Introduction
Methods
Data extracted
LOE | Type of study | Validation |
---|---|---|
I-A | RCT specifically to assess the utility of the biomarker. The samples are collected and analysed in real-time | Not necessary but could be useful |
I-B | RCT not specifically to assess the utility of the biomarker. The samples are stored during the study and analysed after the study is finished, following a protocol | One or more studies with consistent results |
II-B | RCT not specifically to assess the utility of the biomarker. The samples are stored during the study and analysed after the study is finished, following a protocol | Only one study, or several studies with inconsistent results |
II-C | Non-randomized retrospective study aimed to assess the utility of the biomarker using samples from patients in an observational setting (standard treatment and follow-up) | Two or more studies with consistent results |
III-C | Non-randomized retrospective study aimed to assess the utility of the biomarker using samples from patients in an observational setting (standard treatment and follow-up) | Only one study, or several studies with inconsistent results |
IV–V-D | No aspect of the study is prospective | Not necessary because these types of studies do not enable the clinical utility of the biomarker to be assessed |
Results
Reference/study design | Study details | Patients/treatment | Type of specimen fixation/storage | Antibody/controls/counting/cut off/double reading (Y/N) | Results | Conclusions REMARK [71] score/20 |
---|---|---|---|---|---|---|
Ki-67 as a prognostic factor: neoadjuvant chemotherapy | ||||||
EORTC-NCIC-SAKK trial [11] Retrospective analyses from RCT | 12 countries/May 1993–April 1996 No pts: 179/448 (40%) FU: 5.5 years (median) Outcomes: RFS OS | Any T4, any N, M0 or any T, N2/N3, M0 or inflammatory breast carcinoma/CEF; EC+ G-CSF | Pre-treatment samples from primary tumour/No details/FFPE, Bouin Holland-fixed-PE | IHC—MIB-1 (Immunotech)/(1) −ve control slide—no MIB-1 (2) +ve control—breast carcinomas known to contain high levels of Ki-67 Central laboratory/Percentage of cells with clear nuclear staining among 200 tumour cells/≥20%/ND | Univariate: Results for 20% cut-off: PFS: HR (95% CI)—1.22 (0.78–1.91) P = 0.38 OS: HR (95% CI)—1.67 (0.98–2.85) P = 0.06 Results for analyses as a continuous variable: PFS: 1.23 (0.97–1.56) P = 0.09 OS: 1.45 (1.13–A.88) P = 0.004 Multivariate: results for analyses as continuous variable: PFS: P = 0.57 OS: P = 0.32 | Not a statistically significant prognostic factor 13 |
Ki-67 as a prognostic factor: neoadjuvant hormonotherapy | ||||||
Decensi et al. [29] RCT but with other samples | Italy/Sept 1999–Aug 2001 No pts: 116/120 (97%) FU: 7.2 years Outcomes: RFS OS | Early stage ER+ breast cancer, <5 cm, N0–N1, M0/Two low-doses versus standard dose Tam (4 weeks) | Core cut biopsy pre-treatment biopsy at surgery/6–12 h fixation/FFPE) | IHC—MIB-1 (Dako) followed by high-sensitivity detection kit (EnVision Plus-HRP; Dako)/Assessor blinded to treatment group/% of +ve tumour cells in core biopsy or over at least 2,000 cells in surgical biopsy/Quantiles calculated from a data of 6,853 women (ER +ve or PgR +ve) who underwent surgery from 2004 to 2007/No | Univariate: Post-treatment Ki-67 for RFS (HR (95% CI)): 4th quartile (≥30%): 6.05 (2.07–17.65) 3rd quartile (20–29%): 4.37 (1.56–12.25) 2nd quartile (14–19%): 2.92 (0.95–8.96) 1st