反对美国临床肿瘤学会临床实践指南最新推荐意见？
剧情梗概：去年夏天，全球医学期刊影响因子排名第二的美国《新英格兰医学杂志》发表乳腺癌70基因检测MINDACT研究结果；今年夏天，美国临床肿瘤学会根据MINDACT研究结果更新指南推荐乳腺癌70基因检测；昨天，美国临床肿瘤学会《临床肿瘤学杂志》连发三篇关于MINDACT研究的文章：德国独立研究机构完全否定MINDACT研究和指南推荐；美国自由执业医生集团认为指南用词不当，而且乳腺癌70基因检测不如乳腺癌21基因检测；指南更新专家组细致答复。
2016年8月25日，美国麻省医学会《新英格兰医学杂志》发表的多基因检测用于淋巴结阴性和1～3个淋巴结阳性病变可能避免化疗（MINDACT）前瞻随机研究认为，70基因检测（MammaPrint）有可能使原发性、激素受体阳性、HER2阴性、腋窝淋巴结阳性≤3枚的乳腺癌女性避免化疗，根据这些高度相关性数据，应该对乳腺癌指南和推荐意见进行更新。于是，2017年7月11日，美国临床肿瘤学会《临床肿瘤学杂志》在线发表美国临床肿瘤学会临床实践指南专项更新：使用生物标志指导早期浸润性乳腺癌女性全身辅助治疗决策。
前情提要
2017年12月11日，美国临床肿瘤学会《临床肿瘤学杂志》在线发表德国医疗质量与效益研究所的来信，对早期乳腺癌70基因检测研究结果以及美国临床肿瘤学会临床实践指南后续推荐意见更新进行了解读。
德国医疗质量与效益研究所（IQWiG）并不开展临床研究，而是对临床研究进行独立分析，其职能类似英国的NICE，不接受医药企业资助的研究课题，仅接受联邦卫生部和联邦联合委员会（其职能类似英国的NHS）委托的研究课题，着重分析药物或非药物干预方法的医疗质量和卫生经济学。
对此，IQWiG表示反对MINDACT研究或美国临床肿瘤学会临床实践指南后续推荐意见的说法，即根据生物标志（70基因）决定某些亚组患者免于化疗的策略。理由如下：
MINDACT研究的相关性问题是一个非劣效性问题，即对于临床风险高而基因组风险低的患者，化疗是否没有（相关）获益。对于此类问题，符合方案分析至关重要，该分析也显见于MINDACT研究报告的表2。MINDACT研究的主要终点为无远处转移生存，次要终点为无病生存。但是，任何癌症复发或任何继发癌症发生，对于患者而言都极其重要。因此，应该主要考虑无病生存。根据符合方案分析，对于临床风险高而基因组风险低的亚组患者，不化疗与化疗相比，虽然5年无远处转移生存率从统计学而言相似（94.8%比96.7%，P=0.11），但是5年无病生存率从统计学而言显著较低（90.3%比93.3%，P=0.03），绝对风险相差3%。该数量级差异看起来被研究者认为与MINDACT研究根据无病生存次要终点样本量计算相关。因此，MINDACT数据表明，对于临床风险高而基因组风险低的亚组患者，化疗后5年也有相关获益。既往美国临床肿瘤学会指南认为10年后绝对风险减少2%～3%即有临床相关性；在MINDACT研究中，该值在5年后已经达到。
MINDACT数据还表明，基因表达检测没有提供超出临床风险分类的额外信息：根据用于不化疗患者5年内无病生存事件检出或排除的接受者操作特征（ROC）曲线，个体和汇总结果的估算值几乎相似（预后准确性；图1和表1）。因为临床风险分类也可以被认为是定量检验（评分），所以对于该分类，推测可以通过评分项目组成的适当选择和加权，到达ROC曲线的每个点。不过，需要来自MINDACT研究的原始数据验证该假设。
2017年12月11日，美国临床肿瘤学会《临床肿瘤学杂志》还在线发表了美国麦克森专科医生集团、贝勒萨蒙斯癌症中心、美国肿瘤医生集团、德克萨斯肿瘤医生集团、康巴斯肿瘤医生集团、弗吉尼亚癌症专科医生集团的意见，虽然表示赞成美国临床肿瘤学会指南专家组成员根据MINDACT研究更新生物标志物使用以指导早期浸润性乳腺癌女性全身辅助治疗的决策，但是除了两个问题：
医生集团：新型医疗团队联合执业形式，类似医生经纪公司，与自由执业医生签约，提供医疗相关服务，可能属于医院，也可能为独立民间组织、独立法人机构，以股份制形式运作，大多深入社区，部分还会开展临床研究。
首先，用词不当。关于指南更新的大部分具体推荐意见用词，其中声明医师“应该”或“不应该”在特定情况下使用特定的检测，这超出了指南推荐的范畴，并且成为对临床医生的命令。不过，推荐意见1.2.1声明医师“可以”使用70基因检测的用词，并未使用更有命令口吻的“应该”或“不应该”。
