Group C: abnormal baseline coagulation/anticoagulated thrombophilia patient sample group - samples derived from thrombophilia patients who gave abnormal baseline coagulation results (i.e. high APTT and/or high INR) or who were identified as on anticoagulant therapy at time of testing.

All APCR samples tested were sodium citrate plasma, using venous blood collected into 0.11 mol/ L trisodium citrate (1 in 10 dilution), subsequently centrifuged (2000 x g, 15 min), and stored frozen until tested (typically, less than 1-2 weeks). All three clot-based APCR assays were performed, wherever possible, on each specimen received. However, not all specimens were tested using all assays because: (i) the RVVT-based procedures have only recently become available; (ii) in some instances there was insufficient plasma to perform all tests; and (iii) the PCI assay was withdrawn recently for legal reasons.

Coagulation-based test procedures

Evaluation of three different functional coagulationbased laboratory APCR procedures was performed, as described below. All assays were commercial kit methods and were performed as automated procedures, using an ACL-300R instrument (Coulter-IL, Sydney, Australia).

APTT-based procedure: Used according to manufacturer's instructions (Chromogenix, Sweden). As recommended, controls (normal and abnormal) were run in all assays. Recently, and in line with published findings,""'9 the manufacturer has recommended the use of factor V-deficient plasma as a patient plasma diluent, and this was also performed for many test samples (see below, and Results).

R VVT-based procedures: Used according to manufacturer's instructions (Gradipore Pty Ltd, Sydney, Australia). Both kit methods are based on an RVVT clot procedure. As recommended, controls were used and this permitted comparison with APTT-based assay controls. In some runs, factor V-deficient plasma was used as a patient plasma diluent, to allow comparison with results generated without the use of factor V-deficient plasma, or with other assay procedure results.

* Protein C impedance (PCI) test procedure: The kit provided RVVT reagent and a source of APC. An RVVT assay was performed on patient plasma, using the RVVT reagent and saline, and then repeated using the RVVT reagent/APC combination, and a ratio of test results then calculated (i.e. RVVT[+APC]/RVVT[-APC]).

* Protein C pathway (PCP) test procedure: The kit provided RVVT reagent and a source of protein C activator (PCA). This source is different from that used in the two previously described assay procedures. An RVVT assay was performed on patient plasma, using the RVVT reagent and a water/ saline (equal parts) mixture, then repeated using the RVVT reagent/PCA combination, and a ratio of test results calculated (i.e. RVVT[+PCA]/ RVVT[-PCA]).

Factor TV-deficient plasma

Typically, the factor V-deficient plasma used as a patient sample diluent was that manufactured and recommended specifically for use in the APTT-based method (Chromogenix). It was used in a patient plasma sample volume to total plasma volume ratio of 1 to 5.

Calculation of cut-off values for test comparisons and estimation of test sensitivity and specificity

For the APTT-based APCR assay, the manufacturer provided recommendations regarding the cut-off values, and these were followed to generate an APTT-based APCR cut-off value of 2.0 (very close to those published in the literature by users of this methodology). For the RVVT-based methods, the manufacturer recommends that laboratories generate their own cut-off values, but suggests target values between 1.5 and 2.0. In order to compare test procedures for predictive power, we generated various comparative cut-off values as follows: (i) in order to maximise the sensitivity of the test procedure, the cut-off value was the upper limit of the abnormal reference range (i.e. the mean of the factor V Leidenaffected group +2 SD; calculated as 3.1 for the APTT-APCR method, 1.6 for the PCI-RVVT method, and 1.9 for the PCP-RVVT method); (ii) in order to maximise the specificity of the test procedure, the cut-off value was the lower limit of the expected normal reference range (i.e. the mean value for the normal group -2 SD; calculated as 1.3 for the APTT-APCR method, 1.1 for the PCI-RVVT method, and 0.7 for the PCP-RVVT method); (iii) in order to equalise the resultant specificity and sensitivity of the test procedures, the cut-off value was the midpoint between the two values (i.e. lower normal limit and upper abnormal limit; calculated as 2.2 for the APTT-APCR method, 1.35 for the PCI-RVVT method, and 1.3 for the PCP-RVVT method); (iv) in order to capture all factor V Leiden-affected individuals, sensitivity set to 100%; and, (v) in order to capture all factor V-normal individuals, specificity set to 100%.

