Great expectations: use of molecular tests and computerised prognostic tools in New Zealand cancer care

There is an international drive to improve outcomes for patients with cancer by individualising cancer treatment using technologies including molecular tests (MT) and computerised prognostic tools (CPT).1,2 MT utilise molecular information, for example variations in DNA sequence or RNA expression levels, to diagnose disease or to predict susceptibility or treatment outcome. CPT use computerised statistical models to combine large datasets with individuals clinical details to infer individualised prognoses.MT and CPT designed to aid clinical decision making for patients with a range of malignancies have been described.3 Molecular tests available in New Zealand (NZ) include: MammaPrint4 and Oncotype DX,5 which use gene expression analysis to derive a recurrence risk score for patients with early breast cancer; FLT3, NPM1 and CEBPA mutation analysis which provide prognostic information for patients with cytogenetically normal acute myeloid leukaemia (CN-AML) and are recommended in World Health Organization (WHO) guidelines;6,7 KRAS mutation analysis, which predicts response to cetuximab, an unfunded treatment for metastatic colorectal cancer;8 UGT1A1mutation analysis to predict irinotecan toxicity;9 EGFR mutation analysis to predict response to gefinitib and erlotinib for patients with non-small cell lung cancer.10In NZ we also have free online access to a number of CPT including Adjuvant!, which estimates recurrence risk and treatment benefit for patients with breast, colon or lung cancer.11 Further details of these examples of MT and CPT are given in Table 1. Table 1 Examples of molecular tests and computerised prognostic tools currently available in New Zealand for the care of patients with cancer Molecular test Type of cancer Clinical significance Method of detection Sensitivity Specificity Ref Oncotype DX Breast 21-gene test used to assign a tripartite recurrence risk score for ER-positive, lymph node negative breast cancers using a continuous variable algorithm. qRT-PCR 77% 55% 12 MammaPrint Breast 70-gene test use to assign dichotomous high or low risk of metastatic recurrence from a continuous variable. Microarray 90% 42% 13 FLT3mutation analysis CN-AML Internal tandem duplication is associated with constitutional activation of the FLT3 tyrosine kinase receptor and shorter disease free survival. PCR - - 7 NPM1mutation analysis CN-AML NPM1 mutations are associated with improved prognosis in the absence ofFLT3 mutation PCR - - 7 CEBPAmutation analysis CN-AML CEBPA mutations are associated with improved prognosis. PCR - - 7 KRASmutation analysis Metastatic CRC KRAS mutation predicts lack of response to anti-EGFR- antibodies (e.g. cetuximab) PCR 49% 93% 8 UGT1A1mutation analysis Metastatic CRC Presence of theUGT1A1*28 mutation predicts risk of severe neutrophenia in patients treated with irinotecan. PCR 23% 92% 9 EGFRmutation analysis Non-Small Cell Lung Cancer EGFR activating mutations predict response to EGFR tyrosine kinase inhibitors (e.g. gefinitib and erlotinib) PCR 77% 93% 14 Adjuvant! for breast cancer Breast cancer Uses clinicopathological data to predict overall and disease free survival, and the impact of endocrine therapy and polychemotherapy. Web-based computerised prognostic tool 70% 57% 12 CN-AML=cytogenetically normal acute myeloid leukaemia; CRC=colorectal cancer; ER=(o)estrogen receptor; qRT-PCR=quantitative reverse transcription polymerase chain reaction. The utilisation of MT and CPT during the management of patients with solid organ and haematological malignancy is likely to have a significant impact on clinical practice and health economics in NZ, however it has not been evaluated to date. The intent of this study is to determine the awareness and specific utilisation of MT and CPT amongst NZ cancer clinicians treating solid organ and haematological malignancy, and to ascertain their predictions for the impact of these technologies over the next 10 years. Methods An anonymous online questionnaire was used to survey clinicians who treat patients with cancer in NZ. The questionnaire was implemented using LimeSurvey software (Carsten Schmitz, Germany), a free open source survey application. It comprised 185 questions in three sections. Most questions in sections one and two had fixed click button answer options and a free text other option; where a numeric answer was required a free text box or slide rule was provided. In section 3, participants were shown clinical scenarios relating to their area of specialty. The scenarios presented situations in which molecular tests are purported to assist with clinical decision-making: stage II breast cancer, stage II colon cancer and CN-AML in remission after chemotherapy. Participants were invited to leave free text comments at the end of each section of the survey. The questions presented to each participant were determined by their previous responses such that each participant saw only those questions relevant