其次，70基因不如21基因。虽然70基因检测可有助于雌激素受体阳性、HER2阴性、1～3枚淋巴结阳性的乳腺癌患者，但是该推荐意见依据来自731例1～3个淋巴结阳性、临床风险高而基因组风险低、被随机分配接受化疗与否的女性亚组前瞻数据。虽然整个组包括1550例临床风险高而基因组风险低、被随机分配接受化疗与否的女性，但是一半患者淋巴结为阴性。此外，该研究随访时间仅5年，后续随访结果可能变化。2010年发表于《柳叶刀》肿瘤学分册的SWOG-8814研究证实，对于1～3个淋巴结阳性且雌激素受体阳性乳腺癌绝经后女性，21基因复发评分对他莫昔芬治疗患者可以预后，并且预测环磷酰胺＋多柔比星＋氟尿嘧啶用于高风险评分患者的显著获益。虽然该回顾分析对两个研究分组927例患者其中40%的肿瘤标本进行，但是长达10年的随访，加强了这些结果的临床影响。2017年发表于《乳腺癌研究与治疗》的西德研究组前瞻研究（PlanB）还支持21基因复发评分用于淋巴结阴性患者和1～3个淋巴结阳性患者。因此，这些临床研究结果支持21基因复发评分用于1～3个淋巴结阳性雌激素受体阳性乳腺癌绝经后女性，反对指南推荐意见1.2关于临床医师不应该使用21基因复发评分指导激素受体阳性、HER2阴性、淋巴结阳性乳腺癌患者术后全身治疗决策的声明。
2017年12月11日，美国临床肿瘤学会、哈佛大学医学院和达纳法伯癌症研究所、华盛顿大学、约翰霍普金斯大学的相关专家对上述两篇文章进行了逐条答复，认为上述两篇文章提出了经过深思熟虑的评论。
首先，关于推荐意见起草用词，尤其术语“应该”和“不应该”的使用。美国临床肿瘤学会使用“决策支持指南”（GLIDES）方法制定其推荐意见。关于定义责任等级（必须、应该、可以）的术语信息参见指南“方法附录”。根据GLIDES方法，诸如“应该”或“可以”之类声明具有特定含义，故按指南方法统一发表。
编者按：指南免责声明强调了指南并不强制任何特定医疗流程。而且，指南并非取代治疗提供者的独立专业判断，因为指南并未考虑患者之间的个体差异。“必须、应该、可以”反映推荐意见置信程度的高、中、低，此类用词表示对于大多数或许多患者推荐或不推荐，但是医师对于治疗个别情况有选择余地。在任何情况下，行动方案选择应该由治疗个别患者的治疗提供者进行考虑。指南信息仅供自愿使用。
其次，关于21基因复发评分对1～3个淋巴结阳性、雌激素受体阳性、HER2阴性乳腺癌绝经后女性的作用。由于指南进行更新时，MINDACT研究结果是当时唯一的新数据，促使美国临床肿瘤学会根据改变实践的研究结果对2016年指南进行专项更新。2010年发表于《柳叶刀》肿瘤学分册的SWOG-8814研究回顾分析已记载于2016年指南。指南更新专家组更迫切地期待RxPONDER研究结果，该研究将1～3个淋巴结阳性和21个基因复发评分≤25的绝经前和绝经后女性随机分配接受现代化疗序贯内分泌治疗或单用内分泌治疗。西德研究组PlanB研究的最新数据在线发表于2017年6月29日，在本次专项更新被接收发表之后；而且由于该研究相关亚组样本量太小而无法改变临床，不符合美国临床肿瘤学会标准。PlanB研究仅入组了110例pN1且21基因复发评分<11且未化疗患者。PlanB研究数据将在以后完整更新时考虑。
最后，IQWiG指出MINDACT符合方案分析对于临床风险高而基因组风险低的患者进行化疗可使无病生存率绝对值提高3%（P=0.03）。不过，MINDACT预设的主要重点分析，基于初试人群的无远处复发生存（无远处转移生存），其与符合方案分析人群略有不同。符合方案分析是预设二次分析的敏感性分析。因此，无病生存二次符合方案分析虽然发现化疗可使无病生存相差3%，但是忽略了该研究所有其他结果，未见与化疗相关的统计学显著获益。指南更新专家组承认，对于比较两个随机分组的非劣效性研究，符合方案分析确实应该与意向治疗分析进行对比。对于非劣效性研究，两种分析均存在偏差，并且对这两种分析之间结局差异的理解很重要。无远处转移生存是MINDACT研究的主要结局指标，通常被认为最有临床意义。无病生存通常被作为无远处转移生存或总生存的替代指标，以缩短研究持续时间，但是指南更新专家组并不同意所有或甚至大多数患者会认为局部复发与远处复发一样重要。由于化疗大部分获益在前5年，总体相差不大，预计也不会变大。但是，指南更新专家组承认（正如指南专项更新所述）更长随访将很重要。IQWiG还提供了一个有趣的ROC分析，根据指南假设了一些参数值。指南更新专家组同意，如果可以获得数据，更完整的分析将很有用。指南更新专家组会推荐一种为生存数据而非二进制结局数据设计的方法。
