Prognosis Analysis of Service Time and Performance in Major League Baseball Players after Ulnar Collateral Ligament Reconstruction

研究背景 尺骨附屬韌帶重建術為美國職棒大聯盟界韌帶斷裂之主要治療方法。在1974年至2000年期間，尺骨附屬韌帶重建術每年僅有些微案例的增加；而到了2000年代，每年則約有10至20例手術個案，但特別的是到了2012年單年，手術案例卻突然暴增到36例，尺骨附屬韌帶重建術快速增加的情形也因此在運動期刊中引起關注與討論。 研究目的 而先前研究重視術前、術後1至3年大聯盟投手表現數據的比較，但未分析何種預後因子可能會造成選手術後重返大聯盟登板天數長短差異。因此，本研究的目的在於建立一長期觀察性的預後模型，以此模型評估大聯盟選手之術後重登板天數，同時使用本模型憑以預測這些優秀棒球員於術後之表現。 方法 本研究從Baseball-reference (http://www.baseball-reference.com)及Baseball Prospectus (http://www.baseballprospectus.com) 兩網站中收集1974年9月25日至2014年6月30日期間，曾執行過尺骨附屬韌帶重建術之大聯盟球員公開資訊，並追蹤這些選取球員資料至2015年4月28日。 本研究並定義一個時間反應變數為「尺骨附屬韌帶重建術後之大聯盟重登板天數」，其等同球員於術後回歸大聯盟日期直到該球員退休日期之期間。而重登板天數長短則使用於代表特定球員之表現價值。此外，因為收集球員資料中，部分屬於現役球員，這些球員並無退休日期，故使用存活分析方法以處理右設限問題。 研究結果 在276筆資料中，有236筆於尺骨附屬韌帶重建術後成功回歸登板至大聯盟(85.1%)，其中位數回歸重登板天數為1,675天/4.59年(95%CI：1,163-1,588天/2.94-4.35年)。單變數分析部分，年紀大於等於28歲之選手顯著性地有較短的重登板天數(P<0.0001)，而其它達到顯著促成較短重登板天數的預後因子包含較長(≥4.7年)的手術前大聯盟年資(P=0.0003)、較低(≤26.5)的身體質量指數(P=0.0085)、投手(相較於其它守備位置，P=0.0159)、執行過兩次手術(相較於執行過一次手術，P=0.0063)及非全明星選手(相較於明星選手，P=0.0494)。另Kaplan-Meiner方法則依照不同的預後因子，用於估計選手術後持續於大聯盟登板之機率。 多變數分析部分，年紀大於等於28歲之選手顯著性地有較高的風險(P=0.0070)於重登板後離開大聯盟，其它導致較高風險的預後因子包含較長(≥4.7年)的手術前大聯盟年資(P=0.0319)、較低(≤26.5)的身體質量指數(P=0.0054)、投手(相較於其它守備位置，P=0.0055)、美國聯盟選手(相較於國家聯盟選手，P=0.0437)及非全明星選手(相較於明星選手，P=0.0020) 研究結論 本研究透過存活分析方法，探討影響球員離開大聯盟風險之顯著預後因子，並用以處理球員不完整的術後重登板觀察時間。並分析發現，相對年輕且術前大聯盟年資較短、有較高的身體質量指數、屬於野手、美國聯盟及全明星選手，皆可能在大聯盟有較長的術後重登板天數。Background Ulnar collateral ligament reconstruction (UCLR) is the mainly medical treatment for arms ligament injury in American Major League Baseall (MLB). The cases of UCLR slightly increased between 1974 and 2000. After 2000s, UCLR rapidly increased to about 10 to 20 cases annually. Particuarly in 2012, the performed cases dramatcially increasd to 36 cases. Therefore, the recent literatures have payed increased attention in studying the UCLR in the sport journals. Objectives Prior literatures focued on the performance of pitcher before and after UCLR for 1 to 3 years, but did not analyze what prognosis factors may affect the the performance after UCLR based on the service time after returning the MLB. Therefore, the aim of this thesis was to establish a long-term prognsistic model for the service time after UCLR, and use this model to predict the dynamic post-UCLR performance of these elite players. Methods We selected 276 major league baseball players’ record with performing UCLR between September 25, 1974 and June 30, 2014 from public data in Baseball-reference (http://www.baseball-reference.com) and Baseball Prospectus (http://www.baseballprospectus.com). All the selected players were followed up to April 28, 2015.We define a time response variable, called ‘service time after UCLR’, as to the numbers of days from players’return dates to MLB after UCLR until the date of retirement from MLB. This service time was used to stand for the performance of a player. In addition, beause some players are still active in the collected data, survival analysis was applied to solve such a right-censored problem. Result. In the totally 276 players’record, there were 236 players(85.1%) who had successfully returned in MLB after UCLR, and the medium service time after UCLR was 1,675 days/4.59 years (95%CI：1163-1588 days/2.94-4.35 years). In the univariate analyses, the players with age≥ 28 years old had significantly shorter service days (P<0.0001), and the other significant prognosis factors induced shorter service time included longer (≥4.7years) MLB experience before UCLR (P=0.0003), lower (≤26.5) BMI (P=0.0085), pitchers vs. players with other positions (P=0.0159), the second vs. first UCLR (P=0.0063) and other players vs. all-star players (P=0.0494). The Kaplan-Meier method was used to estimate the probabilities of playing in MLB at the days after returning in MLB with repsect to different prognosis factors. In the multivariate analysis, we use Cox-regression model and backward stepwise method to select the prognosis factors in the model.We found that the players with age≥28 years old had significantly higher hazard (P=0.0070), and the other significant prognosis factors induced higher hazard included longer (≥4.7years) MLB experience before UCLR (P=0.0319), lower (≤26.5) BMI (P=0.0054), pitchers vs. players with other positions (P=0.0055), American Leagues vs. National Leagues (P=0.0437) and other players vs. all-star players (P=0.0020). Conclusion We adopt the survival analysis to identify the significant prognostic factors related to the risk of leaving MLB because the issue of the incomplete post-UCLR service times can be tackled. We found the player with the younger, shorter MLB experience before UCLR, higher BMI, with other positions than pitcher and National League all-star players are most likey to have a longer post-UCLR service times in MLB