br METHODS br Study Sample br The authors performed
The authors performed a cross-sectional trend analysis of the impact of the 2009 guideline change on breast cancer incidence, using data from women aged 40−74 years who had a diagnosis of a breast cancer in U.S. Cancer Statistics (USCS); the combined data from the Centers for Disease Control and Prevention’s National Program for Cancer Registries (NPCR); and the National Cancer Institute’s (NCI’s) Surveillance, Epidemiology, and End Results (SEER) Program.21 The USCS 2001−2014 database includes cancer incidence and NPS-2143 data for all 50 states and the District of Columbia. Hospitals, physicians, and laborato-ries across the nation report data on demographic characteristics and tumor characteristics to central cancer registries supported by the Centers for Disease Control and Prevention and NCI. The NPCR and SEER Incidence−USCS Public Use Database (2001 −2014 database) covered essentially the entire U.S. population between 2001 and 2014. This study was not considered human subjects research by the IRB at the University of Texas Medical Branch, Galveston, Texas.
The information collected about each incident of cancer diagnosis included demographic characteristics, year of cancer diagnosis, and cancer histology. Stages of breast cancer were classified into four groups comprising in situ, localized, regional, and distant. Ages were grouped into 40−49 and 50−74 years, as screening guidelines differed in those two age groups. This study included in the analyses information about race (non-Hispanic white, non-Hispanic black, Asian/Pacific Islander, and other) and ethnicity (Hispanic or non-Hispanic). Hispanic ethnicity for all cancer cases was identified by the North American Association of Central Cancer Registries Hispanic/Latino Identification Algorithm.22 Statistical Analysis
All analyses were carried out using the SEER*Stat statistical soft-ware package, version 8.3.5. Women were divided into the follow-ing groups according to age 40−49 and 50−74 years. Breast cancer incidence rates were calculated as cases per 100,000 people and age adjusted to the respective standard population distribu-tion in these age ranges of the 2000 U.S. standard population. The crude incidence rate was the number of new cases of breast cancer (numerator) occurring in a specified population (denominator) in a given year, and the age-adjusted rate was a weighted average of crude rates, where the crude rates are calculated for different age groups and the weights are the proportions of people in the corre-sponding age groups of the 2000 U.S. standard population. CIs were calculated using the Tiwari method,23 which produces confi-dence limits that are similar to those with standard normal approximation when the counts are large and the population being studied is similar to the standard population, and is more likely to ensure proper coverage in other cases (e.g., the counts are small). Joinpoint regression models24 were fitted based on annual incidence data of 2006−2014 using the NCI’s Joinpoint Regres-sion Analysis program, version 4.6.0. This analysis program selected the best-fitting log linear regression model to identify the joinpoints (calendar year at diagnosis) when annual percentage changes (APC) differed significantly, allowing for the minimum number of joinpoints necessary to fit the data. APC was calculated as (exp[b]−1) £ 100, where the regression coefficient (b) was esti-mated by fitting a least-squares regression line to the natural loga-rithm of the rates, using the calendar year as a regressor variable. The number of significant joinpoints is determined by the permu-tation test. The grid search method was used to fit the segmented regression function, and the p-value of each permutation test is estimated using Monte Carlo methods; the overall asymptotic sig-nificance level is maintained through a Bonferroni adjustment.24 The p-value for the comparison of two segmented line regression functions (trends for two time periods) is estimated using the per-mutation procedure.25 Subgroup analyses were performed in age groups, races/ethnicities, and breast cancer stages. Additionally, 4-year average annual rates were calculated for 2006−2009 and 2011−2014 and compared the differences across age groups and races/ethnicities. Differences in age-adjusted rates were evaluated using RR and the corresponding 95% CI.26 Poisson model of vari-ation was selected for the analyses of breast cancer data as breast cancer is a rare event. Age-adjusted incidences were calculated for each cancer stage when comparing the differences between 2006−2009 and 2011−2014. When stage composition of breast cancer was calculated, the number of cases in each stage was age adjusted to the 2000 U.S. standard population. Chi-square tests were used to compare differences in cancer stage composition between 2006−2009 and 2011−2014. Statistical significances were determined as two-sided p-values <0.05. Data used in this study were collected in 2001−2014, released in 2017, and analyzed in 2018.