quartile (<14%): P-trend = 0.001 Invasive disease recurrent per point increase: Pre-treatment Ki-67: 2.2 (0.9–5.0) Post tamoxifen Ki-67: 5.0 (2.3–7.76) P-trend = 0.076 OS: Post tamoxifen Ki-67 ≥20% versus <20%: 5.5 (1.3–23.2) P-trend = 0.006 Multivariate: RFS per point increase in: Pre-treatment Ki-67: 2.2% (0.99–4.7) P = 0.076 Post-treatment Ki-67: 5.0% (2.3–7.7) P < 0.001 OS per point increase in: pre-treatment Ki-67: 4.52% (0.9–9.6) P = 0.066 Post-treatment Ki-67: 5.7% (1.1–10.4) P-trend = 0.014 | Ki-67 response after short-term neoadjuvant tamoxifen is a good predictor of RFS and OS 12 |
RCT with planned sub-study | Multinational/Mar 1998–Aug 1999 No pts: 158/337 (69%) FU: 61.2 months (median) Outcomes: RFS BCSS | ER+, T2-4a-c, N0–2, MO breast cancers/LET-Tam 4 months prior to scheduled surgery | Core biopsy pre-treatment and last visit (before surgery) Samples shipped at ambient temperature 10% buffered formalin/FFPE | Ki-67 antibody (Zymed) ABC detection/assessors blind to patient ID, treatment and outcomes −ve control: no primary Ab +ve control/% +ve stained tumour cell among 200–1,000 cells/Ki-67 per 2.7 fold increase (natural log intervals)/Yes | Univariate: RFS (HR (95% CI)): Post-treat Ki-67 per 2.7× increase: 1.4 (1.2–1.6); P < 0.001 BCSS: 1.4 (1.1–1.7); P = 0.009 Multivariate: RFS: Post-treat Ki-67 per 2.7× increase: 1.3 (1.1–1.5); P = 0.01 BCSS: 1.4 (1.1–1.8); P = 0.02 | Post-treatment Ki-67 is independently associated with RFS and BCSS 16 |
Ki-67 as a prognostic factor: adjuvant chemotherapy | ||||||
EORTC 10854 [69] RCT—retrospective selection of some patients | Multinational/May 1986–March 1991 No pts: 441/674 (65%) FU: 82 months (median) Outcomes: OS, DFS, MFS | N0, invasive breast cancer, stage I-IIIA/Surgery (all) treatment group: FDC—1 cycle | Tumour samples ND/FFPE | IHC—MIB-1 (Immunotech; ref cited)/one reference lab/‘hot spot’ at high magnification/≥20%/ND | Univariate: OS: P < 0.001 DFS: P < 0.01 MFS: P < 0.001 Multivariate: Ki-67 not independent variable in model | Mitotic index was a better prognostic indicator in multivariate analysis than Ki-67 (or S-phase) 10 |
PACS01 [83] RCT—planned sub-study | France/Jun 1977–Mar 2000 No pts: 1,190/1,999 (60%) FU: 58.7 months (median) Outcomes: DFS | Stage <T4a, ER +ve, N+/FEC—6 cycles versus FEC-D—3 cycles) | Tumour blocks ND/ND | IHC—MIB-1 (Dako) using Ventana NeXes automat/centralised lab/visual grading system; estimated % of +ve cells/>20%/yes, assessment by 10 pathologists | Univariate: DFS: 1.7 (1.2–2.4) P = 0.002 Docetaxel efficacy for relapse: ER +ve/Ki-67 +ve—0.5 (0.3–1.0) ER +ve/Ki-67 −ve—1.0 (0.9–1.6) HR for interaction with docetaxel—0.5 (0.2–1.2) STEPP analysis for 5-year DFS shows maximum benefit for highest Ki-67 values Multivariate: Model with only biomarkers: 1.6 (1.2–2.3) P = 0.01 Model with treatment, biomarkers, clinical characteristics: 1.5 (1.0–2.22) P = 0.046 | Ki-67 is a candidate for predicting docetaxel efficacy in ER +ve breast cancer 18 |
