Costs of cervical cancer treatment: population-based estimates from Ontario

Original Article

Costs of cervical cancer treatment: population-based estimates from Ontario

C. Pendrith, MSc*, A. Thind, MD PhD*,, G.S. Zaric, PhD*, S. Sarma, PhD,*§
*Department of Epidemiology and Biostatistics, Schulich School of Medicine and Dentistry, University of Western Ontario, London, ON;, Centre for Studies in Family Medicine, and Schulich Interfaculty Program in Public Health, University of Western Ontario, London, ON;, Ivey Business School, University of Western Ontario, London, ON;, §Institute for Clinical Evaluative Sciences, Toronto, ON..




The objectives of the present study were to estimate the overall and specific medical care costs associated with cervical cancer in the first 5 years after diagnosis in Ontario.


Incident cases of invasive cervical cancer during 2007–2010 were identified from the Ontario Cancer Registry and linked to administrative databases held at the Institute for Clinical Evaluative Sciences. Mean costs in 2010 Canadian dollars were estimated using the arithmetic mean and estimators that adjust for censored data.


Mean age of the patients in the study cohort (779 cases) was 49.3 years. The mean overall medical care cost was $39,187 [standard error (se): $1,327] in the 1st year after diagnosis. Costs in year 1 ranged from $34,648 (se: $1,275) for those who survived at least 1 year to $69,142 (se: $4,818) for those who died from cervical cancer within 1 year. At 5 years after diagnosis, the mean overall unadjusted cost was $63,131 (se: $3,131), and the cost adjusted for censoring was $68,745 (se: $2,963). Inpatient hospitalizations and cancer-related care were the two largest components of cancer treatment costs.


We found that the estimated mean costs that did not account for censoring were consistently undervalued, highlighting the importance of estimates based on censoring-adjusted costs in cervical cancer. Our results are reliable for estimating the economic burden of cervical cancer and the cost-effectiveness of cervical cancer prevention strategies.

KEYWORDS: Cervical cancer, cost estimates, censoring, Ontario, population-based


Cancer is the leading cause of death in Canada, accounting for nearly 30% of all mortality1. Cancer is one of the most costly diseases2, and its economic burden is substantial in Canada. The direct cost of cancer care in Canada in 2008 was estimated at $4 billion3.

Cervical cancer is the 2nd leading cause of cancer death among Ontario women 20–44 years of age4, and the 4th most common cause of cancer death among women worldwide5. Of every 145 Ontario women, 1 will be diagnosed with cervical cancer during her lifetime, and each year in Ontario, approximately 610 women are diagnosed with cervical cancer, and approximately 150 die from the disease1.

Treatment for cervical cancer is complex and can include surgery, chemotherapy, and radiation therapy. Of Ontario women diagnosed with cervical cancer in 2003–2004, more than 30% received chemotherapy, and an estimated 55% received radiation therapy6. More than half of Ontario’s cervical cancer patients had a cancer-related surgical procedure, and each patient had an average of 1.5 hospital admissions within 12 months of diagnosis6. In the United States, resource consumption for cancer patients is highest during the initial phase of treatment and the terminal phase before death7, because in the 1st year after diagnosis, patients undergo primary treatment and experience the greatest mortality8,9.

Accurate estimates of the cost of cancer treatment are crucial for economic evaluations, policy decisions, and forecasting future medical care expenditures relating to cancer treatment. Although a prior study estimated the costs of cervical cancer treatment in Canada10, no Canadian study has, to the best of our knowledge, examined cervical cancer costs beyond the 1st year after diagnosis or accounted for censoring. In addition, prior publications using Ontario data for cost estimates failed to capture costs associated with visits to cancer clinics or dialysis clinics or those associated with mental health admissions, because those data were not available before 2007. The objective of the present study was therefore to fill those important gaps by using the most recent Ontario data and by accounting for censoring, thereby providing estimates of the total direct medical care costs of treating cervical cancer during the first 5 years after diagnosis. We estimated costs from the perspective of the Ontario Ministry of Health and Long-Term Care.


Ontario women 35–69 years of age with incident primary cases of cervical cancer (International Classification of Diseases, 9th revision, code group 180.x) diagnosed between 1 January 2007 and 31 December 2010 were identified from the Ontario Cancer Registry. Records of those cervical cancer cases were linked with several population-based administrative databases held at the Institute for Clinical Evaluative Sciences (ices). The linkages used unique encoded identifiers and were analyzed at ices. Demographic data were obtained from the Registered Persons Database, and dates and causes of death were obtained from the Vital Statistics Registry. Patients without a valid Ontario Health Insurance Plan number or whose date of death preceded the date of diagnosis were excluded.