J Clin Oncol. 2017 Dec 11. [Epub ahead of print]
Interpretation of the Results of the MINDACT Study and Consequent Recommendations in the Updated ASCO Clinical Practice Guideline.
Stefan Lange, Fueloep Scheibler, Daniel Fleer, Juergen Windeler.
Institute for Quality and Efficiency in Health Care, Cologne, Germany.
The 5-year results of the Microarray in Node-Negative and One to Three Positive Lymph Node Disease May Avoid Chemotherapy (MINDACT) study present the first ever prospective, randomized comparison of the effect of a biomarker-based strategy on the decision for or against chemotherapy in women suffering from primary, hormone receptor-positive and human epidermal growth factor receptor 2-negative breast cancer with up to three positive axillary nodes. These highly relevant data should lead to an update of breast cancer guidelines and recommendations. However, we do not agree with the interpretation of the MINDACT study or the subsequent recommendation in the ASCO Clinical Practice Guideline, namely, that a biomarker-based strategy may be used to inform decisions on withholding chemotherapy in certain subgroups of patients. The reasons for this are as follows.
The relevant question in MINDACT, that is, whether chemotherapy in patients with a high clinical but low genomic risk shows no (relevant) benefit, is a noninferiority question. For this type of question, the per-protocol analysis is crucial, and this analysis was also featured prominently in the MINDACT publication (Table 2).
Distant metastasis-free survival was the primary end point of the MINDACT study and disease-free survival (DFS) was a secondary end point. However, any recurrence of cancer or any occurrence of secondary cancer is of paramount importance to patients; therefore, DFS should be considered primarily.