Factor V DNA analysis

Anticoagulated whole blood was stored at -20oC until processed. After thawing, a 50 gL sample was mixed with an equal volume of water, boiled for 15 min, and centrifuged at 12 000 x g for 10 min. The supernatant was used for DNA amplification, without further treatment. Primers used in this study were as described by Zoller and Dahlback.8 A positive control (heterozygous patient) and negative control (water) were included in each batch. The gel was stained with ethidium bromide and visualised with ultraviolet light. The Mnll restriction enzyme digest employed produces two fragments (43 bp and 118 bp) in patients without the factor V Leiden mutation; however, in patients with this mutation, one (heterozygous) or both (homozygous) alleles do not cut, and the 161 bp PCR product remains visible.

Normal ranges and pooled normal plasma (PNP)

PNP was prepared from more than 60 healthy donors and stored in samples (0.5 mL) at -80C until used. Samples of individual normal plasma were also processed and stored similarly for comparative assessment and derivation of normal ranges. The normal INR range for our laboratory is 0.8-1.3; thus, for the purpose of this study, INR values

Results

Factor lV DNA studies and incidence of factor V mutation

Concurrent APCR testing and factor V DNA analysis was performed on 958 thrombophilia patient specimens. The relative incidence of the factor V DNA (heterozygote) mutation in our patient group was 21% (200/958).

Clot-based APCR assays

APTT-based APCR procedure: Concurrent APTTbased APCR testing and factor V DNA analysis were performed on all specimens (n = 958). Using the unmodified APTT-APCR test procedure, APCR showed a high degree of overlap between factor V DNA heterozygous and normal individuals (Fig. 1, Group A). Exclusion of data from individuals known to be on anticoagulant therapy, or with high baseline APTT or INR values suggesting anticoagulation, factor deficiency or presence of lupus inhibitor, resulted in minor improvement in discrimination (Fig. 1, Group B). In particular, we identified some high APC ratios in the heterozygous factor V DNA group, and some low APC ratios in the normal factor V DNA group, that were excluded on this basis (Fig. 1, Group B), and also showed that overlap in test results was worse in the abnormal baseline coagulation/anticoagulated patient group (Fig. 1, Group C). Nevertheless, considerable overlap was still evident after exclusion of such data (Fig. 1, Group B). Normalisation procedures were also performed using these plasma samples (i.e. patient APTT-APC ratio result/PNP APTT-APC ratio result) with little improvement in discrimination (Fig. 2).

RVVT-based procedures: Concurrent with factor V DNA analysis, the PCI test was performed using plasma samples from >550 thrombophilic patients (Fig. 3), and the PCP test using plasma from >650 patient samples (Fig. 4). Although a small overlap in results is evident between factor V DNA normals and heterozygotes, most normals yielded results clearly outside the heterozygous group values. Analogous to the case for the APTT-based APCR procedure, relatively more overlap in test results was evident in the high coagulation baseline/anticoagulated patient group (Group C, Figs 3 and 4) than in the normal coagulation baseline/non-anticoagulated patient group (Group B, Figs 3 and 4).

Use of factor lj deficient plasma as sample diluent: It is now recognised that use of factor V-deficient plasma (as a test plasma sample diluent) improves discrimination of results between normals and heterozygotes in the APTT-based APCR procedure, and this was performed using many patient samples (Fig. 5). Its use produced a noticeable improvement in discrimination, particularly in the APTT-based assay, although some overlap between patient groups was still evident for all coagulation assay procedures.

Relative predictive value of coagulation-based APCR tests for the factor lV DNA mutation: Using the above data, the relative predictive values of each test procedure and one combination test procedure are summarised in Table 1.