RCT—Retrospective | UK—Oct 1996–Apr 2001 No pts: NEAT: 1,623/2,021 (80%); BR9601: 318/370 (86%) FU: 48 months (median) Outcomes: RFS, OS | Completely excised early breast cancer (N+ and N0)/E-mod-CMF (n = 183) mod- CMF (n = 191) | Triple TMA prepared from stored tissue blocks ND/ND | IHC/ND/Scoring by one experience observer, blinded to patient ID and outcome/13%/ND | Univariate: Ki-67 +ve versus Ki-67 −ve: RFS: HR = 1.12 (95% CI: 0.95–1.32) P = 0.19 OS: HR = 1.11 (95% CI 0.93–1.33) EPI-CMF versus CMF RFS: Ki-67 high: 30.3% versus 36.5%—0.78 (0.59–1.01) Ki-67 low: 28.3% versus 34.4%—0.77 (0.55–1.08) Overall: 29.5% versus 35.6% (0.77 (0.66–0.90), P = 0.001, P-interaction = 0.95 OS: Ki-67 high: 24.7% versus 29.9%—0.78 (0.58–1.05) Ki-67 low: 24.1% versus 27.7%—0.82 (0.56–1.19) Overall: 24.5% versus 29.0% (0.80 (0.67–0.95), P = 0.01, P-interaction = 0.80 Multivariate: ND | Ki-67 is strongly prognostic but not predictive of additive benefit from EPI-CMF versus CMF 13 |
Mottolese et al. [77] Samples from 1 centre from RCT | Rome, Italy/1991–1993 No pts: 157/506 (31%) FU: 37 months (median); 4–88 months (range) Outcomes: DFS, OS | Premenopausal women: invasive breast cancer >1 cm grade 2–3, any N and HR status Postmenopausal women: same but ER/PgR negative Adjuvant chemotherapy G-CSF versus EC | Tumour samples/ND/PE | IHC polyclonal Ki-67 (DAKO)/+ve control: breast cancer with known high level −ve control: no primary Ab/+ve nuclei in four random fields of ≥200 cells/≥10%/ND | Univariate: High versus low Ki-67: DFS: RR = 1.52 (0.82–2.82); P = 0.18 OS: RR = 1.83 (0.91–3.70); P = 0.08 Multivariate: ND | Ki-67 not a significant prognostic factor 9 |
Ki-67 as a prognostic factor: adjuvant hormonotherapy | ||||||
BIG 1–98 [108] Retrospective tissue collection from patients in RCT | International/Mar 1998–Mar 2000 No pts: 2,685/4,922 (55%) FU: 51 months (median) Outcomes: RFS; OS | Early invasive breast cancer, ER +ve ± PgR +ve (N+ and N0)/Tam (n = 1,361) or Let (n = 1,324) | Primary tumour samples (whole tissue sections) ND/PE | IHC—MIB-1 (Dako) Cut and stained centrally with automated immunostainer (Autostainer, Dako)/central review/% of +ve cells from 2,000 tumour cells, in randomly selected fields at the periphery of tumour/>11%/ND | Univariate: DFS: 1.8 (1.4–2.3) P = 0.0001 Multivariate: DFS adjusted for age, PgR status, tumour size, tumour grade, nodal status, HER-2 status and peritumoural vascular involvement: 1.4 (1.1–1.9) P = 0.02 | Ki-67 confirmed as prognostic factor 12 |
IKA TAMOXIFEN [72] Random sample of patients in RCT | Netherlands/1982–1994 No pts: 394/1,662 (24%) FU: ~10 years (median) Outcomes: DFS | Postmenopausal, T1–4N0–3M0/1st year: Tam versus no treatment 2nd–3rd year: Tam group randomised to stop or continue 2 years | Tumour tissue Fixed >24 h in neutral buffered 4% formaldehyde/PE on silane coated slides | IHC MIB1 (Immunotech)/−ve control: no primary Ab/scanning for +ve cells (distinct nuclear staining) at medium and high resolution/>5%/ND | Univariate: High versus low Ki-67: DFS: log-rank P = 0.0023; unadjusted HR = 2.069 (1.284–3.333); adjusted Multivariate: High versus low Ki-67: DFS: HR = 1.717 (0.992–2.969); P = 0.0533 | No conclusions for Ki-67 10 |
Ki-67 as a prognostic factor: adjuvant chemo-hormonotherapy | ||||||
IBCSG TRAILS VIII and IX [109] Two RCTs—retrospective collection of samples | International/1988–1999 No pts: 1,924/2,732 (70%) FU: 10 years (median) | N0 invasive tumour in premenopausal (Trial VIII) and postmenopausal (Trial IX) women/Trial VIII: Gos versus CMF versus CMF +Gos Trial IX: CMF/Tam versus Tam | Primary tumour samples ND/FFPE | IHC—MIB-1 (Dako)/central laboratory—blinded to treatment and outcomes/% of definite +ve cells among 2,000 tumour cells in randomly selected high-power (×400) fields at the periphery of the tumour/≥19%/ND | Univariate: Trial VIII DFS (high versus low Ki-67) (HR (95% CI)): 1.66 (1.20–2.29); P = 0.002 Trial IX DFS (high versus low Ki-67) (HR (95% CI)): 1.60 (1.26–2.03); P < 0.001 Multivariate: Trial VIII DFS (adjusted for other tumour features): P < 0.05—independent of treatment received Trial IX DFS (adjusted for other tumour features): P < 0.05—independent of treatment received | Ki-67 labelling index is an independent prognostic factor but not a predictive factor 12 |
