British Columbia Cancer Research Articles

Follicular Non-Hodgkin Lymphoma Grade 3a and 3b Subtypes Exhibit Similar Clinical Behaviour and Treatment Outcome

The clinical usefulness of separating follicular lymphoma (FL) grade 3 into FL grade 3A (FL3a) and 3B (FL3b) subtypes remains controversial. The relative aggressiveness of these two subtypes and their potential curability with anthracycline-based therapy continues to be debated. Discrepancies in the literature may be related to the relatively small number of patients in most series and to the variable inclusion of patients with a diffuse histologic component on biopsy, likely representing early transformation. We present a retrospective review of clinical characteristics and outcome in a large cohort of patients with FL3a and FL3b, as strictly defined by WHO diagnostic criteria.Patients Patients diagnosed with FL grade 3 in the province of British Columbia (BC) between the years 1982 and 2008 were identified through the BC Cancer Agency Lymphoid Cancer Database. A total of 161 patients with pathology available for central review were included for analysis, and were categorized as FL3a (n=139) or FL3b (n=22) using current WHO criteria. Cases containing a diffuse component on biopsy were excluded.Results Median age of the entire cohort was 63 years (range, 18–88). There were no significant differences in age, sex, stage, serum LDH levels, or number of extranodal sites between the subgroups. However, there was a trend toward worse performance status in patients with FL3b compared with FL3a (PS 2–4 27% vs. 12%, respectively; p=0.054). Overall, more patients in the FL3b subgroup had a high-intermediate or high-risk IPI score compared with patients in the FL3a subgroup (36% vs. 17%, p=0.028). A significantly higher number of patients with FL3b received an anthracycline-based regimen as front-line therapy (82% vs. 36%, p <0.001) and approximately one third of the entire cohort received rituximab. With a median follow-up of 38 months (range, 1–224), no difference in overall survival (OS) (p=0.79) or disease-specific survival (DSS) (p=0.38) was observed between FL3a and FL3b patients (see Figure below) (5-year OS, 72% vs. 70% and 5-year DSS, 77% vs. 70%, respectively). Analysis limited to the FL3a subgroup showed no OS advantage for patients who received an anthracycline-containing regimen as initial therapy compared with those who did not (p=0.24).Conclusions Overall no significant differences in clinical outcome were noted between patients with FL3a and FL3b. No plateau in the survival curves was seen, suggesting that neither subtype is curable even with the use of anthracycline-based therapy. This large retrospective analysis, which expands on an earlier cohort from the province of British Columbia, calls into question the clinical relevance of this histological grading designation. [Display omitted]

Incidence and Survival for Gastric and Esophageal Cancer Diagnosed in British Columbia, 1990 to 1999

Geographical variation and temporal trends in the incidence of esophageal and gastric cancers vary according to both tumour morphology and organ subsite. Both diseases are among the deadliest forms of cancer. The incidence and survival rates for gastric and esophageal carcinoma in British Columbia (BC) between 1990 and 1999 are described. Incidence data for the period 1990 to 1999 were obtained from the BC Cancer Registry. Age-adjusted incidence and survival rates were computed by anatomical subsite, histological type and sex. All rates were standardized to the 1996 Canadian population. The estimated annual percentage change (EAPC) was used to measure incidence changes over time. Kaplan-Meier curves were used to show survival rates, and log-rank tests were used to test for differences in the curves among various groups. Between 1990 and 1999, 1741 esophageal cancer cases and 3431 gastric cancer cases were registered in BC. There was an increase in the incidence of adenocarcinoma of the esophagus over time (EAPC=9.6%) among men, and of gastric cardia cancer among both women (EAPC=9.2%) and men (EAPC=3.8%). Patients with proximal gastric (cardia) cancer had significantly better survival rates than patients with cancer in the lower one-third of the esophagus. Among gastric cancers, patients with distal tumours had a significantly better survival rate than patients with proximal tumours. The incidences of proximal gastric cancer and esophageal adenocarcinoma are increasing, and their survival patterns are different. Examining these cancers together may elucidate new etiological and prognostic factors.

