Cancer stage at diagnosis, treatment, and survival among American indians and non-American Indians in Montana


  • Terry D. Dennis M.D., M.P.H.

    Corresponding author
    1. Billings Area Indian Health Service, Billings, Montana
    • Suite 120W, 2900 12th Avenue North, Billings, MT 59101.
    Search for more papers by this author

  • The opinions expressed in this article are those of the author and not the Indian Health Service.



The intent of the current study was to ascertain whether differences in cancer survival between Montana non-American Indians (non-AI) and Montana American Indians (AI) were related to differences in stage of disease at diagnosis or in the type of treatment received.


A case-control design was utilized using data from the Montana Central Tumor Registry and the Indian Health Service medical records. AIs diagnosed between January 1, 1984 and December 31, 1993 were the cases in the study, and non-AIs diagnosed in the same period were the controls. Chi-square tests and life table techniques were used to analyze the data.


Five hundred twenty-two cases were matched with controls. The 5-year cancer survival rate for AIs was 36% and was 47% among non-AIs. The stage at the time of diagnosis was local in 34% of AIS and 36% of non-AIs. The stage was regional in 30% of AIs and 26% of non-AIs. Distant disease at the time of diagnosis was present in 25% of AIs and 24% of non-AIs, whereas an unknown extent of disease was present in 11% of AIs and 14% of non-AIs. AIs underwent surgery less frequently than non-AIs (79% vs. 86%), but this did not appear to contribute to the survival differences observed.


The survival differences observed in the current study cannot be explained easily by differences in the cancer stage at diagnosis or the type of treatment received. Cancer 2000;89:181–6. © 2000 American Cancer Society.

The 1989–91 age-adjusted malignant neoplasm mortality rate among American Indians in Montana and Wyoming was higher than that of U.S. All Races rate1 even though previous studies in various populations of American Indians have indicated a lower overall incidence of malignant neoplasms.2, 3 Previous studies also have indicated a lower survival rate among native populations.4–7 The study by Bleed et al. indicated a significantly poorer survival rate among Montana American Indians when compared with U.S. whites.4 He hypothesized that the rate of advanced disease at diagnosis or the use of different treatment modalities may explain the observed differences in survival. Others who have observed a lower survival rate among American Indians also have speculated that the differences are due to differences in stage at diagnosis or treatment received.8

The primary intent of this study was to explore two potential hypotheses that could explain the survival differences observed by Bleed. The first hypothesis is that stage at diagnosis was less favorable in the Montana American Indians. The second hypothesis is that differences in treatment may explain the survival differences reported by Bleed et al. This study uses matching with controls from the same geographic area to address the possibility that the observed survival difference represents a regional or urban/rural survival difference because U.S. whites are more likely to be in an urban environment than Montana American Indians.


A case-control model was used for the study. Montana was divided into four regions by using county boundaries that corresponded to Indian Health Service Unit boundaries. The use of regions was to facilitate matching of American Indian cancer cases with non-American Indians who resided in close proximity. Indian Health Service in Montana provides primary care services on the reservations and operates three small inpatient facilities that provide limited surgical services. Oncology services generally are referred to larger private facilities within the state. Montana's sparse population and great distances greatly limit the options for oncology referral for Indians and non-Indians alike. An informal canvas in 1999 indicated approximately 26 medical oncologists and six radiation oncologists in the state. Given the limited options for referral, matching within regions also should control somewhat for different sources of referral oncology care. Three regions included six reservations that had extensive medical records available in Indian Health Service facilities on Indian Health Service beneficiaries. The fourth region included all other Montana counties. One reservation is included in the fourth region because limited Indian Health Service facilities serve the population in this region. This article includes data only from the three regions with Indian Health Service records.

Case definition included residence in one of the three reservation containing regions at the time of diagnosis, and diagnosis of a malignant neoplasm between January 1, 1984 and December 31, 1993. All in situ diagnoses and nonmelanoma skin carcinomas were excluded.