Ki-67 as a prognostic and predictive factor: neoadjuvant chemotherapy | ||||||
Chang et al. [20] Retrospective analyses of some patients from RCT and some others who received same treatment | UK—Feb 1990–Aug 1995 No pts: 109/158 (131 from RCT + 27 other patients who received same treatment) FU: 48 months (median) Outcomes: GCR (CR or min residual disease) at 3 months, OS, RFS, Change in Ki-67 expression from pre-treat to day 10 or day 21 after 1st course of treatment | Operable (including T4) breast cancer/Mit-M ± Mi | Fine needle aspirate ND/Cytospin slides: air-dried and stored at −80°C | MIB -1 (Dako) with biotin-anti-mouse IgG and avidin–biotin-peroxidase complex/Experience assessor blinded to patient ID and outcome/change in Ki-67 between pre-treatment and day 10 or 21 after 1st chemotherapy/Continuous/No | Univariate: Decrease in Ki-67 expression: GCR at 3 months: 2.3 (95% CI 0.9–6.0) higher (P < 0.05) Pretreatment Ki-67: GCR: 1.0 (95% CI 0.8–1.2) RFS: 1.7 95% CI (0.6–5.2) OS: 2.0 (95% CI 1.0–3.8) Multivariate: Pre-treatment Ki-67 expression not statistically significant for RFS or OS | Change in Ki-67 expression is predictive of achieving GCR which seems to be a valid surrogate marker for survival 13 |
Chemotherapy arm of RCT | Bordeaux, France/Jan 1985–Apr 1988 No pts: 128/134 (96%) FU: 124 months (median); 47–148 months (range) Outcomes: Response (complete or ≥50% tumour regression), OS, DFS, MFS | Operable tumour >3 cm/3 cycles; EViMi and 3 cycles, MTV | Core biopsy before randomisation ND/FFPE | IHC—MIB-1 (Immunotech), ABC complex/Objective % of +ve tumour cells, semi-quantitative from 0 to 100%/>40% (75th percentile)/No | Univariate: Predictive for tumour response : 4.1 (1.4–11.5) P = 0.007 Prognostic: OS: 77.8% versus 81.5%—NS DFS: 56.6% versus 63.0%—NS MFS: 63.6% versus 77.8%— = 0.05 Multivariate: Ki-67 remained independent predictive factor but not prognostic | High Ki-67 was associated with responsiveness to chemotherapy Ki-67 was only statistically significantly associated with metastasis-free survival 11 |
Ki-67 as a prognostic and predictive factor: neoadjuvant hormonotherapy | ||||||
RCT—unplanned exploratory analyses | UK and Germany/Oct 1997–Oct 2002 No pts: 174/330 (53%) FU: 37 months (median); 4–88 months (range) Outcomes: Objective response (values and change in Ki-67 expression baseline to 2 weeks) RFS | ER+ invasive operable, or locally advanced, no evidence of metastasis, N0/(median duration: 30 months) (n = randomised/biopsy available/per protocol) ANA (n = 113/98/86) Tam (n = 108/98/88) ANA + Tam (n = 109/96/85) | Core cut biopsy: pre-treatment and at 2 weeks (not obligatory), excision biopsy at surgery/24–48 h fixation/FFPE | IHC—MIB-1 (Dako)/ND/% of Ki-67 +ve tumour cells scored across 1,000 cells/Ki-67 expression per 2.7-fold increase (geometric mean percentage change from baseline)/ND | Univariate: RFS: baseline Ki-67, per 2.7× increase: 1.85 (1.06–3.22) P = 0.03 2-week Ki-67 expression, per 2.7× increase: 2.09 (1.41–3.08) P < 0.001 Multivariate: RFS: 2-week Ki-67 expression, per 2.7× increase: 1.95 (1.23–3.07) P = 0.004 | Ki-67 level at 2 weeks is a better predictor of RFS than pre-treatment levels 17 |