CD20 Mutations at the Rituximab Binding Site Are Rare and Are Not a Significant Cause of R-CHOP Resistance in Patients with De Novo Diffuse Large B-Cell Lymphoma.

Background: Diffuse Large B Cell Lymphoma (DLBCL) is the most common non-Hodgkin's lymphoma and is not cured in 40% of patients who receive combined Rituximab + CHOP (R-CHOP) immunochemotherapy. The mechanisms of resistance to R-CHOP therapy are poorly understood. Rituximab is a humanized monoclonal antibody directed against the CD20 antigen on B lymphocytes. Since its addition to CHOP chemotherapy in 2001, it has reduced the mortality of patients with DLBCL by 50% in British Columbia (BC). Given this significant improvement in survival, rituximab must contribute an important role in the neoplastic B cell death. Its precise binding site on the CD20 antigen has recently been elucidated (Binder et al. Blood 2006). We hypothesized that mutations at this site could be a cause of failure to cure the DLBCL with R-CHOP. If so, detection of CD20 mutations could help risk-stratify patients and identify a group who may not benefit from the addition of rituximab to their chemotherapy regimen.Methods: We extracted DNA from 282 frozen DLBCL specimens (including 21 patients with Primary Mediastinal B cell lymphoma) at the BC Cancer Agency performed after March 2001, the date when the provincial treatment policy for advanced DLBCL was changed to R-CHOP. We amplified exon 6 of the CD20 gene which contains the rituximab epitope with the following primers: 5′-TGTAAAACGACGGCCAGTTTGGAATTCCCTCCCAGATT-3′ and 5′-CAGGAAACAGCTATGACGGATCCAGAGTTCATGCTCA-3′. In italics are the sequencing primers -21M13F and M13R. The purified 431 base pair product was bi-directionally sequenced using BigDye® Terminator v3.1 Cycle Sequencing Kit and a 3730 XL Applied Biosystems sequencer. The sequence reads where then analyzed using Polyphred/Consed.Results: 264 patients had successful sequences for this analysis. The clinical characteristics were available on only 241 pts and were as follows: median age 63 yrs (range 16–101); 129 (62%) were male; 64% IPI 0–2, 36% IPI 3–5. 197 pts received R-CHOP chemotherapy and were evaluated for outcome. The remaining patients were recorded as not receiving rituximab either because, the information was not available; they had limited-stage disease (in 2001), or were too frail to receive chemotherapy. 20% of patients relapsed or progressed after R-CHOP after a median follow-up time of 2 years (range 0.1–6.3 years). The sequencing analysis revealed 2/264 (0.008%) cases of CD20 mutations in exon 6, both in R-CHOP treated pts. One indicates a heterozygous 4 base pair (bp) deletion in nucleotides 353–356, upstream of the epitope. Clinically, this patient progressed on R-CHOP therapy. The other indicates a heterozygous 13 bp deletion at position 722 which is downstream of the epitope. This patient achieved a complete remission with R-CHOP. Interestingly, there were no Single Nucleotide Polymorphisms (SNPs) in this 431 base pair sequence which could also potentially impact rituximab binding at this site.Conclusions: The incidence of CD20 mutations at the rituximab binding site in 264 pts with de novo DLBCL is extremely low. Mutations at this site are therefore not a significant cause of R-CHOP resistance in this group of pts.

Both Discordant and Concordant Bone Marrow (BM) Involvement Predict for a Poorer Outcome Independent of the IPI in Patients with Diffuse Large B-Cell Lymphoma (DLBCL) Treated with R-CHOP.

