1、经典临床数据陷阱分析疑夕内容提要内容提要 2013-06-03:SNTA的ganetespib肺癌II/III期试验结果亚组分析数据一般不被认可 2014-03-10:LJPC的GCS-100慢性肾病II期试验结果标准偏差过大导致均值失去意义 2014-10-06:ADHD的MDX成人ADHD III期试验结果安慰剂组里的“刁民”不能随便剔除 2015-03-12:CUR的NSI-566渐冻症II期试验结果一不小心就被绕进去的逻辑陷阱 2015-04-26:CLDN的MYDICAR心衰IIb期试验失败治疗组患者比安慰剂组更健康造成有效假象2013-06-03:SNTA公布公布ganetespi
2、b肺癌肺癌II/III期试验结果期试验结果Synta announces positive overall survival results from GALAXY-1 Phase 2b/3 trial of ganetespib in second-line non-small cell lung cancer.Median overall survival improved 32%in the all adenocarcinoma,intent-to-treat population.In the population selected for the ongoing GALAXY-2 P
3、hase 3 trial,median overall survival improved 67%,Hazard Ratio=0.61(p=0.009).The magnitude and consistency of the activity reported today are very encouraging and bode well for the outcome of the GALAXY-2 Phase 3 study.Press Release中充斥着positive,encouraging,improve,事实果真如此吗?市场对市场对ganetespib肺癌肺癌II/III期
4、试验结果的反应期试验结果的反应结果公布当日跌34%,随后半个月连续下跌。Ganetespib肺癌肺癌II/III期试验真的成功了吗?期试验真的成功了吗?ITT组OS数据未达到统计学显著差异,亚组分析数据一般不被投资人认可。2014-03-10:LJPC公布公布GCS-100慢性肾病慢性肾病II期试验结果期试验结果La Jolla Pharmaceutical Company reports positive,top-line results from Phase 2 clinical trial of GCS-100 in chronic kidney disease.The trial me
5、t its primary efficacy endpoint of a statistically significant improvement in kidney function.Key secondary endpoints were also met,and the effect on circulating galectin-3 levels was consistent with the effect on eGFR.Our experience with GCS-100 in this trial was very encouraging,and my patients to
6、lerated the therapy well.Press Release中充斥着positive,encouraging,主要临床终点、次要临床终点都达到,事实果真如此吗?市场对市场对GCS-100慢性肾病慢性肾病II期试验结果的反应期试验结果的反应市场表现非常滑稽,数据公布当天涨65%,但是在随后的半个月内跌回原点。GCS-100真的升高肾小球滤过率吗?真的升高肾小球滤过率吗?GFR,肾小球滤过率,GFR下降是慢性肾病的标志,GCS-100慢性肾病II期临床以此为主要临床终点,安慰剂组、1.5 mg/m2组、30 mg/m2组分别-0.58、+1.26、+0.06,初看起来1.5 mg/
7、m2组的GFR确实升高了,而且达到统计学显著差异。GCS-100真的降低真的降低galectin-3水平水平吗?吗?galectin-3,GCS-100的靶点的靶点,慢性肾病慢性肾病galectin-3水平升高,水平升高,GCS-100慢性肾病II期临床以此为次要临床终点,安慰剂组、1.5 mg/m2组、30 mg/m2组分别+1.03、-0.88、+1.29,初看起来1.5 mg/m2组的galectin-3水平水平确实降低了。低剂量组有效,高剂量组无效,事出反常必有妖!低剂量组有效,高剂量组无效,事出反常必有妖!你注意到两个表格第6行的StDev值了吗?中文名叫标准偏差,反映数据相对于平均
8、值的离散程度,将GCS-100的肾小球滤过率数据用图表示如下,你还认为GCS-100与安慰剂有显著差异吗?2015年05月,LJPC宣布放弃GCS-100的研发。2014-10-06:ADHD公布公布MDX成人成人ADHD III期试验结果期试验结果In a modified Intent To Treat(mITT)population(n=293),MDX demonstrated a statistically significant improvement in ADHD symptoms compared to placebo(p0.03).This is a key milesto