Ki-67 as a predictive factor: neoadjuvant chemotherapy | ||||||
Learn et al. [60] RCT with data collected prospectively during trial | ND/Feb 1996–Aug 2000 No pts: 121/144 (84%) FU: ND Outcomes: CR, PR | Invasive breast cancer, T1C-T3, N0, M0 or T1–T3, N1, M0/C-A-D | Pretreatment fine-needle aspiration or core biopsy ND/ND | IHC—MIB-1 (Dako Cytomation)/ND/+ve cells among 200 tumour cells/Continuous/Yes | Univariate: No association between Ki-67 for CRR Multivariate: No association between Ki-67 for CRR | No statistically significant association with CRR 8 |
Ki-67 as a predictive factor: neoadjuvant chemo-hormonotherapy | ||||||
Bottini [13] RCT (planned) | Italy—Jan 1997–Jan 2002 No pts: 210/211 (99.5%) FU: ND Outcomes: CRR, Changes in Ki-67 expression before treatment and at definitive surgery | T2–4, N0–1, M0/E versus E-Tam | Incision biopsy ND/ND | IHC—MIB-1 (Dako) with biotin-anti-mouse IgG and avidin–biotin-peroxidase complex/−ve control—no MIB-1; +ve control—known sample with high Ki-67 expression/% of +ve stained tumour cells (≥1,000 cells) across several representative fields iwth 10 × 10 graticule/3 categories: <10%, 11–29%, >30% | Univariate: CCR versus not: median 23.5% (range 7–90%) versus 16% (range 1–90%) P < 0.01 PCR versus not: median 30.0% (range 7–90%) versus 18% (range 1–90%) P < 0.01 Lower Ki-67 expression at post-operative residual histology in E-TAM group but no difference in response rate (P = 0.0041) Multivariate: Ki-67 only independent variable for complete pathological and clinical responses (P = 0.005 and P = 0.006, respectively) | Baseline elevated Ki-67 expression is associated with greater chance of PCR. E-Tam did not improve clinical response but reduced Ki-67 expression, compared with E alone 6 |
Generali et al. [46] RCT | Single centre/Nov 2000–Jan 2004 No pts: 114/114 (100%) FU: ND Outcomes: CCR, PCR, NCR | ER+, T2–4, N0–1/LET versus LET-C | Whole tumour sections taken at diagnosis ND/FFPE | IHC—MIB-1 (Dakopatts) Biotinylated horse antimouse IgG and avidin–biotin-peroxidase complex (Vectastatin ABC kit; Vector Laboratories)/−ve control—no MIB-1; +ve control—known breast tumour with high Ki-67 expression/% +ve stained tumour cells (≥1,000 cells) across several representative fields/≥10%/yes, Rescoring of 10 slides by 2nd investigator | Univariate: Post-treatment Ki-67 –significant inverse correlation with clinical response: NR versus PR versus CR χ2 = 10.85, P = 0.001 Multivariate: ND (skewed distribution) | No conclusion for Ki-67 12 |
GPAD-GBGCS trial [110] RCT with prospectively defined biomarker outcomes | Germany (56 centres)/Apr 1998–Jun 1999 No pts: 196/250 (78%) FU: ND Outcomes: CR | Operable T2–3 (≥3 cm), N0–2, M0/ddAT ± Tam | Core cut needle or incisional biopsy, and surgical sample ND/FFPE, ICH staining within 1 week after mounting on slides | IHC—MIB-1 (Dianova) + automated capillary gap Dako kit, staining with AEC/ND/Semi-quantitative assessment of % of stained cells/3 categories of proliferation activity: low: 0–15%; medium: 16–30%; high: 31–100%/yes | Univariate: PCR: Ki-67 ≤ 15%: 3/42; > 15%: 14/56: 0.32 (0.09–1.15) Multivariate: PCR: 0.43 (0.11–1.61), P = 0.208 | Ki-67 was not an independent predictive factor 10 |