to their clinical practice. The questionnaire took approximately 15 minutes to complete. Please visit http:\/\/www.bioinformatics.auckland.ac.nz\/doc\/project_data\/Supplemetary_figure_and_tables_FINAL.docx to view the questionnaire in full and all supplementary figures and tables. The University of Auckland Human Participants Ethics Committee granted ethical approval for this study. Medical and radiation oncologists, haematologists, pathologists and general surgeons practicing in NZ at specialist and trainee level were invited to participate by email via their professional societies and colleges. All trainees were enrolled in college-approved training programmes. Reminder emails were sent out 2 and 4 weeks after the initial invitation. Participation was incentivised with an iPad (Apple Inc., California, USA), won by a participant selected using a random number generator. The survey was conducted over 11 weeks, from 17th May to 1st August 2010. Responses from clinicians practicing outside NZ were excluded from analysis, as were incomplete responses that did not include details of the participants specialty and seniority. Data analysis was carried out using PASW Statistics 18.0 (IBM Corp., NY, USA), Excel 2008 version 12.2.9 (Microsoft Corp., Washington, USA) and VassarStats (faculty.vassar.edu\/lowry\/VassarStats.html). Relationships between independent categorical variables were analysed using the chi-square test for independence of association, relationships between non-independent variables were analysed using McNemars test. Where multiple tests were performed the Bonferroni correction was used. A P value of <0.05 was held to be significant and P<0.01 as highly significant. Results Survey participants - 739 clinicians were invited to participate in the survey. 186 clinicians accessed the online questionnaire; 137 completed it (Figure 1). Participants represented all invited specialties and included both specialists and trainees (Table 2). Specialists were significantly under represented relative to trainees (P<0.01); pathologists were significantly under represented relative to other specialties (P<0.05 for both pathology specialists and trainees). Participants worked in secondary, tertiary, academic and private practice settings. Figure 1. Participation in a survey investigating utilisation of molecular tests and computerised prognostic tools Table 2. Seniority and specialty of survey participants Seniority Specialty Number of participants (n=137) Number of clinicians invited to participate (n=739) Participation rate (%) Specialists General surgery Medical oncology Radiation oncology Haematology Pathology 26 17 15 15 11 140 71 45 61 217 (19) (24) (33) (25) (5) Registrars\/Fellows General surgery Medical oncology Radiation oncology Haematology Pathology 27 11 4 4 7 69 23 17 17 79 (39) (48) (24) (24) (9) Current practice\u2014A greater proportion of participants were aware of MT than CPT (92% vs. 69%, P<0.01) (Table 3). Awareness of MT by specialists vs. registrars showed no statistically significant difference (6% and 10%, respectively), however specialists were significantly less likely to be aware of CPT than registrars (60% vs. 81%, P <0.05). Fewer participants had ever used MT than CPT (43 vs. 78, P<0.01). Of participants aware of MT, 59\/126 (47%) reported that they had never used MT relevant to their clinical practice. Of participants aware of CPT, 12\/94 (13%) reported that they had never used CPT relevant to their clinical practice. Table 3. Awareness and utilisation of molecular tests and computerised prognostic tools amongst New Zealand cancer clinicians Variables Molecular tests n=137 Computerised prognostic tools n=137 Not aware of any tools\/tests n (%) 11 (8) 43 (31) Aware of tools\/tests n (%) 126 (92) 94 (69) Utilisation by those aware of tools\/tests Never used them Previously used them Currently used them 83 7 36 16 13 65 Table 4 presents data on factors that limited the use of those MT and CPT most commonly used in NZ. Supplementary Tables 1 and 2 present this data for all of the MT and CPT included in the survey. Factors reported to limit the use of MT and CPT varied. For example, awareness of both the CPT Adjuvant! and the MT Oncotype DX was high (78% and 86%) amongst participants who managed breast cancer (n=94) but while the use of Oncotype DX was most commonly limited by cost, use of Adjuvant! was most commonly limited by lack of clinical time (Table 4). For participants who prescribed chemotherapy, both the cost of mutation analysis and, in some instances, the cost of unfunded medications (e.g. cetuximab) limited their uptake of MT. Table 4. Factors that limited the use\u2020 of molecular tests and computerised prognostic models for the management of patients with cancer in New Zealand OncotypeDX FLT3 mutation analysis KRAS mutation analysis Adjuvant! for breast cancer (n=94) (n=26) (n=34) (n= 94) Current or previous user Never used Current or previous user Never used Current or previous user Never used Current or previous user Never used (n=10)