Appendix 4 to NZMJ Herceptin - Clinical#2 (text extract)

This is the text extract for Appendix 4 to NZMJ Herceptin - Clinical#2, browse documents here.

Appendix Four. Clinical effectiveness

Contents

1. Treatment regimens and available trial data ............................................................................................... 1 Treatment regimes in randomised controlled trials.................................................................................... 1 Trial outcomes as reported....................................................................................................................... 10 2. Increasing uncertainty around sequence and durations ............................................................................ 13 Non-publication of data for 12 month sequential treatment from N9831................................................ 13 Initial HERA results overstate effects on disease free survival ............................................................... 16 Impact of N9831 Arm B and the updated HERA results on disease free survival benefits of 12 month sequential treatment ................................................................................................................................. 17 The impact of sequencing on the effectiveness of 12 month treatment ................................................... 19 The impact of sequencing on the cardiotoxicity of 12 month treatment.................................................. 21 Durability of response.............................................................................................................................. 21 3. Concurrent treatment pre-anthracyclines (FinHer and E2198)................................................................. 26 FinHer...................................................................................................................................................... 26 Reliability and validity of the FinHer trial and nine-week concurrent treatment..................................... 27 Patient numbers................................................................................................................................... 28 Overall survival................................................................................................................................... 30 Data from other clinical trials (E2198) ............................................................................................... 31 4. The SOLD study....................................................................................................................................... 33 References and endnotes .............................................................................................................................. 36

1. Treatment regimens and available trial data Clinical effectiveness is one of PHARMAC’s decision criteria (http://www.pharmac.govt.nz/pdf/231205.pdf ‘The clinical benefits and risks of pharmaceuticals’). Treatment regimes in randomised controlled trials To date there have been five open-label randomised controlled trials (RCTs) reporting outcomes for adjuvant trastuzumab given in addition to standard chemotherapy treatment against standard chemotherapy treatment alone in HER2-positive early breast cancer 1—HERA2 3 4, NASBP B315 6 , NCCTG N98315 7, BCIRG 0068, and FinHer.9 (A sixth study, ECOG E219810, which compared 12 months with 10 weeks trastuzumab given concurrently with paclitaxel, was presented as a poster at the San Antonio Breast Cancer Symposium in 2006; however this study has not reported outcomes against standard chemotherapy treatment.) In these studies, trastuzumab was investigated in one of two broad treatment regimens (Figure 1): 1. ‘sequential treatment’—trastuzumab for 12 months following completion of chemotherapy (anthracycline +/- taxane): HERA and NCCTG N9831 Arm B; 2. ‘concurrent treatment’—trastuzumab for 9-10 weeks, or 12 months, started in combination with taxane chemotherapy (either preceding or following completion of anthracycline chemotherapy): NASBP B31, NCCTG N9831 Arm C, BCIRG 006, and FinHer.

Figure 1. Clinical trials of trastuzumab

13,609 patients in all Clinical trials

trastuzumab treated (8724)

Sequential treatment (post-chemotherapy) Data from 2679 patients reported

Concurrent therapy (with chemotherapy) Data from 4351 patients reported

HERA 1 year treatment arm 1694 patients (12 months Post-chemo)

HERA 2 year treatment arm 1694 patients (not yet reported)

N9831 arm B

N9831 arm C

B31

BCIRG 006

Finher

E2198

985 patients 12 months post-chemo

840 patients 52 weeks

1019 patients 52 weeks

2149 patients 52 weeks

116 patients 9 weeks

227 patients 52 weeks or 10 weeks

Key to graphic: √ statistically significant improvement in disease-free survival (DFS) × no statistically significant improvement in DFS ? results awaited Notes to graphic: HERA patient numbers derive from the first (12 month median f/u) interim analysis (Piccart-Gebhart 2005), where n=1694 patients in the 1 year trastuzumab arm and n=1693 in the standard treatment arm; this compares with n=1703 and n=1698 for those respective arms reported in the second (23 month median f/u) interim analysis (Smith 2007). BCIRG 006 includes both Arm AC-TH (post anthracycline treatment, n=1074) and Arm TCH (concurrent trastuzumab with docetaxel and carboplatin, no anthracycline,n=1075). Includes study ECOG E2198.