Discussion

The relative incidence of the factor V DNA mutation reported by us in this group of thrombophilia patients (200 factor V heterozygotes/958 total tested; 21%) is similar to that reported in another Australian study in a smaller patient group24 (45 patients; 26% had factor V mutation). The current report confirmed the findings of others,'7`i-ll 13 tlt low APC ratio does not necessarily indicate a ) cd factor V gene, nor does a normal APC ratio nccc;sarilv indicate a normal factor V gene. As we also showed in this report, correlation with the standard APTT-based APCR assay is poor unless the patient plasma is prediluted in factor V-deficient plasma (compare Figs 1 and 5).

Some improvement in discrimination between factor V DNA groups (compare Fig. 1, Groups A and B) was evident after exclusion of APC ratio results derived from patients yielding high baseline coagulation results values or on anticoagulant therapy (Group C). As a result of this exclusion, some obvious artifact-high APC ratios in the heterozygous group and artifact-low APC ratios in the normal group became evident. Although additional improvement in discrimination was obtained following the factor V-deficient plasma dilution step (Fig. 5), some overlap in ratios between normals and heterozygotes was still evident, limiting the ability of this assay to predict mutated factor V gene (see also Table 1). This sample group included a high number of anticoagulated samples that did not yield APC-APTT clot values in the absence of factor V-deficient plasma (i.e. worst-case testing scenarios). Nevertheless, the presence of this overlap indicated that 100% sensitivity and specificity is not achievable with this test system. Additionally, the factor V-deficient plasma predilution procedure adds considerable expense to the overall testing cost, adds to the assay performance time, and necessitates additional specimen handling steps.

Thus, our data confirmed, as others have found,'23 that anticoagulated samples are largely unsuitable for analysis in the APTT-based APCR coagulation tests, unless a factor V-deficient plasma predilution step is undertaken; even then, discrimination for the factor V Leiden mutation is not assured. It was of interest to learn that 37% of samples submitted to a US reference laboratory were unsuitable for APCR testing because they were from patients on concomitant anticoagulant therapy.23 In our experience, some 30% of samples were from such patients, and most did not provide reliable APCR results by the APTT-based method (Fig. 1), although some improvement in reliability was seen when using the factor V-deficient plasma predilution step (Fig. 5). As noted previously, the use of factor V-deficient plasma may be of benefit in screening patients on oral anticoagulant therapy.""

RVVT-based clotting procedures were found to yield better discrimination between patients with the normal and mutated factor V genes. Almost complete separation between the DNA groups was evident, even without the factor V-deficient plasma predilution step (Figs 3 and 4; Table 1). As exclusion of data from high baseline coagulation/anticoagulated patients did not improve discrimination of heterozygotes and normals as markedly as for the APTTbased assay (Fig. 5), it can be concluded that RVVT assay procedures are more reliable markers of a potential factor V mutation using plasma from these patients. However, discrimination is improved further by using a factor V-deficient plasma dilution step (Fig. 5).

In our study, use of a normalisation procedure did not assist in discriminating heterozygotes and normals, unlike previous reports.'3'2' As previously reported, we confirmed that most heterozygotes gave normalised APC ratios

Our data also confirmed (using a large patient population) recently published findings using small sample populations and RVVT-based assays.zz,zs Thus, in a recent interlaboratory study evaluating the ability of different coagulation-based assays (as used in eight test laboratories) to discriminate the factor V DNA mutation, the RVVT test was confirmed as one of two best discriminating coagulation procedures (80 plasma samples tested).25 In another study, Aboud and Ma22 showed that, in a variety of clinical samples (70 test samples tested, including samples from individuals on anticoagulant therapy and with lupus inhibitors) the RVVT-based procedure was superior to the APTT-based procedure.