Recently, it has been suggested that concordant but not discordant BM involvement is a negative predictor of outcome in DLBCL patients treated with CHOP (Chung et al, Blood 2007). The significance of BM involvement in pts treated in the current era of therapy including rituximab has not been fully examined. We evaluated the prognostic impact of BM involvement in DLBCL patients treated with the current standard of care, R-CHOP.Patients: We identified 282 patients with biopsy proven de novo DLBCL treated in British Columbia (BC) with an R-CHOP regimen between 01/01/2001 and 01/01/2005 with complete clinical information and staging bone marrows available for review. Cases were identified using the Lymphoid Cancer Database of the BC Cancer Agency. Median follow-up is 44 mos (range 1–77).Results: 234/282 (83%) had a negative staging BM, 27 pts (10%) had a positive BM with concordant histology (Con-BM) and 21 (7%) had a positive BM with discordant histology (Dis-BM) with predominantly small B-cells present. Clinical characteristics for the entire cohort were as follows: median age 64 y (range 18–88); 66% male; 65% stage III/IV; 39% PS>1; 51% elevated LDH; 31% >1 extranodal site. IPI risk factors: 10% 0; 42% 1–2; 48% 3–5. In addition to higher stage and greater extranodal involvement, pts with a positive BM were more likely to have a higher LDH and a poorer PS than pts with a negative BM. Compared to pts with Con-BM, pts with Dis-BM were more likely to be elderly, but otherwise exhibited a similar distribution of clinical prognostic factors. The Kaplan-Meier 4-year PFS was significantly worse for pts with both Dis-BM (32%) and Con-BM (46%) involvement compared to pts with a negative BM (75%) (p<0.0001) (see Figure). Similarly, 4-year overall survival was significantly worse for pts with Dis-BM and Con-BM involvement (53% and 54%) compared to pts with a negative BM (74%) (p=0.005). All patients with Dis-BM involvement who developed progressive or relapsed disease were believed to have progression of their aggressive lymphoma based on clinical behavior or biopsy proof. In a multivariate analysis controlling for the IPI, BM involvement (Dis-BM and Con-BM) remained an independent predictor of PFS (p=0.03).Conclusions: In DLBCL pts treated with R-CHOP, both discordant and concordant BM involvement predict for a poorer outcome independent of the IPI. It is possible that pts with DLBCL and discordant BM involvement represent a subset of pts with transformed disease which is inherently less treatment-sensitive. [Display omitted]

Evaluation of Cockcroft-Gault (CG) and abbreviated modified diet in renal disease (MDRD) formulae in dosing of cancer drugs other than carboplatin

2522 Background: Glomerular filtration rate (GFR) is often used to determine initial dosing of renally-excreted cancer drugs. Estimated GFR (eGFR) can be calculated using serum creatinine (SrCr) -based formulae such as CG and MDRD. MDRD is more accurate in non-cancer patients, does not require patient weight, and is reported automatically by all laboratories in British Columbia (BC). We previously showed that CG and MDRD have similar accuracy for carboplatin dosing in patients with gynecological malignancies. We now examine dosing of all renally-excreted cancer drugs in the general cancer population. Since this setting does not include routine measurement of GFR, we report the concordance of eGFR derived from CG and MDRD. Methods: Patient data were collected retrospectively at the BC Cancer Agency. The primary outcome was concordance of eGFR derived from CG and MDRD, using the method of Bland and Altman. A difference of ≥ 30% was assumed to be clinically significant because this difference would usually lead to dose adjustment based on reclassification of renal function. The secondary outcome was the proportion of patients who would have received a different initial dose due to difference in eGFR. Each patient’s dose was determined from dose adjustment tables stated in specific treatment protocols. Results: 313 patients were evaluated: 40% male, median 56y, 67.5kg, 166cm, SrCr 74micromol/L. Median eGFR derived from CG and MDRD were 86.8mL/min and 87.6mL/min, respectively. A difference of ≥ 30% in eGFR was found in 17.9% (56/313) of patients. 8.6% (27/313) of patients would have received a different dose due to difference in eGFR; of these, 67% (18/27) would have received a higher dose. Conclusions: There is good concordance of eGFR derived from CG and MDRD for most cancer patients, with less than 10% of patients expected to receive a different initial dose of chemotherapy. MDRD may be a reasonable alternative to CG for dosing of any renally-excreted cancer drug. No significant financial relationships to disclose.

