Agency, Northwest Portland Area Indian Health Board, Aberdeen Tribal Chairmen’s Health Board, Fond Du Lac Reservation, and Cherokee Nation) and all 50 states, the District of Columbia, and several U.S. territories. The National Partnership for Com- prehensive Cancer Control, a group of national orga- nizations that includes NAACCR, the Intercultural Cancer Council, and C-Change, now promotes com- prehensive cancer control at federal, state, local, tribal, and territorial levels; they also support Com- prehensive Cancer Control Leadership Institutes to improve implementation of cancer plans in tribal communities. Nonprofit organizations such as ACS and the Lance Armstrong Foundation are providing new resources to AI/AN cancer control efforts. The ACS program Circle of Life is designed to decrease the breast cancer incidence and death rates among American Indian women. The Lance Armstrong Foundation has developed a strategic plan for AI/AN survivors to help overcome the fatalism that often accompanies a diagnosis of cancer in the community setting. All these efforts and more are essential to reducing the burden of cancer in AI/AN populations.
Issues in Data Interpretation The cancer surveillance infrastructure in the U.S. now provides cancer incidence data for most of the population by site, sex, race, and Hispanic origin. However, several limitations in the data collection and analyses may influence the interpretation of results in this report.
First, the 2 methods that were used for trend analysis—joinpoint and fixed-interval trends—some- times appear to yield different results. The joinpoint method is used for long-term trends but also has the flexibility to identify changes in magnitude and direction of the trend within shorter intervals. The fixed-interval method facilitates comparison between groups but lacks the flexibility to identify changes in the trend over the fixed interval. Although they were not presented, when joinpoint analyses were applied to the fixed-interval time period, the results were more comparable to the long-term trends despite the differences in the U.S. population coverage of 59% for the fixed-interval analyses and 10% for the long- term joinpoint analyses and despite the lack of delay adjustment in the fixed-interval approach.
Second, as discussed earlier, it was discovered that many AI/AN persons were misclassified as another race in cancer registry data, and the extent of misclassification varied by registry, ranging from 36% in Minnesota to 57% in California in published
Although linkage with the IHS
patient registration database improves the race clas-
Report to the Nation, 19752004/Espey et al.
sification for AI/AN cases, the issue is not resolved completely, because AI/AN persons who are not members of the federally recognized tribes and are not eligible for IHS services are under-represented in the IHS database. Because the denominators used in the rate calculations were derived from the U.S. Cen- sus estimates, which are based on self-identification of race (because it is preferred as the most accurate classification), the actual rates for AI/AN persons may be higher than those reported in our study.
Third, the analyses presented here for AI/AN populations are based on residents of CHSDA coun- ties and exclude many AI/AN residents in urban areas that are not included in the CHSDA counties; therefore, the findings may not represent all AI/AN populations in the U.S. or in individual IHS regions (see Supplementary Table 1B; available at URL: www.seer.cancer.gov/report_to_nation/). In particu- lar, the East region includes only 12.2% of the total AI/AN population in that region.
Fourth, the current analysis revealed less varia- tion for NHW populations than for AI/AN popula- tions by IHS regions using data from CHSDA counties only. This approach may obscure variation in cancer rates for NHW populations by alternate state or county groupings.
Finally, the median BRFSS response rate is approximately 50%, the response rates vary widely by state, and the surveys rely exclusively on telephone interviews.35 Telephone coverage also varies within the AI/AN population and is lower in this population compared with other racial/ethnic groups.35 Gener- ally, the questions regarding cancer screening tests do not distinguish between tests conducted for screening and tests conducted for diagnostic pur- poses. Mixed-mode sampling approaches (ie, com- bining random digit dialing, computerized address databases, web surveys, etc) are being investigated to address concerns about nonresponse and the declin- ing coverage of the population with traditional land- line household phones. 160
Future Directions Each year, the annual report to the nation on cancer brings together key agencies and organizations engaged in cancer surveillance in the U.S. to describe the latest cancer incidence and mortality data for the general population and to focus on a topic that war- rants special attention. This year’s report represents the 10th anniversary of this collaboration and pro- vides improved data for AI/AN, highlighting the re- gional burden of cancer and the distinctive patterns of specific cancers in this population.