Data about direct medical care costs were obtained from the ices administrative databases using the costing methodology developed at ices11; Table i describes in detail the data sources and costing methodologies for all health care services included in the present study. We captured the costs of all health care services covered by the province that were provided to our patient cohort: inpatient hospitalizations, same-day surgeries, emergency department visits, ambulatory visits to hospital (for example, to cancer clinics and dialysis clinics), stays in long-term care, inpatient rehabilitation stays, physician services, home care services, and community laboratory services. We captured the costs of publicly insured prescription drugs for patients 65 years of age and older or meeting other eligibility criteria for the Ontario Drug Benefit. Costs from a patient’s index date (date of diagnosis) to date of death or 31 December 2012 were included. Costs for each year were adjusted for inflation and are presented in 2010 Canadian dollars (adjusted using Statistics Canada’s Consumer Price Index for health care).

TABLE I Data sources and costing methodology


We estimated overall and specific costs during the first 5 years after diagnosis. Annual costs were estimated for patients who survived longer than 1 year and for patients who died in a given year from any cause, from a cervical cancer-related cause, and from another cancer cause of death. Overall and cancer-clinic costs were also estimated by year of diagnosis. First, we used the arithmetic mean (that is, a simple mean) to estimate costs. However, the simple mean estimation is likely to be inaccurate because length of follow-up varied for the patients. To account for censored data because of the varying follow-up, we used the Bang and Tsiatis (B&T) estimator for weighted and improved estimates1214. All analyses were conducted using the SAS software application (version 9.2: SAS Institute, Cary, NC, U.S.A). Appendix A presents the details of the statistical methods.


The women in the study cohort, representing 779 cases of cervical cancer, had been diagnosed between 2007 and 2010. Mean age at diagnosis was 49 years (95% confidence interval: 47 years to 53 years). About 36% of the patients (n = 279) died within 5 years of diagnosis, and of those 279 deaths, 79% (n = 221) were caused by cervical cancer. The entire cohort was observed for a minimum of 2 years after diagnosis or until death, and therefore no cases were censored during years 1 or 2 after diagnosis. In the 3rd year after diagnosis, 21% of the cohort had censored cost data, and by the end of year 5, 52% of patients were censored.

Table ii reports mean costs for the study cohort overall and by 1-year vital status without taking censoring into account. Overall mean cost during the 1st year post-diagnosis was $39,187 [standard error (se): $1,327]. The mean cost was much higher for patients who died within 1 year of diagnosis ($66,790; se: $4,482) than for those who survived for longer than 1 year ($34,648; se: $1,275). The cost was higher for patients who died from cervical cancer ($69,142; se: $4,483) than for those who died from other causes ($56,824; se: $11,338). For patients who survived 1 year or longer, the highest cost category was cancer clinic costs; for patients who died within 1 year, it was inpatient hospitalization (Table ii). Cancer clinic costs were lower for patients who died within 1 year of diagnosis from non-cervical-cancer causes ($5,238; se: $1,365) than for patients who died from cervical cancer ($12,440; se: $1,143) or who survived at least 1 year ($14,130; se: $516).

TABLE II Costs associated with cervical cancer in the first year after diagnosis


Table iii reports mean cumulative costs in cervical cancer patients (simple arithmetic mean and weighted and improved estimates). Annual costs were highest during year 1 ($39,187; se: $1,327) and declined during subsequent years (year 2: $14,425; se: $1,346; year 3: $11,280; se: $1,677; year 4: $8,444; se: $1,023; year 5: $5,480; se $1,074).

TABLE III Medical care costs associated with cervical cancer during years 1–5 after diagnosis


Mean 1-year costs varied greatly by year of diagnosis, ranging from $35,519 (se: $2,257) in 2010 to $45,369 (se: $3,383) in 2009 (Table IV). Table V reports costs by year after diagnosis and survival during that year. Our results showed that the estimated mean costs without accounting for censoring were consistently lower than the B&T estimates. Mean cumulative 3- and 4-year costs were $58,702 (se: $2,710) and $63,131 ($3,131) respectively when estimated using the simple mean. Using the improved estimator, the corresponding cumulative 3- and 5-year costs were $59,768 ($3,016) and $68,745 ($2,963) respectively. Estimates using weighted and improved estimators were similar, but the variance with the improved estimator was generally smaller and thus more efficient.