Cost-effectiveness analysis (CEA) of third-line erlotinib therapy compared to best supportive care for advanced non-small cell lung cancer (NSCLC) in British Columbia (BC)

7678 Background: Erlotinib was approved for funding as a systemic therapy treatment for 3rd line management of advanced NSCLC by the BC Cancer Agency (BCCA) in April 2004. BCCA patient outcome and cost data are routinely collected to verify the therapeutic effectiveness and cost-effectiveness of systemic treatment policies. Methods: This was a pragmatic retrospective analysis of all patients who received 3rd line erlotinib compared to a historical group treated with 2nd-line docetaxel then no further active treatment, both according to BCCA protocol. The primary end-point was cost-effectiveness, measured in terms of cost per-life-year-gained. Secondary end-points included: median overall survival (MOS); overall survival (OS) at 1 year; and comparison to phase III efficacy results. Data was retrieved from the Cancer Agency Information System (CAIS) and Systemic Therapy Data Warehouse. Life-years- gained were calculated from the area under the survival function curve. CEA took the BCCA perspective and costs included all direct drug costs for treatment of advanced disease. Sensitivity analyses included varying life expectancy across its 95% CI, cost to the extremes of ranges, and start date for length of survival: method 1. progression after 2nd line therapy or 3 weeks post last dose of chemotherapy for control group and start date of erlotinib for treatment group; method 2. last date of 2nd line therapy for both groups Results: 75 control and 70 erlotinib patients were included in the analysis. Results are presented in the table . The Incremental Cost-Effectiveness Ratio (ICER) was $28,516 per life-year-gained under method 1, and $17,632 under method 2. The erlotinib group had similar 1-year OS compared to literature (36 vs. 31%). Conclusions: Erlotinib appears to be cost-effective in terms of life-expectancy under a range of assumptions. No significant financial relationships to disclose. [Table: see text]

Cost-effectiveness (CE) analysis of CHOP and rituximab for diffuse large B-cell lymphoma (DLBCL) in British Columbia (BC)

6623 Background: The BC Cancer Agency (BCCA) provides province-wide, population-based care. Outcomes are monitored to verify therapeutic effectiveness and justify funding for systemic treatment policies. The CE of rituximab with CHOP (cyclophosphamide, doxorubicin, vincristine, and prednisone) (CHOP-R) for DLBCL was compared to its predecessor, CHOP alone. Methods: This was a pragmatic population-based CE analysis based on the original cohort of advanced DLBCL patients described by Sehn et al (JCO 2005) who received either CHOP or CHOP-R between Sept 1999 and Aug 2002 (18-months pre and post availability of rituximab in BC) according to standard BCCA treatment policy at the time. The primary endpoint was CE in terms of life-expectancy (LE) at a median follow-up of 4 years (cost-per-life-year-gained). Costs were incorporated into a decision analysis including primary systemic therapy and downstream chemotherapy, radiotherapy, and stem-cell transplant (SCT). Actual incidence of each downstream therapy was converted to a probability for each group. Downstream therapy costs were then multiplied by these probabilities and added to the respective primary treatment costs. The CE analysis took the BCCA perspective which includes all direct costs for active cancer treatment, and hospitalization for SCT, but not ambulatory supportive care. Sensitivity analyses varying LE to the extremes of its 95% CI, modeling out to 15 years and discounting at 0, 3, and 5% were performed. Results: 292 patients were included and categorized to treatment received: 148 CHOP and 144 CHOP-R (median follow- up 5.4 and 4 years respectively). LE to 4 years was 30.18 months for CHOP and 39.44 for CHOP-R. OS at 4 years was 48.8% and 70.1% for CHOP and CHOP-R respectively (p&lt;0.0001) Respective costs of primary and downstream therapy were $4,682 and $7,198 for CHOP versus $26,366 and $6,228 for CHOP-R. The incremental CE ratio at 4 years median follow-up was $26,844 CDN per life year gained. Results were robust across univariate sensitivity analyses conducted. Conclusions: At 4 years median follow-up, CHOP-R improves LE and appears to be economically attractive at conventional thresholds. CE is an increasingly useful tool for the BCCA in making decisions about new cancer therapies. No significant financial relationships to disclose.