Trastuzumab treatment regimens in these studies varied in their association with taxanes (sequential to, concurrent with), duration of trastuzumab (between 9 weeks and 2 years), sequence relative to anthracycline chemotherapy and weekly or every three-weekly dosing (Figure 2). Figure 2. Anthracycline-containing treatment regimens in trials of trastuzumab in early stage HER2-positive breast cancer

1) Sequential

SEQUENTIAL HERA, N9831 Arm B

2) Concurrent

CONCURRENT POST ANTHRACYCLINE N9831 Arm C, B31, BCIRG 006 arm AT-CH CONCURRENT PRE ANTHRACYCLINE FinHer, E2198 Arm B

CONCURRENT PRE ANTHRACYCLINE PLUS 12 MONTH DURATION

E2198 Arm C

Key: taxane chemotherapy anthracycline-containing chemotherapy trastuzumab

Broad schematic only. Other regimens, not illustrated, include concurrent use of trastuzumab with docetaxel and carboplatin chemotherapies in BCIRG 006 (no anthracycline chemotherapy), and 2 years’ sequential trastuzumab (following anthracycline and then taxane chemotherapies) in HERA (no data available to date). HERA permitted the use of a variety of different combinations of chemotherapy, including non-anthracycline regimens using CMF (http://content.nejm.org/cgi/data/353/16/1659/DC1/1).

Specific treatment regimes studied in the trials were as follows:

• HERA—a three-arm open-label RCT comparing 12 months and 24 months sequential with standard chemotherapy: • control (n=1693-169811): Observation alone following completion of standard neoadjuvant or adjuvant chemotherapy • 1 Year arm (n=1694-170311): 12 months sequential trastuzumab treatment following completion of standard neoadjuvant or adjuvant chemotherapy • 2 Year arm (n=1694-170111): 24 months sequential trastuzumab treatment following completion of standard neoadjuvant or adjuvant chemotherapy NCCTG N9831—a three-arm open-label RCT comparing sequential and concurrent trastuzumab treatment with standard chemotherapy: • Arm A (control) (n= 979): 4 cycles of AC treatment followed by 12 weeks paclitaxel • Arm B (12 months sequential trastuzumab treatment) (n=985): 4 cycles of AC treatment followed by 12 weeks paclitaxel and then 52 weeks trastuzumab • Arm C (12 months concurrent trastuzumab treatment) (n=840): 4 cycles AC followed by 12 weeks paclitaxel; trastuzumab started on day 1 of paclitaxel and continued for 12 months. NASBP B31—a two-arm open-label RCT comparing concurrent trastuzumab treatment with standard chemotherapy: • Control (n=1024): 4 cycles of AC treatment followed by 12 weeks paclitaxel • 12 months concurrent trastuzumab treatment (n=1019): 4 cycles AC followed by 12 weeks paclitaxel; trastuzumab started on day 1 of paclitaxel and continued for 12 months BCIRG 006—a three-arm open-label RCT comparing concurrent trastuzumab treatment (with 2 different chemotherapy regimens) with standard chemotherapy: • Arm AC-T (control) (n=1073): 4 cycles of AC treatment followed by 4 cycles docetaxel • Arm AC-TH (12 months concurrent trastuzumab) (n=1074): 4 cycles of AC treatment followed by 4 cycles docetaxel; trastuzumab started on day 1 of docetaxel and continued for 12 months • Arm TCH (12 months concurrent trastuzumab) (n=1075): 6 cycles docetaxel and carboplatin; trastuzumab started on day 1 of docetaxel/carboplatin and continued for 12 months