9、ne for Alcobra.We are encouraged by these findings,as they build upon our Phase II studies showing that MDX significantly improved symptoms of ADHD without many of the safety and tolerability issues commonly associated with currently available ADHD medications.Alcobra公司对III期临床结果进行了非常美好的解读,认为这是一个重要的里
10、程碑,足以让人为之欢欣鼓舞,那么市场会作何反应呢?市场对市场对MDX成人成人ADHD III期试验结果的反应期试验结果的反应结果公布当日跌57%,随后大半个月持续下跌。mITT与与ITT差之毫厘谬以千里差之毫厘谬以千里The mITT population was derived by a post hoc exclusion of four subjects with extreme placebo responses(The ITT analysis before exclusion yielded a positive trend,p=0.15;n=297).We conducted t
11、he mITT analysis after observing the disproportional effect of a few extremely large placebo responses which were inconsistent with what has been reported in previous ADHD trials of MDX or other agents.ITT全称意向性治疗,患者来参加临床试验就是为了治病,最终不管什么原因失败(副作用太大退出、试验过程中改主意、被车撞死了)都算作试验药物无效。Alcobra公司首先分析了ITT数据,发现没有达到统
12、计学显著差异,于是将安慰剂组的4个“刁民”(看起来比吃了药还有效)的数据踢出去,然后做mITT分析达到统计学显著差异,mITT数据一般是不被认可的。2015-03-12:CUR公布公布NSI-566渐冻症渐冻症II期试验结果期试验结果The study met primary safety endpoints.Secondary efficacy endpoints at nine months post-surgery indicate a 47%response rate to the stem cell treatment.The average ALSFRS score for res
13、ponders at 9 months after treatment was 37.Non-responders scored an average of 14.(statistically significant)Responders disease progression was-0.007 point per day,while non-responders disease progression was-0.1 per day,which was again statistically significant.The top-line data show disease stabil
14、ization in a subgroup of patients.The top-line data look very positive and encouraging.Press Release中充斥着positive,encouraging,应答率,统计学显著差异,事实果真如此吗?市场对市场对NSI-566渐冻症渐冻症II期试验结果的反应期试验结果的反应结果公布当日跌37%,随后半个月持续下跌。应答者NSI-566渐冻症渐冻症II期试验具体数据期试验具体数据这是一项单组试验,共入组15例患者,主要临床终点是安全性,次要临床终点是有效性,采用ALSFRS评分评价有效性,ALSFRS评分下
15、降越快说明患者病情恶化越快。应答者平均每天-0.007分,非应答者平均每天-0.1分,应答者显著优于非应答者。为什么说为什么说Neuralstem故意误导投资人?故意误导投资人?正确的做法是统计15例患者ALSFRS评分下降的平均值,然后对比历史数据,说明经NSI-566治疗后,患者病情恶化速度是加快了还是减慢了。如此计算出的结果是平均每月-1.70分,不优于甚至劣于历史数据,于是Neuralstem决定想办法掩盖这个不太好的数据。统计分析时Neuralstem将15例患者分成两组,表现好的7人是应答者,表现不好的8人是非应答者,应答者与非应答者之间存在统计学显著差异。Neuralstem故意
16、制造了一种错觉,该试验存在对照且治疗组显著优于对照组。应答者与非应答者对比没有意义,就如同在抗癌药试验中,用活着的患者与死了的患者对比,然后得出结论:活着的患者比死了的患者生存期更长且达到统计学显著差异,这种恒真结论与药物有效性无关。2015-04-26:CLDN宣布宣布MYDICAR心衰心衰IIb期试验失败期试验失败宣布失败当日跌80%,这样一种完全无效的疗法是如何拿到优秀的早期临床数据,并且成为第一个获得FDA突破性药物资格的基因疗法?治疗组患者比安慰剂组更健康造成有效假象治疗组患者比安慰剂组更健康造成有效假象Placebo(n=14)Low Dose(n=8)Mid Dose(n=8)High Dose(n=9)Age,mean6160.363.956.6Sex,Female1103NT-proBNP4072135333102141Risk reduction 60%(p=0.11)54%(p=0.12)88%(p=0.03)MYDICAR凭借高剂量组降低88%心衰风险的IIa期数据获得FDA突破性药物资格,但是这一结果没能在IIb期试验中重复出来,原因可能是IIa期试验中高剂量组患者比安慰剂组更健康。(1)年龄越小,心衰风险越小;(2)女性心衰风险更小;(3)NT-proBNP水平越高说明心脏功能越差。