Risk of a second malignant neoplasm among 5‐year survivors of cancer in childhood and adolescence in British Columbia, Canada

We examined second malignancies, a recognized late effect of therapy among survivors of childhood and adolescent cancer, among a recent, population-based cohort of 2,322 5-year survivors diagnosed before 20 years of age in British Columbia (BC), Canada between 1970 and 1995. Survivors and second malignancies were identified from the BC Cancer Registry. Risk of second malignancy was evaluated using standardized incidence ratios (SIRs), absolute excess risk (AER), and cumulative risk. The effect of demographic, temporal, and disease-related characteristics on risk was assessed. Fifty-five second malignancies were observed after 26,071 person-years of follow-up. Relative rate of developing a second malignancy among survivors was 5 times higher than expected (SIR = 5.0, 95% CI, 3.8-6.5), and absolute excess risk was 1.7 deaths per 1,000 person-years. Cumulative incidence of a second malignancy was 5.1% at 25 years after diagnosis of the first cancer. SIRs and absolute excess risk of subsequent cancer was higher among females (SIR = 5.9, 95% CI, 4.5-8.3 and AER = 2.66). While relative risk of second cancer was higher for those diagnosed before 10 years of age (SIR = 10.6, 95% CI, 7.1-16.0), absolute excess risk was slightly higher for those diagnosed after 10 years of age. SIRs were significantly elevated for all follow-up periods, but absolute excess risk of a second cancer was highest among patients surviving more than 15 years. Increased risk of a subsequent neoplasm is evident among childhood cancer survivors diagnosed in more recent periods than has been previously reported, continues years after diagnosis, and varies according to several risk factors. Continued surveillance is essential to quantify and characterize long-term and changing risks for appropriate follow-up.

Mortality among 5‐year survivors of cancer diagnosed during childhood or adolescence in British Columbia, Canada

Ongoing monitoring of late mortality among survivors of a childhood or adolescent cancer is essential to appropriately evaluate risk in more recent cohorts and with longer follow-up. We examined overall and cause-specific mortality in a population-based cohort of 2,354 individuals diagnosed with a cancer or tumor prior to 20 years of age between 1970 and 1995 in British Columbia (BC), Canada who survived at least 5 years. Late deaths in a survivor cohort ascertained from the BC Cancer Registry were identified using death registrations. Standardized mortality ratios, absolute excess risk of death, and cumulative risk of death were determined. Demographic, temporal, and disease-related factors in risk of late mortality were also assessed. After 24,491 person-years of follow-up, there were 181 deaths, 139 of which were cancer related. Excess risk of late mortality among survivors was 7 deaths per 1,000 person-years at risk (AER = 6.6). Standardized mortality ratio (SMR) was ninefold higher relative to the underlying BC population (SMR = 9.1, 95% CI, 7.8-10.5), and was greatest for those with a recurrence within 5 years of diagnosis, and for those diagnosed with acute lymphoblastic leukemia and nervous system tumors. Absolute excess risk of late death was significantly higher for males and for those diagnosed prior to 1980, but did not vary according to age at diagnosis. Relative mortality was significantly increased due to cancer-related causes of death (SMR = 81.7, 95% CI, 68.6-95.8), as well as circulatory (SMR = 9.7, 95% CI, 4.2-19.1) and respiratory (SMR = 16.8, 95% CI, 4.6-43.0) diseases. In this population-based cohort with long follow-up, there continues to be excess late mortality among childhood and adolescent cancer survivors due to both cancer and non-cancer causes, even among more recently diagnosed survivors.