The per-protocol analysis showed a statistically significant difference in DFS to the disadvantage of the high clinical but low genomic risk group, which omitted chemotherapy, with an absolute difference of 3%. A difference of this magnitude was seemingly defined as relevant in the sample size calculation of the MINDACT study for secondary, DFS-based end points. The MINDACT data thus suggest that after 5 years, chemotherapy also shows a relevant benefit in patients with a high clinical but low genomic risk. The previous ASCO guideline considered an absolute risk reduction of 2% to 3% after 10 years as clinically relevant; in the MINDACT study, this value was already reached after 5 years.
The MINDACT data also suggest that the gene expression test provides no information gain over the clinical risk classification: individual and pooled results lie on almost the same estimated receiver operating characteristic curve for the detection or exclusion of DFS events within 5 years in patients without chemotherapy (prognostic accuracy; Fig 1 and Table 1). Because the clinical risk classification can also be considered a quantitative test (score), for this classification, presumably each point on a receiver operating characteristic curve can be reached by an appropriate selection and weighting of components. However, the original data from the MINDACT study would be required to verify this assumption.
PMID: 29227725
DOI: 10.1200/JCO.2017.75.9506
J Clin Oncol. 2017 Dec 11. [Epub ahead of print]
Use of Biomarkers to Guide Decisions on Adjuvant Systemic Therapy in Early-Stage Invasive Breast Cancer.
Joanne L. Blum, Nicholas Robert, Jay Andersen, Anne Favret, Patrick Ward, Cynthia Osborne, John Pippen.
Texas Oncology, Baylor Sammons Cancer Center, US Oncology, Dallas, TX; McKesson Specialty Health/US Oncology Network, The Woodlands, TX; Compass Oncology, US Oncology, Portland, OR; Virginia Cancer Specialist, US Oncology, Fairfax, VA; US Oncology, Cincinnati, OH.
Although we agree with the decision of the ASCO Guideline Panel members to update the use of biomarkers to guide decisions on adjuvant systemic therapy for women with early-stage invasive breast cancer on the basis of the published results of the MINDACT (Microarray in Node-Negative and 1 to 3 Positive Lymph Node Disease May Avoid Chemotherapy) trial, we take exception with two issues. One is the language used in most of the detailed recommendations in the update, which state that physicians "should" or "should not" use specific tests in specific situations. We believe that the use of directive language, such as should and should not, goes beyond the concept of recommended guidelines and becomes a directive to clinicians. We note that recommendation 1.2.1 includes language that states that physicians "may" use the 70-gene signature assay and does not use the more directive should or should not language.
Second, we agree that the 70-gene assay may provide some utility for decision making for patients with breast cancer with estrogen receptor (ER)-positive, human epidermal growth factor receptor 2-negative, lymph node-positive disease with one to three positive lymph nodes. However, we note that this recommendation is based on a small prospective data set of 731 women with lymph node-positive disease with one to three positive lymph nodes and high clinical and low genomic risk who were randomly assigned to receive chemotherapy or not. Although the entire high clinical and low genomic risk group included 1,550 women randomly assigned to receive chemotherapy or not, this group included both patients with positive and patients with negative lymph nodes. The outcome data provided for the entire group as noted in Figure 2 and Table 3 of the report included one half of the patients who had lymph node-negative disease. Moreover, follow-up for this study is still short at only 5 years, and the results may change with later follow-up.
As previously reported by Albain et al, who demonstrated that among postmenopausal women with ER-positive breast cancer and one to three positive lymph nodes treated in SWOG-8814, the 21-gene recurrence score was prognostic for tamoxifen-treated patients and predicted significant benefit of cyclophosphamide, doxorubicin, and fluorouracil among patients with high-risk scores. Although this retrospective analysis was performed for 40% of the tumors in 927 patients in the two trial arms, the long follow-up of 10 years contributed to the clinical impact of these results. The prospective West German PlanB trial also supported the utility of the 21-gene recurrence score among patients with negative lymph nodes and those with one to three positive lymph nodes. We believe that the findings of these clinical trials support the use of the 21-gene recurrence score among postmenopausal women with ER-positive breast cancer with one to three positive lymph nodes and disagree with recommendation 1.2 that states that if a patient has ER/progesterone receptor-positive, human epidermal growth factor receptor 2-negative, node-positive breast cancer, the clinician should not use the 21-gene recurrence score to guide decisions on adjuvant systemic therapy.