Although, in our study, each RVVT-based assay was found to provide a good degree of factor V group separation, we found the best approach was the use of a combination of the different clotting-based assays. Table 1 summarises and compares tests for their relative predictive value (or discriminatory power), using objective criteria for calculating `cutoff' values. As can be seen, use of the APTT-based procedure alone provided little ability to discriminate between patient groups, and sensitivity and specificity was better with the RV VT-based assays. Factor V-deficient plasma predilution markedly improved the predictive value of the APTT-based test, but not the predictive value of the RVVT-based tests, using the precalculated `cut-off' values. In part, this was due to the factor V-deficient plasma prediluted test group containing proportionally higher numbers of anticoagulated samples that did not yield clot values in the APC-based assay performed without factor V-deficient predilution (i.e. worst-case testing scenario).

In no single testing case (i.e. APTT-based procedure or RVVT-based procedure, with or without factor V-deficient plasma predilution) could we obtain total separation of test groups (i.e. 100% sensitivity, 100% specificity). However, appropriate use of results from the two RVVT-based assays or from a combination of clotting assays provided an excellent predictive basis for a normal factor V gene. As the PCI assay has now been withdrawn from sale, we currently use data from both the APTT-APCR and RVVT-PCP procedures (as screening assays) to determine the need for further laboratory testing or DNA analysis. Thus, using a cut-off value of 2.2 in both assays, the combination provided excellent specificity and predictive basis for a normal factor V gene (Table 1). Thus, we were able to avoid the need to perform DNA analysis in over 50% of samples received. In addition, co-use of the APTT-based assay helped to identify patients likely to be on anticoagulant therapy (e.g. high baseline APTT, often giving rise to artificially low APC ratios) and thus those samples potentially benefiting from a factor V-deficient plasma predilution (Figs 1 and 5). Based on our experience, this approach is more costeffective than performing factor V DNA analysis on all patient samples received.

Conclusions

Our current strategy is to utilise laboratory data from two different clotting-based assays. The likelihood of a normal factor V DNA result can be predicted with high probability, using a combination of assay results (in particular, RV VT-PCP and APTT-APCR test ratio results both 32.2). The APTT-based assay procedure supplements this approach by helping to determine whether a predilution step involving factor V-deficient plasma is likely to be of benefit. Accordingly, we concur with Wasserman and colleagues,l3 and suggest the use of a testing algorithm, using coagulation-based assay results to determine the need to perform the more specialised genetic testing procedure. Our own algorithm is shown as Fig. 6.

References

1 Dahlback B, Carlsson NI, Svensson PJ. Familial thrombophilia due to a previously unrecognised mechanism characterised by poor anticoagulant response to activated protein C: prediction of a cofactor to activated protein C. Proc Natl Acad Sci USA 1993; 90:1004-8.

Griffin JH, Evatt B, Wideman C, Fernandez JA. Anticoagulant protein C pathway defective in majority of thrombophilic patients. Blood 1993; 82:1989-93.

Koster T, Rosendaal F, de Ronde H, Briet E, Vandenbroucke JP, Bertina RM. Venous thrombosis due to poor anticoagulant response to activated protein C: Leiden thrombophilia study. Lancet 1993; 342:1503-6.

Stevsson PJ, Dahlback B. Resistance to activated protein C as

a basis for venous thrombosis. N Engl J Med 1994; 330:517-22. Cadroy Y, Sie P, Boneu B. Frequency of a defective response to activated protein C in patients with a history of venous thrombosis. Letters to the editor. Blood 1994; 83:2008-9. Bertina RM, Koeleman BPC, Koster T et al. Mutation in blood coagulation factor V associated with resistance to activated protein C. Nature 1994; 69:647. 7 Voorberg J, Roelse J, Koopman R et al. Association of idiopathic venous thromboembolism with single point-mutation at Arg-6 of factor V. Lancet 1994; 343:1535 6. Zoller B, Dahlback B. Linkage between inherited resistance to activated protein C and factor V gene mutation in venous thrombosis. Lancet 1994; 343:1536-8.

Zoller B, Svensson PJ, He X, Dahlback B. Identification of the same factor V gene mutation in 47 out of 50 thrombosis-prone families with inherited resistance to activated protein C. 3 Clin Invest 1994; 94:2521-4.