Transformed Lymphoma: Incidence and Long-Term Outcome.

Background: Development of large cell lymphoma in patients with an underlying indolent lymphoma is usually referred to as transformation and is thought to carry a poor prognosis. Its true incidence remains unknown with reported frequencies ranging between 20–70%.Objective: To determine the annual risk of transformation in a population-based group of patients diagnosed with indolent NHL in the province of British Columbia(BC) and to evaluate their outcome after transformation.Patients and Methods: The BC Cancer Agency lymphoma database was used to identify all patients diagnosed with follicular lymphoma (FL), lymphoplasmacytic lymphoma (LPL) or small lymphocytic lymphoma (SLL) between 1986 and 2001. Exclusion criteria: diagnosis outside BC; marginal zone or MALT histology; HIV positivity; age <15 or ≥61 y; incomplete staging at diagnosis or refusal of initial recommended treatment. The definition of transformation was based on either histological confirmation (HIST) or clinical features (CLIN), which included rapid discordant nodal or extranodal growth; sudden rise in LDH to > twice previous baseline; involvement of unusual extranodal sites other than bone marrow, spleen or peripheral blood; or new hypercalcemia. The extent of transformation (limited vs advanced) and treatment were also noted.Results: 406 records have been reviewed to date with a median follow-up for living patients of 84 months (range, 1–220). Initial characteristics: median age 49 y (range, 21–60); M:F ratio 1:1; histology, FL 345 (85%), SLL 32 (8%), LPL 29 (7%). 80 patients developed transformation, 58 (73%) based on HIST and 22 (27%) based on CLIN. The likelihood of developing transformation was approximately equal for each of the three histologic subtypes: FL 20%, SLL 15%, and LPL 10%. The majority of transformations were considered advanced, 58 (73%) vs 22 (27%) limited. Most patients, 66 (83%), were treated with CHOP-like chemotherapy at transformation and, of those, 14 patients also received radiation. The 5 y, 10 y and 15 y risks of transformation were 14%, 21% and 39%, respectively, with a continuous annual incidence of transformation of 2%. The 5 y, 10 y and 15 y overall survivals for the 326 patients whose lymphoma never transformed were 92%, 84% and 72% compared to 69%, 50%, and 26% for the 80 patients whose lymphoma did transform (p<0.0001). The post-transformation 5 y overall survival (post-T 5 y OS) of all 80 patients with transformed disease was 39% and the 5 y progression free survival (post-T 5 y PFS), 33%. Patients had a better outcome after limited than advanced transformation (post-T 5 y PFS (55% vs 22%, p = 0.02) and post-T 5 y OS (63% vs 30%, p=0.007)). Patients with transformation based on CLIN vs HIST criteria did not differ in their outcome (post-T 5 y PFS 39% vs 29%, p=NS, and 5 y OS 39% vs 38%, p=NS).Conclusion: The annual risk of transformation of indolent NHL lymphoma is 2% per year continuously for at least 15 y after initial diagnosis. Patients with limited extent of transformation have a much better post-transformation survival (63% vs 30%). A clinical diagnosis of transformation is just as accurate for predicting subsequent natural history as histologic proof. The development of transformation has a significant negative impact on the survival of patients with indolent lymphoma (10 y OS reduced from 84% to 50%).

Development of an agricultural job-exposure matrix for British Columbia, Canada.

