1. Impact of Local Control on Cause-Specific Survival After SBRT for Early-Stage NSCLC: Dynamic Prediction With Landmarking

對早期NSCLC關於SBRT後特異性病因生存率局部控製的影響:具有標致性的動態預測

Introduction
介紹

Stereotactic body radiotherapy (SBRT) is an effective treatment for early-stage non-small cell lung cancer (NSCLC), especially in inoperable patients. Although previous studies have indicated that local control improves with higher doses above the biologically effective dose (BED) 100 Gy, the effect of improved local control on survival remains unclear. The purpose of this study was to assess the impact of local recurrence (LR) on cause-specific survival with a dynamic prediction model that incorporates LR as a time-dependent covariate.

立體定向體放射治療(SBRT)是治療早期非小細胞肺癌(NSCLC)的一種有效方法，尤其是對無法手術的患者。雖然以前的研究表明，當劑量高於生物有效劑量(BED) 100 Gy時，局部控製會改善，但改進的局部控製對存活率的影響仍不清楚。本研究的目的是評估局部複發(LR)對病因特異性生存的影響，采用動態預測模型，將LR作為一個時間依賴的協變量。

Methods
研究方法

This study included 386 stage IA NSCLC patients treated with SBRT from two centers, one using a high BED of 140 Gy or more and the other using a conventional BED of 105 Gy. We developed landmark dynamic prediction models for the probability of cause-specific survival. This model provides the probability of surviving an additional 2 years at different prediction time points during follow-up, given the history of recurrent status. Baseline covariates included in the model were age, gender and tumor diameter, and the time-dependent covariates were LR and regional or distant recurrence (RDR). The interactions between prediction time points and covariates were also considered in the model. LR was defined as recurrence within the radiation field.

該研究包括386名IA期NSCLC患者，他們接受了來自兩個中心的SBRT治療，一個使用140戈瑞或以上的高床位，另一個使用105戈瑞的常規床位。我們開發了具有裏程碑意義的原因特異性生存概率的動態預測模型。該模型提供了在隨訪期間不同預測時間點存活2年的概率，考慮到複發病史。模型中納入的基線協變量為年齡、性別和腫瘤直徑，時間依賴的協變量為LR和局部或遠處複發(RDR)。模型中還考慮了預測時間點與協變量之間的相互作用。LR定義為輻射場內的複發。

Results
結果

With the median follow-up of 4.3 years, 89 patients (23%) died of lung cancer. In a total of 127 patients who developed recurrence, 18 had LR only, 81 had RDR only, and 28 had both. The landmark model showed that age, tumor diameter, LR and RDR were significantly associated with increased odds of shorter cause-specific survival. Among these covariates, LR (adjusted odds ratio [aOR], 16.1; 95% CI, 9.7-26.7; P < .001) and RDR (aOR, 16.0; 95%CI, 11.6-22.0; P < .001)) demonstrated a strong effect on cause-specific death 2 years after the prediction time points.

中位隨訪時間為4.3年，89例患者(23%)死於肺癌。在127例複發患者中，18例僅LR, 81例僅RDR, 28例兩者均有。裏程碑模型顯示，年齡、腫瘤直徑、LR和RDR與較短病因特異性生存率的增加顯著相關。在這些協變量中，LR(調整優勢比[aOR]， 16.1;95%置信區間,9.7 - -26.7;P < .001)和RDR (aOR, 16.0;95%置信區間,11.6 - -22.0;(P < .001))顯示了對預測時間點2年後的死因特異性死亡的強烈影響。

Conclusion
結論

The dynamic prediction using landmark model showed that LR had a strong impact on subsequent cause-specific deaths. This result suggests that improving local control with higher doses is a reasonable strategy.

使用標誌性模型進行的動態預測表明，LR對隨後的死因特異性死亡有很強的影響。這一結果表明，提高劑量以改善局部控製是一項合理的策略。