Skip to main content

Advertisement

Log in

Projecting the number of patients with colorectal carcinoma by phases of care in the US: 2000–2020

  • Original Paper
  • Published:
Cancer Causes & Control Aims and scope Submit manuscript

Abstract

Objective

This study provides projections of colorectal cancer prevalence by phases of care (initial, monitoring, and last year of life) to the year 2020 and describes the estimation method.

Methods

Cancer prevalence by phase of care was estimated from colorectal cancer incidence and survival from the Surveillance, Epidemiology, and End Results (SEER) Program data, population estimates and projections from the US Census Bureau, and all cause mortality data from the Human Mortality Life Tables. Assumptions of constant incidence and survival were used for projections from 2000 to 2020. Modeled and directly observed patient months by phase of care were compared for 1996 −1998 to provide validation of estimates.

Results

Prevalence of colorectal cancer is estimated to increase from 1,002,786 (0.36%) patients to 1,522,348 (0.46%) patients between 2000 and 2020. The estimated number of person-months in the initial and last year of life phases of care will increase 43%, while the monitoring phase of care will increase 54%. Modeled person-months by phase of care were consistent with directly observed measures of person months by phase of care in 1996–1998.

Conclusions

Under assumptions of current cancer control strategies we project that colorectal cancer prevalence will increase more rapidly than the US population, largely due to the aging of the US population. This suggests that considerable resources will be needed in the future for initial, continuing and last year of life treatment of colorectal cancer patients unless notable breakthroughs in primary prevention occur in the future years.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5

Similar content being viewed by others

References

  1. Brown ML, Riley GF, Potosky AL, Etzioni RD (1999) Obtaining long-term disease specific costs of care: application to medicare enrollees diagnosed with colorectal cancer. Med Care 37(12):1249–1259

    Article  PubMed  CAS  Google Scholar 

  2. CDC. Cancer Survivorship – United States (1971–2001) MMWR 6-25-2004; 53(24):526–529

  3. SEER Cancer Statistics Review (1975–2002) Ries LAG, Eisner MP, Kosary CL, et al. (2005) National Cancer Institute (http://seer.cancer.gov/csr/1975_2002).

  4. Garcia-Rodriguez LA, Huerta-Alvarez C (2001) Reduced risk of colorectal cancer among long-term users of aspirin and nonaspirin nonsteroidal antiinflammatory drugs. Epidemiology 12(1):88–93

    Article  PubMed  CAS  Google Scholar 

  5. Pickhardt PJ, Choi JR, Hwang I, et al. (2003) Computed tomographic virtual colonoscopy to screen for colorectal neoplasia in asymptomatic adults. N Engl J Med 349(23):2191–2200

    Article  PubMed  CAS  Google Scholar 

  6. Cunningham D, Humblet Y, Siena S, et al. (2004) Cetuximab monotherapy and cetuximab plus irinotecan in irinotecan-refractory metastatic colorectal cancer. N Engl J Med 351(4):337–345

    Article  PubMed  CAS  Google Scholar 

  7. Poynter JN, Gruber SB, Higgins PD, et al. (2005) Statins and the risk of colorectal cancer. N Engl J Med 352(21):2184–2192

    Article  PubMed  CAS  Google Scholar 

  8. Gatta G, Capocaccia R, Berrino F, Ruzza MR, Contiero P (2004) The EUROPREVAL working group. Colon cancer prevalence and estimation of differing care needs of colon cancer patients. Ann Oncol 15(7):1136–1142

    Article  PubMed  CAS  Google Scholar 

  9. Mariotto A, Warren JL, Knopf KB, Feuer EJ (2003) The prevalence of patients with colorectal carcinoma under care in the US. Cancer 98(6):1253–1261

    Google Scholar 

  10. U.S. Census Bureau, Population Division. Interim projections consistent with Census 2000 (released March 2002) http://www.census.gov/population/www/projections/popproj.html.2002

  11. Verdecchia A, De Angelis G, Capocaccia R (2002) Estimation and projections of cancer prevalence from cancer registry data. Stat Med 21(22):3511–3526