Farmers in British Columbia (BC), Canada have been shown to have unexplained elevated proportional mortality rates for several cancers. Because agricultural exposures have never been documented systematically in BC, a quantitative agricultural Job-exposure matrix (JEM) was developed containing exposure assessments from 1950 to 1998. This JEM was developed to document historical exposures and to facilitate future epidemiological studies. Available information regarding BC farming practices was compiled and checklists of potential exposures were produced for each crop. Exposures identified included chemical, biological, and physical agents. Interviews with farmers and agricultural experts were conducted using the checklists as a starting point. This allowed the creation of an initial or 'potential' JEM based on three axes: exposure agent, 'type of work' and time. The 'type of work' axis was determined by combining several variables: region, crop, job title and task. This allowed for a complete description of exposures. Exposure assessments were made quantitatively, where data allowed, or by a dichotomous variable (exposed/unexposed). Quantitative calculations were divided into re-entry and application scenarios. 'Re-entry' exposures were quantified using a standard exposure model with some modification while application exposure estimates were derived using data from the North American Pesticide Handlers Exposure Database (PHED). As expected, exposures differed between crops and job titles both quantitatively and qualitatively. Of the 290 agents included in the exposure axis; 180 were pesticides. Over 3000 estimates of exposure were conducted; 50% of these were quantitative. Each quantitative estimate was at the daily absorbed dose level. Exposure estimates were then rated as high, medium, or low based on comparing them with their respective oral chemical reference dose (RfD) or Acceptable Daily Intake (ADI). This data was mainly obtained from the US Environmental Protection Agency (EPA) Integrated Risk Information System database. Of the quantitative estimates, 74% were rated as low (< 100%) and only 10% were rated as high (>500%). The JEM resulting from this study fills a void concerning exposures for BC farmers and farm workers. While only limited validation of assessments were possible, this JEM can serve as a benchmark for future studies. Preliminary analysis at the BC Cancer Agency (BCCA) using the JEM with prostate cancer records from a large cancer and occupation study/survey has already shown promising results. Development of this JEM provides a useful model for developing historical quantitative exposure estimates where is very little documented information available.

Health System Costs of Metastatic Breast Cancer

To assess the costs of treating patients with incurable breast cancer, all health system costs during the interval from diagnosis of first recurrence or metastasis until death for 75 female subjects randomly selected from those known to have died of breast cancer in British Columbia, Canada between July 1, 1995 and December 31, 1996, were identified. Costs were determined from several databases within the British Columbia (BC) Ministry of Health, as well as from BC Cancer Agency patient charts. The mean total cost to the health system was CDN $36,474.33 (95% confidence interval $29,752-$43,196) per subject. The mean costs were highest for the youngest age group and lowest for the middle age group, but these only differed by $2,300. Inpatient costs accounted for the greatest proportion of the total, over 50% in all age groups. This data may be valuable in assessing the cost-effectiveness of interventions that are known to affect mortality due to breast cancer.

Compliance with practice guidelines for node-negative breast cancer.

Practice guidelines for cancer management have been in use in the province of British Columbia (BC), Canada, since the mid 1970s. To evaluate practice guideline compliance, treatment received was compared with treatment recommended in a population-based cohort of women with breast cancer. All incident cases (n = 939) of invasive, pathologically node-negative breast cancer diagnosed in 1991 were identified from the BC Cancer Registry. Treatment details were abstracted from cancer clinic records for cases referred to the BC Cancer Agency (BCCA) (n = 661) and original source documents for nonreferred cases. Management decisions were considered compliant if the patient received the recommended treatment or was entered onto a randomized trial of the modality being assessed. Overall compliance with adjuvant therapy guidelines was 97% for radiotherapy, 96% for chemotherapy, and 89% for tamoxifen. An oncology specialist was consulted by 94% of patients with an indication for adjuvant treatment and by 58% of those without an indication (odds ratio [OR] = 10.7; 95% confidence interval, 7.0 to 16.4). Compliance with a guideline to deliver radiotherapy was 95%; with chemotherapy, 77%; and with tamoxifen, 68%. Compliance with a guideline that stated no adjuvant treatment was indicated was 99% for radiotherapy, 98% for chemotherapy, and 92% for tamoxifen. Noncompliance among patients with an indication for treatment was related to nonreferral to an oncology specialist and less complete implementation of guideline changes in the community as compared with cancer center practices. Compliance was high, but scheduled updating and more effective community implementation could further improve consistency of care.