The estimated 1993 Indian Health Service Unit population in the three regions was 35,649.9 Because the study population was small, attempts were made to ascertain all incident malignant neoplasms among American Indians residing in the target counties at the time of diagnosis. The initial phase of case ascertainment was to search the Indian Health Service (IHS) Resource Patient Management System (RPMS) for all patients having a problem or purpose of visit coded between ICD-9 140.0 and 208.9. The RPMS is a digital abstract of all IHS provided outpatient and inpatient visits. The charts of all cases identified in this manner were reviewed to ensure that they met the case definition. The third step was to search the Montana Central Tumor Registry to identify American Indian cancer cases not identified in the IHS database. The two databases were then merged to create a single data set.

Controls were selected from all non-American Indian patients with malignant neoplasm identified in the Montana Central Tumor Registry who were residents in the study counties at the time of diagnosis. The control's medical records were not accessible for this study. The controls were matched to the cases by using site of primary tumor, gender, age at diagnosis, and date of diagnosis. Age at diagnosis was considered a match if the control was within 5 years of the age of the case. Date of diagnosis was considered a match if the control's diagnosis was within 5 years of the case's diagnosis.

Cancer stage at diagnosis was classified into the four categories utilized by the Montana Central Tumor Registry: 1) in situ (excluded from this study), 2) local (invasive cancer confined to the organ of origin), 3) regional (spread to adjacent organs by direct extension or to regional lymph nodes), and 4) distant (metastases to distant organs or lymph nodes). In a pilot study, we attempted to find information based on the International Union Against Cancer TNM Classification of Malignant Tumors to determine stage at diagnosis, but the information was too scanty to allow analysis.

Treatment information was obtained from the Montana Central Tumor Registry records for both cases and controls. Treatment included any cancer treatment received subsequent to the diagnosis of cancer. Data from the Indian Health Service chart were used to determine treatment received by those patients not in the Montana Central Tumor Registry. Treatment modality was classified as surgical, radiation, chemotherapy, hormonal, or biologic modifiers. If a patient received multiple modalities, each individual modality was used in the analysis. No data defining if a treatment had been initiated but not completed were available for this study.

Montana vital statistic information is linked to the Montana Central Tumor Registry to ascertain vital status. Cases identified only in the IHS medical record had vital status assessed by review of these records.

Statistical Methods

Abstracted data were collated and analyzed in Epi-Info,10 Computer Programs for Epidemiologic Analysis: PEPI,11 and Microsoft Excel. Categoric data were analyzed using chi-square. Life table analysis was done according to the technique of Kaplan and Meier by using Computer Programs for Epidemiologic Analysis.


During the study period, 570 cases were identified that met the case definition. Matching was successful in 522 cases. The mean age of the American Indians was 62.3 years and the non-American Indians 62.2 years (P = 0.36). The most common tumor types in the matched cases were lung (19.3%), breast (16.8%), prostate (12.0%), colon (10.2%), kidney (5.6%), and oropharynx (5.6%). The forty-eight unmatched cases included 32 women. Eighteen unmatched individuals were younger than 40 years. Thirty-three of the 48 were younger than 60 years. Two were missing critical information on location or histology of the tumor. Stomach carcinoma occurred in eight of the unmatched cases, cervical carcinoma in seven cases, and four were sarcomas.

No difference was found in stage of disease at diagnosis between American Indians and non-American Indians in the study regions. Table 1 describes these findings. Localized disease was found in 34% of cases versus 36 % of controls; regional dissemination in 30% of cases versus 26% of controls; and distant dissemination in 25% of cases versus 24% of controls. No stage could be determined for 11% of American Indians versus 14% of non-American Indians. These differences are not statistically significant (P = 0.32). Stage at diagnosis for the lung, breast, colon, and prostate carcinoma was explored to identify any subgroup variation. Table 1 also details the stage at diagnosis for these tumors. Differences in colon carcinoma stage at diagnosis were the closest to statistical significance.

Table 1. Stage at Diagnosis for Common Neoplasms, Montana American Indian Cases and Non-American Indian Controls, 1984–1993
SiteNumberLocal (%)Regional (%)Distant (%)Unknown (%)P value

Surgery was performed in most cases and more often among non-American Indians (86%) than American Indians (79%). This difference, although relatively small in absolute terms, was statistically significant (P = 0.02). Treatment received is detailed in Table 2. A single individual could and frequently did receive more then one modality during the treatment course. The difference in the proportion of patients who received radiation therapy was not statistically significant (P = 0.10). Chemotherapy, hormonal manipulation, and biologic modifiers were used in almost identical numbers. These treatment categories included patients who received single modality treatment and patients who received these therapies in a multimodality regimen. The difference in the number of patients who received “no treatment” was not statistically significant.