PMID: 29227724
DOI: 10.1200/JCO.2017.75.3756
J Clin Oncol. 2017 Dec 11. [Epub ahead of print]
Reply to J.L. Blum et al and S. Lange et al.
Ian Krop, Nofisat Ismaila, William Barlow, Vered Stearns.
Dana-Farber Cancer Institute, Boston, MA; American Society of Clinical Oncology, Alexandria, VA; University of Washington, Seattle, WA; Johns Hopkins University, Baltimore, MD.
In letters to the editor in response to our recent guideline publication on the use of biomarkers to guide decisions on adjuvant systemic therapy for women with early-stage invasive breast cancer in Journal of Clinical Oncology, Blum et al and Lange et al provide thoughtful comments.
Blum et al raised two issues, the first of which was about the language used in the drafting of the recommendation, particularly the use of the terms "should" and "should not." ASCO uses the Guidelines Into Decision Support (GLIDES) methodology to craft its recommendations. Information about terms that define the level of obligation (must, should, may) is in the Methods Supplement that accompanies the guideline. Statements such as should or may have specific meaning according to the GLIDES methodology, hence the copublication of the methods with the guideline.
Second, Blum et al considered the role of the 21-gene recurrence score among postmenopausal women with estrogen receptor-positive, human epidermal growth factor 2-negative breast cancer with one to three positive lymph nodes. At the time the update was conducted, results from MINDACT (Microarray in Node-Negative and 1 to 3 Positive Lymph Node Disease May Avoid Chemotherapy) were the only new data available that constituted a signal on the basis of ASCO's approach to facilitate focused updates as a result of practice-changing studies. The retrospective analysis by Albain et al was addressed in the 2016 version of the guideline. We eagerly await results of the prospective RxPONDER trial in which both pre- and postmenopausal women with one to three positive nodes and a 21-gene recurrence score ≤ 25 were randomly assigned to receive modern chemotherapy followed by endocrine therapy or endocrine therapy alone. The West German Study Group PlanB study comprised new data published after this focused update was accepted for publication; however, the study does not meet ASCO's criteria as a signal because the size of the relevant subpopulation was too small to be practice changing. The PlanB study included only 110 patients with pN1 disease and a 21-gene recurrence score of < 11 in those who received no chemotherapy. The PlanB data will be considered in a full update in the future.
Lange et al point out that one of the per-protocol analyses (PPAs) in MINDACT showed a 3% difference in disease-free survival (DFS) with the addition of chemotherapy in patients with high clinical but low genomic risk (P = .03). However, the prespecified primary analysis in MINDACT was based on survival without distant recurrence (distant metastasis-free survival [DMFS]) in the primary test population, which was modestly different from the PPA population. The PPAs were sensitivity analyses of prespecified secondary analyses. Thus, although a 3% difference in DFS was seen with the addition of chemotherapy in the secondary PPA of DFS, this ignores all the other findings in the study, none of which demonstrated a statistically significant benefit associated with the addition of chemotherapy. We agree that in a noninferiority study that compares two randomized arms, the PPA should be contrasted with intention-to-treat analyses. Both types of analyses suffer from bias in noninferiority studies, and an understanding of the difference in outcomes between the two types of analysis is important. DMFS was the primary outcome in MINDACT and often is considered to be the most clinically important. DFS often is used as a surrogate for DMFS or overall survival to shorten trial duration, but we do not agree that all or even most patients would consider a local recurrence to be as important as a distant recurrence. Overall differences are small and not expected to become larger because most of the benefit of chemotherapy is in the first 5 years. However, we agree (as we stated in the focused update) that longer follow-up will be important.
The authors also provide an interesting receiver operating characteristic analysis that assumes some values for the parameters on the basis of the guideline. We agree that a more complete analysis would be useful if the data could be obtained. We would recommend an approach designed for survival data rather than for binary outcome data.
PMID: 29227726
DOI: 10.1200/JCO.2017.75.8607