10 Rosendaal FR, Koster T, Vandenbroucke JP, Reitsma PH. High risk of thrombosis in patients homozygous for factor V Leiden (activated protein C resistance). Blood 1995; 6:1504-8. 11 Zehnder JL, Benson RC. Sensitivity and specificity of the APC resistance assay in detection of individuals with factor V Leiden. Am J Clin Pathol 1996; 106:107-11. 12 Voelkerding KV, Wu L, Williams EC et al. Factor V R506Q gene mutation analysis by PCR-RFLP: optimization, comparison with functional testing for resistance to activated protein C, and establishment of cell line controls. Am 7 Clin Pathol 1996; 106:1016.

13 Wasserman I,VI, Edson JR, Key NS, Chibbar R, McGlennen RC. Detection of the factor V Leiden mutation: development

of a testing algorithm combining a coagulation assay and molecular diagnosis. Aim _ Clin Pathol 1997; 108:427-33. 14 Dahlback B. Resistance to activated protein C, the Arg... to Gln mutation in the factor V gene, and venous thrombosis. Functional tests and DNA-based assays, pros and cons. Thromb Haemost 1995; 73:73942.

15 Jorquera JI, Montoro JM, Fernandez MA. Modified test for activated protein C resistance. Letters to the editor. Lancet 1994; 344:1162-3.

16 Trossaert M'I, Conard J, Horellou NIH. Modified APC resistance assay for patients on oral anticoagulant therapy. Letters to the editor. Lancet 1994; 344:1709. 17 Tosetto A, Rodeghiero F. Diagnosis of APC resistance in patients on oral anticoagulant therapy. Letters to the editor. Thromb Haemost 1995; 73:732-3.

18 Cadroy Y, Sie P, Alhenc-Glas M, Aiach M. Evaluation of APC resistance in the plasma of patients with Q06 mutation of factor V (factor V Leiden) and treated by oral anticoagulants. Letters to the editor. Thromb Haemost 1995; 73:734-5. 19 Kapiotis S, Quehenberger P, Jilma B et al Improved characteristics of APC-R assay: Coatest APC resistance by predilution of samples with factor V deficient plasma. Am _ Clin Pathol 1996; 106:588-93.

20 de Ronde H, Bertina RM. Laboratory diagnosis of APCresistance: a critical evaluation of the test and the development of diagnostic criteria. Thromb Haemost 1994; 72:880-6. 21 Bertina RM, Reitsma PH, Rosendaal FR, Vandenbroucke JP. Resistance to activated protein C and factor V Leiden as risk factors for venous thrombosis. Thromb Haemost 1995; 74:44953.

22 Aboud MR, Ma DDF. A comparison between two activated protein C resistance methods as routine diagnostic tests for factor V Leiden mutation. Br* Haematol 1996; 97:798-803. 23 Florell SR, Rodgers GM. Utilization of testing for activated protein C resistance in a reference laboratory. Am Z Clin Pathol 1996; 106:248-52.

24 Ma DDF, Williams BG, Aboud MR, Ibister JP. Activated protein C resistance (APC) and inherited factor V (FV) missense mutation in patients with venous and arterial thrombosis in a haematology clinic. Aust NZ Med 1995; 25:1514. 25 Tripodi A, Negri B, Bertina RM, Mannucci PM. Screening for the FV:Q506 mutation - evaluation of thirteen plasma based methods for their diagnostic efficacy in comparison with DNA analysis. Thromb Haemost 1997; 77:43W9.

EMMANUEL J. FAVALORO, OKSANA MIROCHNIK and DAVID MCDONALD*

Diagnostic Reference Haemostasis nad Molecular Biology Laboratories, Department of Haematology, Institute of Clinical Pathology and Medical Research (ICPMR), Westmeaud Hospital, Western Sydney Area Health Service, Westmead, NSW, 2145, Australia

(Accepted 18 September 1998)

Copyright Royal Society of Medicine Press Ltd. 1999
Provided by ProQuest Information and Learning Company. All rights Reserved