    Google Scholar 

  12. Goldman AI (1984) Survivorship analysis when cure is a possibility: a Monte Carlo study. Stat Med 3(2):153–163

    PubMed  CAS  Google Scholar 

  13. De Angelis R, Capocaccia R, Hakulinen T, Soderman B, Verdecchia A (1999) Mixture models for cancer survival analysis: application to population- based data with covariates. Stat Med 18(4):441–454

    Google Scholar 

  14. Mariotto A, Capocaccia R, Verdecchia A, et al. (2002) Projecting SEER cancer survival rates to the US: an ecological regression approach. Cancer Causes Control 13(2):101–111

    Article  PubMed  Google Scholar 

  15. Human Mortality Database. University of Californian, Berkeley (USA), and Mark Planck Institute for Demographic Research (Germany). Avaialable at http://www.mortality.org (data downloaded on August 2002)

  16. Cronin KA, Feuer EJ (2000) Cause-specific mortality for cancer patients in the presence of other causes: a crude analogue of relative survival. Stat Med 19:1729–1740

    Article  PubMed  CAS  Google Scholar 

  17. Winawer S, Fletcher R, Rex D, et al. (2003) Colorectal cancer screening and surveillance: clinical guidelines and rationale-update based on new evidence. Gastroenterology 124(2):544–560

    Article  PubMed  Google Scholar 

Download references

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Angela B. Mariotto.

Appendix

Appendix

Survival model used in PIAMOD to estimate prevalence

The survival model used in PIAMOD assumes a proportion φ(i) of patients, dependent on age at diagnosis i, who are cured from the cancer and die with the same mortality rate as the general population. The remaining have a relative survival following a Weibull distribution with parameters β(i) and λ(i). Mathematically, the model specifies that for patients in the birth cohort t − i, the probability of surviving age i + d for patients diagnosed at age i and year t is given by,

$$ S(i,t,d)={\left\{{\varphi(\hbox{i})+[1-\varphi (\hbox{i})]}\exp-\left[{\lambda\left(i\right)d} \right]^{\beta (i)} \right\}^{\exp\left[{-\alpha_2(t-t_0)-\alpha_3} \right]}}, $$
(2)

where exp [ −α2] is the annual relative risk of cancer death for patients diagnosed in calendar year t + 1 with respect to period t and exp[ −α3] a relative risk of death of US patients with respect to SEER patients. The parameters φ(i), λ(i), β(i) and α2 are estimated in CANSURV (http://srab.cancer.gov/cansurv/) a statistical software to analyze population-based survival data. Parameters α3 were estimated by using an ecological regression model to the SEER relative rates and extrapolating to the US [14]. More specifically, the association of demographic-socio-economic (ecological) variables at county level on cancer relative survival rates are estimated in SEER and then extrapolated to the US. For all races, the relative risks of cancer death in US with respect to SEER were 1.12 for breast, 1.15 for prostate, 1.06 for colorectal males and 1.05 for colorectal female cancers. Figure 3 displays the observed and modeled relative survival rates some combination of age at diagnosis, period of diagnosis and time from diagnosis. Survival is assumed constant after 1999.

Calculation of the probabilities of death in the presence of all causes of death

For people diagnosed with colorectal cancer at age j, we define h ji the net (in the absence of other causes) probability of dying of cancer between ages i and i + 1, conditioned that they are alive prior to age i. Since we are assuming a constant colorectal cancer survival after 1999 we do not need to consider an index of time. This probability is calculated from s ji the interval relative survival estimated from SEER-9 cases diagnosed in 1991–1999.