Table 2. Treatment Modalities Used by American Indian and Non-American Indian Cancer Patients
TreatmentAmerican Indian (%)Non-American Indian (%)P value
No treatment1080.33

Kaplan–Meier life table survival analysis for all tumor types showed American Indians 5-year survival was 36%, and 5-year survival for non-American Indians was 47% (P < 0.01). One-year survival was 65% among American Indians and 73% among non-American Indians. Three-year survival was 46% among American Indians and 55% among non-American Indians. Table 3 demonstrates that survival is lower among American Indian patients in the four most common neoplasms. However, because of the small numbers, only prostate carcinoma achieves statistical significance. The four most common neoplasms account for 58% of the total number of matched cases.

Table 3. Kaplan–Meier 5-Year Survival for Montana Cancer Patients, 1984–1993
SiteAmerican Indian (%) (n)Non-American IndianP value
Lung5 (101)10 (101)0.21
Breast68 (86)79 (86)0.28
Colon38 (53)55 (53)0.15
Prostate39 (63)63 (63)< 0.01
Total36 (522)47 (522)< 0.01

Seventy-two (14% of cases) of the 325 deaths in the American Indian patients were coded as a noncancer death. Sixty-two (12% of controls) of the 292 deaths among the non-American Indians were coded as noncancer deaths. This was not statistically significant (P = 0.78).

Life table analysis was used to explore the impact of the small difference in the rates of surgery between American Indians and non-American Indians. Five-year survival for the American Indians without surgery was 27% whereas in the non-American Indians it was 34% (P = 0.11). Five-year survival among the American Indians who received surgery was 42% whereas in the non-American Indians it was 50% (P = 0.003)


Malignant neoplasm survival rates for American Indian populations have been shown to be lower then the general population.4–7 Young et al. found lower cancer survival among American Indians in the SEER program from 1973 to 1979.7 Bleed et al.,4 found decreased malignant neoplasm 5-year survival rates for Montana and New Mexico American Indians when compared with U.S. whites. Whether the decreased survival reflected later stage of diagnosis, reduced access to optimal oncologic care, or poorer response to treatment was not studied. Consistent with the findings of the Bleed study, Samet et al.,5 showed that 5-year survival, after adjustment for stage of diagnosis and treatment, was lower among American Indian cancer patients living in Arizona and New Mexico than non-Hispanic whites in New Mexico. Samet et al.,5 did observe a later stage at diagnosis in New Mexico and Arizona, but the 5-year survival difference persisted after adjustment for stage at diagnosis. This Montana study indicated a similar stage at diagnosis for American Indians and non-American Indians, but consistent with other studies a lower survival rate was observed for American Indians.

This study demonstrated a statistically significant decrease in survival but no difference in cancer stage at diagnosis between Montana American Indians and non-American Indians. More non-American Indians received surgical treatment. This difference although statistically significant was relatively small in absolute terms (86% vs. 79%). The difference does not appear to have contributed to the survival differences observed. The American Indian and non-American Indian patients who did not have surgery had a difference in survival rate that was similar to the difference in survival rate between the American Indian and non-American Indian patients who did have surgery. However, the overall survival was lower in the nonsurgery group for both American Indians and non-Indians.

The prostate carcinoma 5-year survival was significantly lower among American Indians. Five-year survival in breast, lung, and colon carcinoma cases tended toward lower survival among American Indians but did not achieve statistical significance. The small number of cases certainly may have obscured significant differences in survival among the other cancers. The tendency to a lower survival rate in each of the four most common cancers indicates the overall difference in survival is not due to lower survival rates in any single cancer.

Case ascertainment during this study included an examination of not only tumor registry cases but also cases identified through Indian Health Service records. A previous internal Indian Health Service study12 of Montana American Indian cancer surveillance in the latter half of the 1980s indicated that 75 (16%) of 455 cases were identified only through the Montana Central Tumor Registry. One hundred and fifteen cases (25%) were identified only through Indian Health Service records, and 265 (58%) were identified in both sources. During this study, a reexploration of this issue on 3 reservations showed that 33% of the cases were identified only in the Montana Central Tumor Registry, 27% only through Indian Health Service records, and 40% in both sources.

