Title : Patterns of chemotherapy treatment for women with invasive epithelial ovarian cancer – a population-based study

ed information including histological subtype, grade, stage at diagnosis, surgical procedures and chemotherapy information including treatment dates; drugs and doses given; and reasons for delays, dose reductions or drug cessation. De-identified data were returned to the investigators. The study was approved by the Human Research Ethics Committees at the Queensland Institute of Medical Research and all participating hospitals and cancer registries. A woman’s postcode was used to classify her degree of accessibility to major urban centers (major cities, inner regional, outer regional, remote, very remote) using the Accessibility/Remoteness Index of Australia (ARIA+) classification for 2006[17] and AC C EP TE D M AN U SC R IP T ACCEPTED MANUSCRIPT 5 socioeconomic status (SES) using the Socio-Economic Indexes for Areas (SEIFA) index of advantage and disadvantage.[18] Information on race/ethnicity was not available. Australia has a publically funded universal health care system which means that public hospital treatment is provided free of charge to everyone. However, private health insurance is also available so that insured people can choose to have treatment provided in private facilities by private specialist doctors. We were interested to see if treatment patterns varied according to health insurance status so recorded this information from the clinical notes. Coding of clinical information: Cancers were classified as serous, mucinous, endometrioid, clear cell (included any with a clear cell component), carcinosarcoma, other subtypes, and unspecified/undifferentiated epithelial cancers. The Australian guidelines recommend chemotherapy for women with stage IC+ cancers and for those with ‘high-risk’ stage IA/IB cancers (i.e. grade 3/poorly-differentiated or of clear cell histology).[16] Women were thus classified into four groups for analysis: early-stage/low-risk cancers (stage IA/IB, grade 1-2 [well/moderately-differentiated], non-clear cell histology); early-stage/high-risk cancers; stage IC or above; or unknown stage/grade. As most early-stage cancers are also low-grade, women with stage IA/IB cancers without grade information (n=12) were assumed to have early-stage/low-risk cancers; conversely, most high-grade cancers are stage IC or higher thus women with high-grade cancers without stage information (n=87) were assumed to have at least stage IC disease. As there is international variation in the recommended carboplatin dose, AUC 5 and 6 were considered standard for these analyses. Furthermore, women for whom dose information was missing were conservatively assumed to have started treatment at the standard dose. Although dose-dense regimens and intraperitoneal chemotherapy are increasingly considered acceptable alternatives, the guidelines current when the women in this study were diagnosed recommended that first line treatment should include a platinum AC C EP TE D M AN U SC R IP T ACCEPTED MANUSCRIPT 6 compound and, ideally, should be carboplatin and paclitaxel given three-weekly for six cycles.[16] Women were therefore classified, firstly, according to whether they were treated with a platinum-based drug (carboplatin/cisplatin); secondly, as to whether they started treatment with combination carboplatin-paclitaxel; and, thirdly, whether they completed at least six cycles of combination treatment at approximately three-weekly intervals. Women who completed six cycles of combination treatment were further categorised according to whether administration of any cycle was delayed by over seven days (defined as ‘delay’) or if the dose of either drug had to be reduced. In order to highlight variations in treatment patterns we chose a strict definition of ‘standard treatment’ and thus women treated with single-agent carboplatin, other regimens of carboplatin and/or paclitaxel and other drugs were classified as not having ‘standard’ treatment, although these treatments are appropriate in specific clinical situations (e.g., Australian guidelines suggest single-agent carboplatin for patients unsuitable for combination treatment on the basis of concurrent medical conditions or poor performance status). Using information from the clinical record, a comorbidity score was derived for each woman based on the Charlson Comorbidity Index.[19] If an index condition was not mentioned in the clinical record we assumed that a woman did not have that condition. Statistical analysis: Continuous variables were compared across groups using t-tests and nonparametric tests. Categorical variables were compared using chi-squared tests and chisquared tests for trend, with Fisher’s exact test used when any expected cell numbers were less than five. Multivariable unconditional logistic regression was used to identify variables associated with a dichotomous outcome (for example, chemotherapy yes vs. no) after adjusting for potential confounding variables. All statistical tests were two-sided and p< 0.05 was considered statistically significant.