Ki67抗原是1983年Gerdes等人首次发现的一种核抗原。研究证实，在浸润性乳腺癌中，Ki67与肿瘤细胞增殖及侵袭转移密切相关。Ki67是重要的乳腺癌预后指标，其高水平表达可反映肿瘤细胞的增殖和侵袭能力强，恶性程度高。研究发现Ki67还可以作为乳腺癌内分泌治疗、化疗等的疗效预测指标。IMPACT、P024、MA27、Z1031、ATAC、BIG1-98等研究都显示Ki67降低可作为（新辅助）内分泌治疗的疗效判断终点。近几年，ki67还被用做乳腺癌分子分型界定LuminalＡ型与Ｂ型的重要因素之一。然而，由于ki67值受太多因素影响，目前为止全世界尚无统一的判读和评价指南，甚至在不同的临床研究中，在不同的检测机构中，所取用的ki67临界值都相差较大，低至1%，高达35%。

　　第10届欧洲乳腺癌大会（EBCC）会议中，国际乳腺癌Ki67协作组(International Ki67 in Breast Cancer Working Group)的一篇摘要吸引了大家对ki67的关注。该研究将空芯针穿刺获取的ER+，?HER2?乳腺癌组织集中进行Ki67染色和评分，在英国皇家马斯登医院(RMH)检测ER、PR，所有Ki67染色玻片被依次送到11个国家的21个实验室，对每份样品选择4个既定的代表肿瘤异质性的判读视野，每个视野包含100个肿瘤细胞，使用整体评分方法（the Global Scoring method）得出Ki67评分。以皇家马斯登医院(RMH)作为中心对照组，其它实验室作为实验组，用相关性分析(ICC)计算RR值，将RR值分为低、中、高三个风险级别。另外，将当地评分的Ki67和中央检测的PR结果结合用于界定LuminalＡ型与Ｂ型。

　　结果显示，相关性分析(ICC)整体评分方法所得Ki67的RR为0.88(95%CI：0.81-0.93)。当把ki67整合进入免疫组化及临床因子风险评估模型中（IHC4+C），所得的RR为0.99(95%CI：0.99-1.00)。风险等级吻合度高达414/420(98.6%)。另外，不同实验室所界定的分子亚型（LuminalＡ型和Ｂ型）与中心实验室界定的分子亚型符合程度也很高，因定义不同或参考的依据不一，吻合度介于347/399(87.0%)与381/418(91.1%)之间。

　　尽管摘要所提供的研究资料细节有限，但至少可以看出，不同实验室之间的ki67值差异客观存在，如果使用整体评分方法（the Global Scoring method）评分，将ki67整合进入免疫组化及临床因子风险评估模型中（IHC4+C）进行风险评估，能最大程度地降低甚至消除因Ki67的实验室之间的差异所导致的对风险评估的影响。事实上，正是因为不同实验室之间的ki67检测值差异较大，一些涉及ki67的临床研究会得出不一样甚至完全相反的结论。比如PACS01及BCIRG001研究认为Ki67有化疗疗效预测价值，而IBCSGVIII和IX的结论为否。

　　这就要求我们在解读这些研究结果或者设计临床研究时要充分考虑到影响Ki67值的因素：①肿瘤组织：取材方式（TMAs最佳 ）、取材位点（热点及肿瘤浸润边缘优选）、取材至放入浸泡液的时间（小于1.5小时）、浸泡液（中性福尔马林优选）、蜡块在室温保存的时间（小于14天）；②实验操作：抗体（MIB1最佳，SP6有待证实）、至少3个高倍视野、计数500至1000个肿瘤细胞、阳性判读（只有细胞核染色才视为阳性）；③先进技术和方法：检测结果的统计矫正和校准、数字影像分析系统的应用等。

　　聂建云，主任医师。博士，硕士研究生导师，加拿大麦吉尔大学博士后。云南省肿瘤医院乳腺二科副主任；中国抗癌协会乳腺癌专业委员会青年委员；中国医药教育协会乳腺疾病专业委员会委员；云南省医学会肿瘤学分会乳腺学组副组长；云南省抗癌协会乳腺癌专业委员会常委、秘书；云南省医师协会肿瘤转化医学分会委员

研究摘要

Between-lab variability in Ki67 scoring by a standardised method in core-cuts has little impact on risk estimates by the IHC4+Clinical (IHC4+C) Score.A study presented on behalf of the International Ki67 in Breast Cancer Working Group of the Breast International Group

Poster Spotlight: A. Dodson (United Kingdom)

Background: immunohistochemical (IHC) assessment of proliferation using Ki67 in breast cancer has been shown to associate with prognosis， surrogate luminal?A/B calling (with progesterone receptor， PgR) and is a marker of residual risk of recurrence (RR) for ER+ cases treated with endocrine therapy. Much attention has been paid to Ki67’s analysis and reporting but until recently modest progress made. The International Ki67 Working Group has reported that a standardised “Global” scoring method can reduce inter-observer variability when assessing Ki67 in breast cancer core biopsies but despite meeting pre-specified success criteria， inter-laboratory variability remained with potential for clinically significant discordance between risk categorisation based on single cut-points. Integration of Ki67 with other prognostic factors into a single molecular-clinicopathologic algorithm leads to a risk-predictor less sensitive to variability in any single component. In particular， combination with ER， PgR， HER2 and clinicopathological(C) parameters in the IHC4+C algorithm produces RR scores with analytical and clinical validity for aiding in chemotherapy decision making. We determined the effects of between-lab variability in use of the Global method on IHC4+C reproducibility and in defining luminal subtypes according to St?Gallen consensus definitions.

Material and Methods: Ki67 data were derived from Phase 3 of the International Ki67 in Breast Cancer Working Party program of standardisation. Core biopsies from 20 ER+，?HER2? breast cancers were centrally stained and scored for Ki67， ER and PgR at Royal Marsden Hospital (RMH) using established standardised methods. The Ki67 stained sections were circulated to 21 labs in 11 countries， and scored using the Global method， comprising selection of 4 fields representative of overall heterogeneity (where present) and scoring of 100 cells in each field. Intraclass correlation (ICC) was calculated for RR scores. RR scores were categorised as low/intermediate/high risk and analysed for agreement between RMH (reference centre) and other centres for risk-category assignment. Local Ki67 and central PgR data were combined in agreement tables to examine luminal?A/B calling.

Results: ICC for the Phase3 Global method Ki67 scores was 0.88 (95%?CI: 0.81–0.93). When integrated into the IHC4+C algorithm ICC for RR was 0.99 (95%?CI: 0.99–1.00); risk category agreement level was 414/420 (98.6%)， both indicating almost perfect agreement. Agreement for intrinsic subtype calls varied between 347/399 (87.0%) and 381/418 (91.1%) dependent on definition and archetype used.

Conclusions: Residual inter-lab variability in Ki67 when scored by the Global method has negligible effect on estimates of RR when integrated into the IHC4+C algorithm.