What Can Colorado’s Cancer Registry Tell Us About Rocky Flats — and What Can’t It Tell Us?

The Colorado Department of Public Health and Environment has stated that it has monitored cancer incidence near Rocky Flats and has not found elevated rates among residents living near the area. But who does “residents living near the area” actually mean? Does it mean people who lived near Rocky Flats during plant operations, the 1957 and 1969 fires, or releases associated with Pad 903? Or does it mean people whose address happened to be near Rocky Flats when they were diagnosed with cancer? Those are very different populations.

Rocky Flats Downwinders is asking CDPHE to explain this distinction whenever it reports the findings of its Rocky Flats cancer studies. We are not asking the Cancer Registry to determine whether Rocky Flats caused an individual person’s cancer. Cancer-registry data cannot make that determination.

We are asking a more fundamental question:

Are the methods being used capable of answering the historical-exposure question that many community members believe is being answered?

The key limitation that should be disclosed

Cancer registries generally classify a cancer case geographically according to the person’s residence at the time of diagnosis. That approach is useful for monitoring current cancer incidence and identifying broad geographic patterns. However, an address at diagnosis is not necessarily an address during a potentially relevant exposure period. Cancer can develop many years or decades after an exposure. During that time, people move.

Someone who grew up near Rocky Flats during its operating years may later move to another Colorado community or another state. If that person develops cancer decades later, the diagnosis generally will not be counted as a cancer occurring among current residents near Rocky Flats.

The opposite problem can also occur. Someone may move into a community near Rocky Flats shortly before being diagnosed. That person can be counted as a local cancer case even though the exposures most relevant to the cancer may have occurred somewhere else.

Therefore, a study of people living near Rocky Flats when diagnosed is not the same as a study of people who lived there during historically relevant exposure periods.

Every public summary of the CDPHE findings should clearly disclose this distinction.

Why this matters

Population movement can produce two important classification problems:

• Historically exposed residents who move away may be excluded from the local cancer count.

• Recent arrivals may be included even though they were not present during Rocky Flats operations or major historical releases.

  
  

As more time passes, residence at diagnosis may become an increasingly weak substitute for residence during the period of potential exposure. This affects how negative or statistically non-significant findings should be explained.

A non-significant result could mean that:

• the observed cancer rate was not higher than expected;

• the observed rate was higher, but the estimate was too imprecise to reach statistical significance;

• no consistent elevation appeared across the cancers, geographic areas or periods studied; or

• the study lacked sufficient statistical power to detect anything except a very large increase.

  
  

These possibilities do not all mean the same thing.

A failure to detect a statistically significant elevation does not automatically demonstrate that cancer incidence was normal among everyone who may have been historically exposed.

Other methodological limitations

Residence at diagnosis is the central concern, but it is not the only limitation that should accompany public descriptions of the studies.

Geographic boundaries

CDPHE’s Regional Statistical Areas include communities extending beyond the principal contamination patterns shown on historical maps associated with the fires and Pad 903.

Environmental contamination does not follow municipal, county, ZIP-code or statistical-area boundaries. Combining areas with different distances, directions and contamination histories can make a localized pattern more difficult to detect.

This does not prove that CDPHE’s broad boundaries changed the results. It means that the possibility should be tested using exposure-informed boundaries based on historical releases, wind direction, soil measurements, drainage and distance from Rocky Flats.

Small numbers and statistical power

Rare cancers, childhood cancers and cancers diagnosed at unusually young ages often involve small numbers of cases. Small numbers produce wide confidence intervals and make statistical conclusions uncertain. When a study lacks statistical power, the absence of a statistically significant result may mean that there were too few observations to distinguish an actual increase from random variation.

CDPHE should report the observed and expected case counts, incidence ratios and confidence intervals—not only whether a result crossed a statistical-significance threshold.

Combining different cancers and age groups

“All cancers” is not one disease. Different cancers have different causes, risk factors and latency periods. Combining cancers can conceal a site-specific pattern. Broad age categories can similarly obscure patterns involving children, adolescents or adults diagnosed unusually young. Public statements should identify exactly which cancers, ages and periods were examined.

Changing populations and denominators

Cancer-rate calculations depend on reliable population estimates. ZIP codes, census tracts and municipal boundaries can change over time. Annexation and rapid development can also alter the number and age distribution of residents. These changes affect the expected number of cancers and complicate comparisons across decades.

