Validation Estimates from Early Reporting of Incidence Data Adjusted for Case Undercounts: Selected SEER Registries

To validate the use of the February 2026 submission data for predicting the eventual rates and trends through 2024, data from one year earlier are used. In this case both the February and subsequent November submissions are both available. The delay-adjusted rates and trends for cases diagnosed through 2023 are derived from the February 2025 submission and compared to those from the November 2025 submission. The comparisons are based on: (1) the delay-adjusted rates for 2023 and (2) the direction and statistical significance of the final Joinpoint segment¹. To compare the 2023 delay-adjusted rates from the November and February 2025 submissions, the ratio of these two rates is computed for each of the cancer site/sex combinations. Ideally the distribution of these ratios should be centered around 1, with the variability as small as possible.

Validation Summary

Three validation metrics are used to summarize the results.

The first metric is the number of series where the statistical significance of the final Joinpoint segment changed (or the direction of the final segment changed from trending significantly upward to significantly downward or vice-versa), when comparing the February and November 2025 submissions. Of the 42 series (18 cancer sites by sex, and 6 single sex cancer sites) only two changes of this type were observed.

The second metric compares the 2023 delay-adjusted rates from the November and February 2024 submissions. The ratio of these two delay-adjusted rates is computed for each cancer site/sex combinations. Ideally, the distribution of these ratios should be centered around 1, with minimal variability. For the twelve SEER 21 registries where the February submission meets the selection criteria, the median of these 42 ratios for the delay-adjusted rates for cases diagnosed in 2022 is 0.992. In contrast, the median of these 42 ratios for the observed (i.e., non-delay-adjusted) rates is 1.061, meaning that the observed rates for the November 2025 submission are, on average, about 6.1% larger than those based on the February 2025 submission. The delay adjustment process successfully aligns the 2023 rates from these two submissions. Note that the delay-adjusted rates obtained from the November 2025 submission are based on the Vintage 2024 population, while the delay-adjusted rates obtained from the February 2025 submission are based on the Vintage 2023 population. When both use the Vintage 2024 population, the median of these 42 ratios for the delay-adjusted rates for cases diagnosed in 2022 is 0.996, which is even closer to 1.

The validation table also includes the trends for cases diagnosed from 2000-2023, based on the November 2025 submission for the same twelve registries. A change in the statistical significance of the final Joinpoint segment—or a reversal in its direction (e.g., from significantly upward to significantly downward, or vice versa) between the 2000–2022 November 2024 submission and the 2000–2023 February 2025 submission—indicated that preliminary incidence estimates were suggesting a trend shift for the most recently diagnosed cases in 2023. The third validation metric evaluates whether this change detected in the preliminary estimates was supported by the results from the November 2025 submission. The table below displays 9 instances (out of the 42 total cancer site and sex combination) where a new trend was suggested by the preliminary estimates. These projected trends were borne out in 8 of the 9 cases (highlighted in green), and were not verified in only 1 case (highlighted in yellow).

Based on these results using the February 2025 and November 2025 submissions for 2000-2023, a decision was made to utilize registries where the February 2026 submission met the 95% completeness criteria. This was done to produce preliminary rates and trends for cases diagnosed between 2000 and 2024. Additionally, included registries were required to have sufficient February submissions to conduct delay modeling, and a ratio of case counts from their initial February and November submissions greater than 0.90 (averaged over the most recent six November submissions). The standard deviation of these six ratios had to be less than 0.05.

After applying all of these criteria, the preliminary estimates from the February 2026 submission for cases diagnosed from 2000-2024 included the following 11 registries: Alaska Natives, Connecticut, Greater Bay, Hawaii, Illinois, Iowa, Louisiana, New Jersey, New York, Seattle, and Utah. These registries represent approximately 22% of the U.S. population and 47% of the population within the Official Release of SEER 21registries. These criteria for including registries in preliminary estimates will continue to be evaluated in future years and are subject to change.

¹Joinpoint is the methodology used to characterize trends in cancer rates using joined linear segments on a log scale, where the locations where the trends change is determined by a data driven algorithm. Because the trends are modeling on a log scale, the trend is each segment is characterized using as an Annual Percent Change (APC), and the change over any user defined period is characterized as an Average Annual Percent Change (AAPC). See the Joinpoint web pages for more details.