Samples from randomised clinical trials
Samples from cohort and case–control studies
Meta-analyses
de Azambuja et al. [27] | Stuart-Harris et al. [100] | |
---|---|---|
Publication year | 2007 | 2008 |
Period for literature search | Up to May 2006 | January 1995–September 2004 |
Exclusion criteria | Non-English publications | Non-English publications Studies with fewer than 100 patients |
Number of studies identifieda
| 46 | 43 |
Included in DFS analysis | 38 | 20 |
Included in OS analysis | 35 | 19 |
Inclusion of studies for meta-analyses | Studies that provided an HR or data that enabled the HR to calculated | Only studies that provided an HR for either OS or DFS, in either univariate or multivariate analysis; if no 95% CI it was calculated |
Reference | Factors studied | Outcome | Results |
---|---|---|---|
Analysis: number of studies (number of patients) | |||
Search strategy described (yes/no) | |||
Date range number of studies identified (number of patients) | |||
de Azambuja et al. [27] | Ki-67 Yes Up to May 2006 Identified: 68 studies (? patients) | DFS | |
All studies: 38 studies (10,954 patients) | Fixed effect HR: 1.88 (1.75–2.02)
P-heterogeneity = 0.01 Random effect HR: 1.93 (1.74–2.14) | ||
Node negative: 15 studies (3,370 patients) | Fixed effect HR: 2.20 (1.88–2.58)
P-heterogeneity = 0.03 Random effect HR: 2.31 (1.83–2.92) | ||
Node positive: 8 studies (1,430 patients) | Fixed effect HR: 1.59 (1.35–1.87)
P-heterogeneity = 0.68 | ||
Node negative (untreated): 6 studies (736 patients) | Fixed effect HR: 2.72 (1.97–3.75)
P-heterogeneity = 0.89 | ||
OS | |||
All studies: 35 studies (9,472 patients) | Fixed effect HR: 1.89 (1.74–2.06)
P-heterogeneity <0.001 Random effect HR: 1.95 (1.70–2.24) | ||
Node negative: 9 studies (1,996 patients) | Fixed effect HR: 2.19 (1.76–2.72)
P-heterogeneity = 0.001 Random effect HR: 2.54 (1.65–3.19) | ||
Node positive: 4 studies (857 patients) | Fixed effect HR: 2.33 (1.83–2.95)
P-heterogeneity = 0.44 | ||
Node negative/positive (untreated): 2 studies (238 patients) | Fixed effect HR: 1.79 (1.22–2.63)
P-heterogeneity = 0.36 | ||
Stuart-Harris et al. [100] | Ki-67, mitotic index, PCNA, LI Yes January 1995–September 2004 Identified: 43 studies (15,790 patients) | DFS | |
Univariate analysis: 15 studies (?) | Unadjusted HR: 2.18 (1.92–2.47) P < 10−5
P-heterogeneity = 0.21
P-publication bias = 0.002 Adjusted HR (4 studies added): 2.05 (1.80–2.33) | ||
Multivariate analysis: 14 studies (?) | Unadjusted HR: 1.84 (1.62–2.10) P < 10−5
P-heterogeneity = 0.93
P-publication bias = 0.019 Adjusted HR (5 studies added): 1.76 (1.56–1.98) | ||
OS | |||
Univariate analysis: 12 studies (?) | Unadjusted HR: 2.09 (1.74–2.52) P < 10−5
P-heterogeneity = 0.037
P-publication bias = 0.074 Adjusted HR (4 studies added): 1.88 (1.55–2.27) | ||
Multivariate analysis: 13 studies (?) | Unadjusted HR: 1.73 (1.37–2.17) P < 10−5
P-heterogeneity <10−5
P-publication bias = 0.001 Adjusted HR (5 studies added): 1.42 (1.14–1.77) |
Narrative reviews
Discussion
Referencea
| Marker | HR (95% CI) |
---|---|---|
Stuart-Harris et al. [100] | Ki-67 | 1.76 (1.56–1.98) |
Rakha et al. [94] | SBR grade (3 vs. 1) | 1.6 (1.3–2.0) |
Look et al. [64] | uPA/PAI-1 (pN0) | 2.37 (1.78–3.16) |
Rakha et al. [94] | Node status | 1.5 (1.4–1.7) |
Wirapati et al. [113] | ER (neg. vs. high) | 2.2 (1.6–3.0) |
Blows et al. [10] | HER2 | 1.55 (1.23–1.96) |