$$ h_{ji}=(1-s_{ji}) $$

Similarly, the probability of dying from other causes, between ages i and i + 1 at year t, conditioned they survived prior to age i, \(\bar{h}_{i}^t\), can be calculated as

$$ \bar {h}_i^t =[1-E_i^t] $$

where \(E_i^t\) is the US life tables from the Human Mortality Database [15]. In this calculation we assume that mortality from other causes is independent of having cancer. In other words, a person diagnosed with cancer has the same hazard of dying of other causes as a person free of cancer. This assumption might not be true if we believe the disease or its treatment might affect other causes mortality. Age and year of birth are the important factors affecting mortality for other causes. The crude (in the presence of all causes of death) probabilities of dying for colorectal cancer and other causes of death between ages i and i + 1 at year t, given they are alive just prior to age i and were diagnosed with cancer at age j, denoted δ j i,t and ɛ j i,t , respectively, are calculated as

$$ \delta_{i,t}^j =h_{ji}-{\frac{1}{2}}h_{ji}\ \bar{h}_i^t\hbox{ and } \varepsilon_{i,t}^j =\bar {h}_i^t-{\frac{1}{2}}h_{ji}\ \bar {h}_i^t. $$

Details of calculations can be found in Cronin and Feuer [16].

Details of calculation to estimate number of person months in last year of life phase of care

We divide prevalent cases alive at January 1st year t by years since diagnosis. We first consider incidence cases diagnosed at January 1st year t, I i,t and apply the probabilities of dying in the same year δ i i,t and ɛ i i,t . We then consider prevalence by years since diagnosis. For prevalent cases diagnosed d  = 1, ... , 5 years prior to the prevalence date (diagnosed at year t − d or age i − d and denoted P td i,t ), the quantities P t-d i,t δ i-d i,t and P t-d i,t ɛ i-d i,t represent the estimated number dying during year t due to cancer and other causes, respectively, and \(\sum\limits_{d=1}^5{P_{i,t}^{t-d} \delta _{i,t}^{i-d}}\) and \(\sum\limits_{d=1}^5{P_{i,t}^{t-d} \varepsilon_{i,t}^{i-d}}\), represent the estimated number dying during year t for individuals diagnosed in the previous 5 years. Prevalent cases diagnosed more than 5 years from the prevalence date can be calculated as \(P_i-\sum\limits_{d=1}^5 {P_{i,t}^{t-d}}\). Assuming that after 5 years from diagnosis the annual hazard of dying of colorectal cancer is constant we apply and average hazard estimated using the hazards at year 6, 7 and 8 from diagnosis to estimate the number dying among those diagnosed 5 years of more from the prevalence date \(\left\{{P_i -\sum\nolimits_{d=1}^5{P_{i,t}^{t-d}}} \right\}\frac{1}{3}\sum\nolimits_{d=6}^8 {\delta_{i,t}^{i-d}}.\)

Thus, the number of CRC deaths at that year t is calculated as

$$ I_{i,t} \delta_{i,t}^i+\sum\limits_{d=1}^5{P_{i,t}^{t-d} \delta_{i,t}^{i-d}}+\left\{{P_i-\sum\limits_{d=1}^5 {P_{i,t}^{t-d}}}\right\}\frac{1}{3}\sum\limits_{d=6}^8 {\delta_{i,t}^{i-d}} $$

and the number of deaths due to causes other than colorectal at year y is

$$ I_{i,t} \varepsilon_{i,t}^i +\sum\limits_{d=1}^5 {P_{i,t}^{t-d} \varepsilon_{i,t}^{i-d}}+\left\{{P_i-\sum\limits_{d=1}^5 {P_{i,t}^{t-d}}}\right\}\frac{1}{3}\sum\limits_{d=6}^8 {\varepsilon_{i,t}^{i-d}}. $$

As a simplifying assumption, we consider that these patients contribute 12 months to the terminal care phase in year t because they represent a snapshot of the histories in the continuous timeline.

Rights and permissions

Reprints and permissions

About this article

Cite this article

Mariotto, A.B., Yabroff, K.R., Feuer, E.J. et al. Projecting the number of patients with colorectal carcinoma by phases of care in the US: 2000–2020. Cancer Causes Control 17, 1215–1226 (2006). https://doi.org/10.1007/s10552-006-0072-0

Download citation

  • Received:

  • Accepted:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10552-006-0072-0

Keywords

Navigation