TABLE IV Average 1-year total medical care and cancer clinic costs by year of diagnosis (uncensored cost estimates)


TABLE V Average annual total medical care costs by year and vital status (uncensored cost estimates)



In the present study, we estimated the direct medical care costs of cervical cancer treatment in Ontario during the first 5 years after a diagnosis of cervical cancer. Cumulative 5-year cancer clinic and overall costs per patient were $17,294 and $68,745 respectively. Cost accumulation was greatest during the 1st year after diagnosis, an unsurprising observation given that treatment is most aggressive during that period7,8,10. Annual total medical care costs declined from $39,187 per patient in year 1 after diagnosis to $14,425 during year 2, $11,280 during year 3, $8,444 during year 4, and $5,480 during year 5. The 1-year costs were much higher for patients who died from cervical cancer within the year after diagnosis ($69,142) than for patients who survived at least 1 year ($34,648). That finding is unsurprising, given that patients who die from cervical cancer are more likely to have late-stage disease and to receive more intensive treatment, which incurs greater costs. Average 1-year costs varied significantly by year of diagnosis; however, those differences are likely the result of differences in the proportion of deaths in the given years.

Our results showed that cancer clinic and hospital admissions were the two largest drivers of costs in the 1st year after diagnosis, corroborating the findings of a prior study10. Those cost categories capture costs associated with cancer-related treatments such as chemotherapy, radiation therapy, and cancer-related surgeries. Our estimated cost of inpatient admissions ($11,539) was nearly double that reported by de Oliveira et al.10 ($6,761), and our estimated cost of cancer-related care ($13,697) was much higher than their estimates of chemotherapy ($804) and radiation therapy ($3,468) combined. Those differences likely reflects the fact that our estimates included all cancer clinic visits, which are not limited to chemotherapy and radiation therapy and can include services such as palliative care, surgical oncology, and supportive services. Furthermore, de Oliveira et al.10 might have underestimated the costs of radiation therapy, because their cost per fraction of radiation was estimated using data from the 1990s.

To the best of our knowledge, our study of medical care costs for Canadian cervical cancer patients is the first to take censoring into account. An earlier Ontario study reported 1-year average costs of $18,055 (2009 Canadian dollars) for cervical cancer patients who survived 1 year and $41,536 for those who died within 1 year10, which are lower than our mean estimates. The discrepancy might be a result of the fact that our estimates captured all cancer-related costs, including costs associated with rehabilitation, mental health admissions, dialysis clinic visits, and all Ontario Health Insurance Plan billings. To the best of our knowledge, no Canadian studies have described costs for cervical cancer beyond year 1, thus permitting a comparison with our results. In our study, resource consumption was highest during year 1, accounting for 67% of cumulative 3-year costs. That result resembles U.S. findings by Insinga et al.9, who concluded that 69% of 3-year costs were incurred during year 19. As in other studies of cervical and other cancer patients, we found that mean costs were much higher for patients who died than for those who survived8,10.

Our study also has several limitations. First, the costs were highly skewed, and estimates of mean cost are influenced by high-cost users. Second, cancer staging data were not available for the study cohort; thus, we were unable to produce stage-specific cost estimates. Third, we were unable to exclude medical care costs unrelated to cancer. However, we expect that costs unrelated to cancer are likely to be small for this patient population. Fourth, our estimates of outpatient prescription drug costs were based on data from the Ontario Drug Benefit program. Because the Ontario Drug Benefit provides coverage only to patients 65 years of age and older or to those meeting other eligibility criteria, our outpatient prescription drug costs could be slightly underestimated. However, chemotherapy, representing the largest pharmacotherapy cost, is administered in hospital and is captured by our estimates of cancer clinic costs. Finally, data limitations precluded us from estimating the costs of treatment beyond 5 years after diagnosis.

Our study has several strengths. First, we estimated the overall and specific medical care costs of cervical cancer treatment for the 1st year and beyond the 1st year after diagnosis. Second, we accounted for censoring to produce accurate estimates of treatment costs beyond 1 year. Our simple mean costs were much lower than those estimated using the weighted and improved estimators, which suggests that earlier published studies of average costs underestimated the true treatment costs. Third, we included all medical care costs covered by the Ontario Ministry of Health and Long-Term Care. Estimates of overall resource utilization might be useful to decision-makers. Finally, our cost estimates are reliable for economic evaluations of interventions to prevent cervical cancer and for calculation of lifetime cervical cancer–related costs.


We analyzed the overall and specific medical care costs of treating cervical cancer in the first 5 years after a diagnosis of cervical cancer in Ontario. By taking censoring into account, our estimates are more likely to reflect the true medical care costs of cervical cancer treatment in Ontario. Overall medical care costs were approximately $40,000 in year 1, $14,000 in year 2, $11,000 in year 3, $9,000 in year 4, and $5,500 in year 5.