This study attempted to reduce variations in available cancer care and local environmental factors that may influence cancer care by matching patients within geographic regions of a state. Incidence and survival comparison of distinct tribes or pooled American Indian populations to a large standard population has several inherent problems. Comparisons are affected by differences in access to primary care physicians, access to oncology care, and environmental factors. Comparison using a control population from a similar geographic location ameliorates several of these variables. First, environmental differences such as a rural versus urban locale are diminished in importance. Second, the effect of variability in access to specialists and treatment recommendations is somewhat reduced, especially when the number of specialists is small as it is in Montana.

Our study attempts to control for the above variables. Cases and controls were from the same regions of the Montana. The specialists were small in number. American Indians had access to both IHS and non-IHS physicians although non-American Indians did not have access to IHS physicians. Specific measures of accessibility and barriers to primary or specialty care were beyond the scope of this study.

The cases in which matching was unsuccessful appeared to be dominated by younger patients. Cervical carcinoma and stomach carcinoma may have a higher incidence in American Indians of the Northern Plains.13 Other differences in cancer epidemiology between the two groups could influence the success of matching. The small population of Montana also may influence the success finding suitable controls, especially for cases with unusual cancer type or younger age. The exclusion of these cases should not influence the overall thrust of this article but would limit the use of the descriptive statistics in estimating the incidence and distribution of cancer among Montana American Indians.

This study involved American Indians in one region of North America. The study is limited by the use of medical record and tumor registry data. The lack of access to medical records of the controls and the use of central tumor registry data represent an inherent weakness of the study. Direct examination of the medical records for both cases and controls certainly would strengthen these findings. However, the resources available precluded a study of that intensity. Appreciation of the limitations of the data sources prevents consideration of a study such as this as the definitive work on these questions. However, the use of relatively accessible data to explore these issues is prudent. Given the limitations on resources, this type of study may be one of the few avenues to explore these questions. Perhaps efforts such as this study will provide some of the groundwork necessary for future studies that may have more resources to improve and expand the available data and delve into the many unanswered questions in greater detail.

The tumor registry definitions of stage and type of treatment provided are global and qualitative in nature. This may decrease the ability to identify important differences that may have influenced survival. Efforts to upgrade and improve the data available to central tumor registries may allow for more specific information in the future. The thoroughness of staging procedures and the intensity of treatment was not addressed in this study. Further exploration of the impact of treatment on the survival differences probably will demand higher quality data at a greater level of detail. The relatively small population limits our ability to study tumors by primary site that may offer important insights. The diversity of cultural, economic, and behavioral (e.g., cigarette smoking) factors among the hundreds of tribes in the U.S. is a limitation of this or any other regionally based study. However, this diversity also undermines the utility of any study that is national in scope.

If the observation in this study that American Indians' cancer survival was less than their neighbors despite a similar stage at diagnosis and, on superficial assessment, treatment is borne out by future studies, then major gains in early diagnosis and treatment, although valuable, may not eliminate the observed survival differences.

The impact of comorbidity may be a significant factor in the observed survival differences. Deaths from other conditions were slightly higher among American Indian cases. This may explain a small portion of the survival differences. Poorer general health status also may contribute to an increase in cancer attributed deaths as well. If control for socioeconomic differences between the American Indians in Montana and the non-American Indians could be conducted, perhaps some of the survival differences would disappear. Other potential questions of interest include the impact of any differences in staging intensity, rates of completion of therapy and the influence of follow-up for recurrent disease. Studies should continue to be done in other North American regions, using various data sets, study designs, and pursuing other hypotheses that may help us further understand the observed survival differences.


The author thanks the Indian Health Service Office of Planning, Evaluation, and Legislation for the funding to conduct this study. This study would not have been possible without the invaluable assistance of Christine Brusco, M.D., Lorraine Allegro, M.D., Carol Luckett, R.N., and Deborah Hellhake in the collection of data and providing valuable insights in the design and conduct of this study.