This study is the first that simultaneously documents the improvements in life expectancy, EYLL, and savings in lifetime healthcare expenditures at different stages of cervical cancer, and the results show that in addition to stage 0, detection of cervical cancer at stages 1–3 can lead to more improvements in life expectancy and costs compared with a more advanced stage (Table 1), and the younger the age of diagnosis, the greater the benefits with regard to EYLL. However, we must carefully examine the accuracy of our estimation before making further inferences. First, since we only included patients with cervical cancer that had been verified with histopathological evidence and registered in the Taiwan Cancer Registry, the diagnoses were highly accurate. Second, because all cases of invasive cervical cancer are registered in the Catastrophic Illnesses database, the waiving of all co-payments has been under the careful monitoring and control of gyneco-oncologists, and all related reimbursements for treating cervical cancer would generally follow the established guidelines, being comprehensive and comparable for different stages. Third, all the extrapolations of survival functions are based on the validated assumption of “constant excess hazard”, which can be obtained by showing a straight line after taking the logit transform of the survival ratio between the index and age- and gender-matched referents [16, 17]. As the assumption of a constant excess hazard may have a strong impact on the estimation of life expectancy for cervical cancer, we conduct a sensitivity analysis. Because the iSQoL software cannot be directly set to zero value of slope for extrapolation, we deliberately chose the second slope value that is closest to zero (either negative or positive) for extrapolation 10 years after follow-up. The results (presented in the Additional file 1) show that all the life expectancies were very close (<15% difference), indicating that our estimates are relatively accurate. Moreover, this study validated this estimation by extrapolating the survival of the first five to 10 years, and the results showed that this approach usually has less than 10% error in comparison with the actual survival based on the Kaplan-Meier method (Table 2). In one of our previous studies , we employed the cohort of cervical cancer patients between 1990 and 2001 in the National Cancer Registry and followed up to 2004, while the current study enrolled the cohort of 2002–2009 and followed up to 2011. Since there have been no major changes with regard to treating cervical cancer during the last two decades, it is perhaps not surprising that we found no major changes in the estimates of life expectancy between the two cohorts (19.77 years in Chu’s study versus 19.85 years in this work). However, as the life expectancy of the general female population has increased from 77.7 yrs in 1995 to 80.8 yrs in 2005, it is not unexpected that the EYLL also increased from 6.33 years to 7.78 years. Therefore, the estimation method can be seen as both consistent and accurate, and, as noted above, we tentatively conclude that detection of invasive cervical cancer before stage 3 compared with a more advanced stage can have benefits with regard to life-years and costs for patients aged below 65, while those aged over 65 must be detected earlier than stage 2 to see the same benefits. Generally speaking, the earlier the stage at diagnosis, the better the outcomes, although we might have over-estimated the effects of early detection because of potential length time bias.
Studies of cervical cancer screening tend to emphasize detection at stage 0. This study, however, provides solid evidence that detection and treatment of invasive cervical cancer at stage 1 or 2 is also very worthwhile. The calculation of EYLL in Table 1 used an age- and gender-matched general population as referents, and provides estimates for the number of life-years lost due to invasive cancer [9, 12]. Because our method takes the age at diagnosis into consideration, the estimations would be less affected by lead time bias and more accurate than direct comparisons of life expectancies for cancer patients diagnosed at different stages. We recommend that the results be used to analyze the cost-effectiveness of screening programs. As Figure 2 indicates that the cervical cancer survival probabilities in Taiwan appear comparable with those reported from other countries [2–5], our findings may also be applicable to them.
Although this study has used the most comprehensive national data currently available in Taiwan, it has the following limitations that need to be addressed: First, the lifetime extrapolation is based on current and prior experiences, especially the national life tables; however, it is clear that such an ex post approach could easily underestimate the actual survival of future cancer populations, because it cannot predict the future development and adoption of newer technologies for cancer diagnosis and management. Therefore, our estimation of the lifetime survival of cancer patients may be a conservative one, while the EYLL might be overestimated. Second, because life expectancy is also a function of co-morbidity, performance states, and recurrence , the current estimates provide only a crude estimation of the average EYLL. Future studies with a larger cohort may stratify them into sub-cohorts based on more clinical data on co-morbidities, performance states, and recurrence, to improve the accuracy of the predictions. Third, we did not consider the growing evidence that those women who decide not to participate in screening may be inherently different from those who decide to participate, and these non-participants might have higher other-cause mortality . If this phenomena occurred in Taiwan, then our EYLL would be overestimated. Fourth, this study adopted the insurer’s perspective, and only direct medical costs were estimated. Because of the lack of empirical data on the costs of out-of-pocket money or lost productivity due to cervical cancer or premature death, our results underestimate the cost of illness to the whole society. Finally, because the healthcare expenditures after the end of the follow-up period were assumed to be the same as the average of the last 10% of measurements based on kernel smoothing, this study might have overestimated the costs after the end of 10 years of follow-up. However, since almost all cases of cervical cancer would be in healthy condition 5–10 years after diagnosis, except those approaching the end of their lives, the average costs due to cervical cancer would generally become smaller given a large number of healthy survivors and higher cumulative discount rates. The potential overestimation due to this would thus be very small.
Policy implications for community healthcare
Pap smears are not very popular among women aged 60 and older in Taiwan, which might have resulted in higher morbidity and mortality rates for cervical cancer among this group . This study provides evidence that early detection of invasive cancer can saves lives and reduce costs for both young and old patients, and that the earlier detection occurs, the better (Table 1), and these facts can be used to encourage those who are otherwise afraid of undergoing cancer screening. However, further evaluations of the cost-effectiveness of this approach are needed in order to optimize the utilization of resources.