We found that costs associated with cancer clinic visits and inpatient admissions were the two largest sources of cervical cancer treatment costs. However, physician services and home care were also significant drivers of costs. Our estimates could be of use for future economic evaluations of human papillomavirus vaccines, screening strategies, or other preventive interventions. Decision-makers might also find our estimates useful for policy planning or projecting future costs.


The authors thank Dr. Rick Glazier, Alex Kopp, and Nathaniel Jembere at the Institute for Clinical Evaluative Sciences (ices) and Dr. Salimah Shariff and ices Western. This study was supported by ices, which is funded by an annual grant from the Ontario Ministry of Health and Long-Term Care (mohltc). The opinions, results, and conclusions reported in this paper are those of the authors and are independent from the funding sources. No endorsement by ices or mohltc is intended or should be inferred.

This work is a revised version of one of CP’s msc thesis chapters submitted to the University of Western Ontario. The study was part of a larger project funded by the Canadian Institutes of Health Research (grant no. MOP-130454).


We have read and understood Current Oncology’s policy on disclosing conflicts of interest, and we declare that we have none.


1. Canadian Cancer Society’s Advisory Committee on Cancer Statistics. Canadian Cancer Statistics 2013. Toronto, ON: Canadian Cancer Society; 2013.

2. Institute of Health Economics (ihe). IHE in Your Pocket 2014: A Handbook of Health Economic Statistics. Edmonton, AB: ihe; 2014. [Downloadable from:; cited 29 August 2014]

3. de Oliveira C, Bremner KE, Pataky R, et al. Trends in use and cost of initial cancer treatment in Ontario: a population-based descriptive study. CMAJ Open 2013;1:E151–8.

4. Cancer Care Ontario (cco). Ontario Cervical Screening Program: 2012 Report. Toronto, ON: cco; 2014.

5. Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D. Global cancer statistics. CA Cancer J Clin 2011;61:69–90.
cross-ref  pubmed  

6. Elit L, Barbera L, Schultz SE, et al. Surgery for cervical cancer. In: Urbach DR, Simunovic M, Schultz SE, eds. Cancer Surgery in Ontario: ICES Atlas. 2nd ed. Toronto, ON: Institute for Clinical Evaluative Sciences; 2008: 167–88.

7. Baker MS, Kessler LG, Smucker RC. Site-specific treatment costs for cancer: an analysis of Medicare continuous history sample file. In: Sheffler RM, Andrews NC, eds. Cancer Care and Costs. Ann Arbor, MI: Health Administration Press; 1989: 127–38.

8. Insinga RP, Ye X, Singhal PK, Carides GW. Healthcare resource use and costs associated with cervical, vaginal and vulvar cancers in a large U.S. health plan. Gynecol Oncol 2008;111:188–96.
cross-ref  pubmed  

9. Akhtar-Danesh N, Lytwyn A, Elit L. Five-year trends in mortality indices among gynaecological cancer patients in Canada. Gynecol Oncol 2012;127:620–4.
cross-ref  pubmed  

10. de Oliveira C, Bremner KE, Pataky R, et al. Understanding the costs of cancer care before and after diagnosis for the 21 most common cancers in Ontario: a population-based descriptive study. CMAJ Open 2013;1:E1–8.
cross-ref  pubmed  pmc  

11. Wodchis WP, Bushmeneva K, Nikitovic M, McKillop I. Guidelines on Person-Level Costing Using Administrative Databases in Ontario. Working Paper Series. Vol 1. Toronto, ON: Health System Performance Research Network; 2013.

12. Bang H, Tsiatis AA. Estimating medical costs with censored data. Biometrika 2000;87:329–43.

13. Pfeifer PE, Bang H. Non-parametric estimator of mean customer lifetime value. J Interact Mark 2005;19:48–66.

14. Zhao H, Bang H, Wang H, Pfeifer PE. On the equivalence of some medical cost estimators with censored data. Stat Med 2007;26:4520–30.
cross-ref  pubmed  

15. Huang Y. Cost analysis with censored data. Med Care 2009;47(suppl 1):S115–19.
cross-ref  pubmed  pmc  

16. Zhao H, Wang H. Cost and cost-effectiveness analysis with censored data. In: Faries DE, Leon AC, Haro JM, Obenchain RL. Analysis of Observational Health Care Data Using SAS. Cary, NC: SAS Institute; 2010: 363–82.

