臺灣子宮頸癌篩檢矛盾之解析：模擬模型與新型致死調整指標之應用

Abstract

背景 儘管自1995年推動全國性子宮頸癌篩檢以來，台灣的發生率已大幅下降，卻觀察到五年相對存活率呈現長期下降的矛盾現象，與其他已實施長期篩檢計畫的國家相似。此現象引發一關鍵問題：當發生率下降卻同時觀察到存活惡化時，癌症的整體負擔是否真的降低？ 方法 本研究建構一個多階段自然史模型，模擬子宮頸癌從癌前病變至晚期的進展過程，並納入篩檢介入，評估其對發生率、存活率與期別分布的長期影響。我們進一步提出一項創新指標—致死調整發生率（fatality-adjusted incidence rate），用以整合癌症的發生頻率與致死風險，更敏感地監測癌症負擔變化。最後，結合1990至2022年台灣癌症登記資料進行實證驗證。 結果 模擬結果顯示，整體五年存活率由75.3%下降至68.4%，主要反映期別組成變化；而各期別內的存活率則維持穩定或呈現上升趨勢。致死調整發生率則持續穩定下降，由8.2降至3.1每十萬人年，較能準確反映癌症負擔降低。實證資料亦呈現相似趨勢：1995年後原位癌明顯上升、侵襲癌逐漸下降；整體存活率自1999年後下降，而致死調整發生率則顯示持續下降，且其下降時點早於死亡率的變化。 結論 五年存活率的下降，並不意味著篩檢失效或治療效果惡化，而是因為篩檢傾向優先發現並移除早期病灶，導致後續被診斷的侵襲性癌症病例相對集中於預後較差的晚期階段所致。致死調整發生率能避免傳統指標因期別轉移所產生的誤導，並能更早、更精確反映癌症負擔變化，適合作為監測篩檢政策成效的補充工具，亦可推廣至其他癌症類型與地區情境。

Background Although Taiwan’s nationwide cervical cancer screening program launched in 1995 led to a marked decline in incidence, five-year relative survival has paradoxically decreased over time. Similar trends have been reported in other countries with mature screening programs, raising the question: does a reduction in incidence truly indicate a lower cancer burden if survival also worsens? Methods We developed a compartmental natural history simulation model of cervical cancer progression, incorporating a screening pathway to assess its impact on incidence, survival, and stage distribution. To better capture disease burden, we introduced a novel metric—the fatality-adjusted incidence rate —which integrates cancer frequency and fatality severity. We validated the model using empirical data from the Taiwan Cancer Registry between 1990 and 2022 to examine whether simulated trends align with observed changes. Results The simulation showed a decline in overall five-year survival from 75.3% to 68.4%, mainly due to a stage shift: screening removes early-stage lesions, resulting in a higher proportion of advanced-stage diagnoses among remaining invasive cases. In contrast, survival within each stage remained stable or improved. Fatality-adjusted incidence rate declined steadily from 8.2 to 3.1 per 100,000 woman-years, more accurately reflecting the reduction in cancer burden. Empirical data showed similar trends: a rise in in situ cases, decline in invasive cancers, and a continuous drop in fatality-adjusted incidence rate —preceding reductions in mortality and capturing early effects of opportunistic screening. Conclusion The decline in survival reflects a statistical consequence of stage redistribution, not a failure of screening or treatment. Fatality-adjusted incidence rate offers a robust and timely indicator that avoids biases in traditional metrics and provides an accurate reflection of population-level burden. It is a promising tool for evaluating cancer control strategies and can be extended to other cancer types and settings.