Ecological study design

Area-level cancer studies compare groups of people rather than linking each person’s diagnosis with that person’s complete residential and exposure history. They generally cannot determine individual radiation dose or fully account for occupational exposures, smoking, medical radiation, screening practices, health-care access and other cancer risk factors. This limits causal interpretation even when the registry’s cancer counts are accurate.

Candelas illustrates the population-turnover problem

The development of Candelas beginning in the early 2010s provides a clear example of why current residence and historical residence should not be treated as equivalent.

Communities surrounding Rocky Flats grew substantially between 1980 and 2020:

Community	1980 population	2020 population	Approximate growth
Arvada	84,576	124,402	47%
Broomfield	20,730	74,112	258%
Westminster	50,211	116,317	132%
Superior	About 250	13,094	More than 5,000%

These municipal figures demonstrate major regional change. They do not independently prove individual migration, exposure or cancer risk. Annexation and changing municipal boundaries also contributed to some of the growth.

The change is even more visible in ZIP Code 80007, which includes Candelas. Census-derived counts show its population increasing from 7,262 in 2010 to 17,997 in 2020—approximately 148 percent. Many people added during this period were not living there during Rocky Flats plant operations.

CDPHE’s principal cancer-incidence update ended in 2014. A family that moved to Candelas in 2012 could have contributed no more than approximately two years to that analysis.

That period is generally inadequate for evaluating most solid cancers potentially associated with ionizing radiation. Epidemiological analyses commonly apply minimum latency periods beginning around 10 years for many solid cancers, and relevant effects may emerge decades later. Some leukemias can have shorter latency, which is why latency should be specified separately for each cancer type.

Candelas also includes newer homes and many families with children. These characteristics make the community important for prospective health surveillance. But they do not establish where individual residents previously lived, whether they knew about Rocky Flats or whether they experienced a particular exposure. Those questions require residential histories or representative surveys.

Population growth does not automatically reduce a standardized cancer-incidence ratio. Expected cancer counts should also rise with population, and age standardization can account for differences in age structure.

The stronger concern is population composition: as longtime residents leave and newer residents arrive, the population living near Rocky Flats at diagnosis becomes less representative of the population that lived there during historical exposure periods.

What CDPHE should report

Rocky Flats Downwinders is asking CDPHE to state clearly:

• Whether “residents living near the area” means residence at diagnosis or residence during a historically relevant exposure period.

• The observed and expected case counts, standardized incidence ratios, confidence intervals and statistical-significance status for each cancer, age group, geographic area and period.

• Whether any cancer rates were higher than expected but statistically imprecise.

• How the study boundaries compare with historical contamination maps associated with the fires and Pad 903.

• Whether the analyses incorporated residential history, migration, year of arrival, duration of residence or cancer-specific latency.

• How small numbers, confidentiality restrictions, population growth, annexation and uncertain historical denominators affected the analysis.

  
  

Recommended disclosure language

CDPHE could provide the public with a more accurate understanding of its findings by including language such as:

For the cancer sites, age groups, geographic areas and calendar periods analyzed, CDPHE did not identify a consistent statistically significant pattern of higher incidence among people classified as residents near Rocky Flats at the time of diagnosis. Because the analyses did not reconstruct complete residential histories, follow all former residents or fully account for migration and cancer-specific latency, they do not determine cancer incidence among every person who lived near Rocky Flats during historically relevant exposure periods.

Our request to CDPHE

Rocky Flats Downwinders is asking CDPHE to publish a technical clarification connecting each public conclusion to the population, geography, cancers, ages and calendar periods actually analyzed. That clarification should accompany previous and future statements that CDPHE has not found elevated cancer rates near Rocky Flats.

This request does not dismiss the value of the Colorado Cancer Registry. Registry surveillance is an important public-health tool. But the results should not be presented as though an area-based analysis using residence at diagnosis were a historical cohort study of everyone who lived near Rocky Flats. Clear disclosure would help the public understand both the value of the studies and their limits.

Sources

• Colorado Central Cancer Registry. Ratios of Cancer Incidence in Ten Areas Around Rocky Flats, Colorado, Compared to the Remainder of Metropolitan Denver, 1980–1989, with Update for Selected Areas, 1990–1995. 1998.

• Colorado Department of Public Health and Environment. Rocky Flats cancer-incidence update covering 1990–2014 and related public summaries.

• North American Association of Central Cancer Registries. Address-at-diagnosis data standards.

• National Academies of Sciences. Health Risks from Exposure to Low Levels of Ionizing Radiation: BEIR VII Phase 2. 2006.

• U.S. Census Bureau. 1980 and 2020 decennial Census population counts for Colorado communities.