17. Lin DY, Feuer EJ, Etzioni R, Wax Y. Estimating medical costs from incomplete follow-up data. Biometrics 1997;53:419–34.
cross-ref  pubmed  

18. Briggs AH, O’Brien BJ, Blackhouse G. Thinking outside the box: recent advances in the analysis and presentation of uncertainty in cost-effectiveness studies. Annu Rev Public Health 2002;23:377–401.
cross-ref  pubmed  


Equation 1

The simple mean for estimating the costs for a cohort of patients is eq where μ̂ is the estimated arithmetic mean cost, Mi is the cost accumulated by patient i during a given time period. Estimates based on Equation 1 will bias the mean downward because costs accrued after observed follow-up are equated to zero15,16. Estimates derived solely from complete observations will be biased upward to patients with a shorter survival time16. Applying standard survival analysis techniques to costing analyses is also invalid because censoring and cost are not independent15,17,18. Censoring is more likely for patients who accumulate costs slowly than for high-cost users; the mean is therefore biased upward17,18. Given that censoring increases with follow-up, appropriate statistical methods that address censoring are required to reduce bias in estimates of costs.

Assumptions of the B&T Estimator

The random variable M represents the costs accumulated by a patient during a specified time T, which is bounded by maximum time L. T is assumed to follow a continuous distribution (0 ≤ TL); if all costs are available, it corresponds to survival time—otherwise, to the period during which costs are observed. If all patients have cost data available for time period L or longer, then the mean cost is the simple arithmetic mean. However, because of staggered dates of diagnosis, costs for all patients are not completely observed. Given the censored nature of the data, we consider a potential time to censoring C, which is assumed to be completely random. C is also assumed to be continuous, and the probability of a censoring time of at least time L is assumed to be greater than zero [pr(CiL) > 0]. The latter assumption is necessary to ensure that, to calculate mean, costs for some patients are observed for the defined study period.

The B&T estimator weights costs using the Kaplan–Meier survival curve and therefore includes the associated assumptions. The Kaplan–Meier estimate assumes non-informative censoring or the independence of censoring from the probability of the outcome of interest. Survival probabilities are also assumed to be the same for patients with early study entry as for those with later entry, and the probability of survival within a time interval is assumed to be constant.

Equation 2

The weighted estimator for estimating costs for a cohort of patients is eq is the weighted estimator for time-restricted costs as proposed by Bang and Tsiatis12, in which complete cases consist of patients who die during the study period or who are observed until the end of the study period (L). Costs of complete cases are weighted by the inverse probability of the Kaplan–Meier estimate not being censored at the end of the interval. Estimates based on Equation 2 allow for continuous death and censoring times and provide a consistent estimate of mean cumulative medical care costs12,14,15. However, this estimator is inefficient because it relies on costs from patients with complete data and could be unstable with heavy censoring12,15.

In Equation 2, is the estimated mean cost based on the simple weighted estimator, Ti indicates a failure time, and Ci indicates a censored time. Observed follow-up time is Xi = min (Ti, Ci), Δi = I (TiC). I(.) is the indicator function, with I = 1 indicating a failure and I = 0 indicating a censored observation. T is bounded by the maximum follow-up time L, where TiL and Pr (CiL) > 0. K (Ti) is the Kaplan–Meier estimate of the probability of not being censored at failure time Ti or censoring time Ci.

Equation 3

The improved estimator for estimating the costs for a cohort of patients is eq and eq is the improved estimator proposed by Pfeifer and Bang13, which attempts to improve efficiency relative to the simple estimator by using data from censored cases12. In Equation 3, is the estimated mean cost based on the improved estimator, M (Ci) is the mean cost for all individuals still under observation at censoring time Ci, Xj indicates that individual j is still under observation beyond individual i’s censoring time, and Mj (Cj) is the cost accumulated by individual j at time Ci. The improved estimator has two parts13:

  • ■ Mean cost of complete cases estimated by the simple weighted B&T estimator

  • ■ An efficiency term that estimates the costs of censored cases

Censored costs are adjusted by subtracting the mean cost for all other cases still under observation at that censoring time. Adjusted censored costs are then weighted by the Kaplan–Meier inverse probability of not being censored at that time. The efficiency term is the average of the weighted censored costs.

Correspondence to: Sisira Sarma, Kresge Building, Room K201, Department of Epidemiology and Biostatistics, Schulich School of Medicine and Dentistry, The University of Western Ontario, London, Ontario N6A 5C1. E-mail:

(Return to Top)

Current Oncology, VOLUME 23, NUMBER 2, April 2016

Copyright © 2019 Multimed Inc.
ISSN: 1198-0052 (Print) ISSN: 1718-7729 (Online)