•

FinHer—an open-label RCT comparing docetaxel with vinorelbine for the adjuvant treatment of early stage breast cancer (n=1010). For the subset of women with HER2-positive cancers, patients were further randomised to four arms comparing concurrent trastuzumab (with 2 different chemotherapy regimens— docetaxel or vinorelbine) with docetaxel or vinorelbine chemotherapy alone: • docetaxel + FEC, no trastuzumab (n=58): 3 cycles of docetaxel followed by 3 cycles of AC • docetaxel + FEC, trastuzumab (n=54): 3 cycles of docetaxel followed by 3 cycles of AC; trastuzumab started on day 1 of docetaxel and continued for 9 weeks • vinorelbine + FEC, no trastuzumab (n=58): 3 cycles of vinorelbine followed by 3 cycles of AC • vinorelbine + FEC, trastuzumab (n=62): 3 cycles of vinorelbine followed by 3 cycles of AC; trastuzumab started on day 1 of vinorelbine and continued for 9 weeks

Nearly all regimens have involved anthracycline chemotherapy, apart from BCIRG 006 which also included an arm that assessed concurrent trastuzumab with docetaxel and carboplatin (not anthracycline—Arm ‘TCH’, 1075 patients), and in HERA (which permitted a variety of chemotherapy regimens) where 6% of primary treatments contained no anthracyclines. In the context of the rest of this appendix, discussions about trastuzumab regimens are restricted to those containing anthracycline, which comprise most of the available evidence and are currently most widely used and applicable to New Zealand as standard care chemotherapy. Results from BCIRG 006 discussed in this appendix are therefore largely restricted to the anthracyclinecontaining arms AC-T and AC-TH (control and added trastuzumab). Study designs and their treatment regimens are summarised further in Figure 3 and at the end of this Appendix.

Figure 3. Schemas of published trials using adjuvant trastuzumab

Abbreviations: CEF, Cyclophosphamide, epirubicin, 5FU; LD, loading dose; PgR, progesterone receptor; qw, weekly; q3w, every 3 weeks

Source: Gonzalez-Angulo AM, Hortobagyi GN, Esteva FJ. Adjuvant therapy with trastuzumab for HER-2/neu-positive breast cancer. Oncologist 2006 Sep;11(8):857-67.

The results from these trials is summarised on the bpacnz website at http://www.bpac.org.nz/magazine/2007/april/herceptin.asp and Adjuvant! online (http://www.adjuvantonline.com/breasthelp0306/Trastuzumab.html). The study designs and the efficacy and cardiotoxicity results are summarised further in Tables 1 to 3 and Figures 4 to 5B below.

Table 1. Disease recurrence in sequential and concurrent trastuzumab regimens in RCTs of trastuzumab vs. standard care in the adjuvant treatment for HER2-positive early breast cancer

Trial Sequential (trastuzumab post taxane)

[n tmt/N tmt vs. n obs/N obs],

Concurrent (trastuzumab with taxane)

[n tmt/N tmt vs. n obs/N obs],

hazard ratio (95% CI), % absolute improvement1 (minimal from 95% CI1)

(HERA one-year treatment arm, 12 mth f/up–interim result) HERA one-year treatment arm, 23 mth median f/up NSABP B31 NCCTG N9831 joint analysis of N9831/B31 BCIRG 006 Arm AC-TH (36 mth year f/u) Overall post-anthracycline treatment FinHer (pre-anthracycline treatment) 36 mth f/u–interim non-anthracycline, concurrent (BCIRG 006 Arm TCH), 36 mth

[127/1694 vs. 220/1693],

0.54 (0.43-0.67), 5.5%

[218/1703 vs. 321/1698],

0.64 (0.54-0.76), 6.1% (3.9%)

[83/864 vs. 171/872], 0.45 (CI not reported; 2P=10-9), 10.0% (7.3%) [50/808 vs. 90/807], 0.55 (CI not reported; 2P=0.0004), 5.0% (2.5%) [134/1672 vs. 261/1679], 0.48 (0.39-0.59), 7.5% (5.8%) [128/1074 vs. 192/1073], 0.61 (0.48-0.76), 6.0% (3.2%) [262/2291 vs. 453/2235], (2.7%)

[103/985 vs. 117/979],

0.87 (0.67-1.13), 1.5% (-1.5%)

[321/2688 vs. 438/2677],

0.70 (0.61-0.81), 4.4%

0.53 (0.46-0.62), 6.9%

[12/115 vs. 27/116],

(5.5%)

0.42 (0.21-0.83), 12.8% (3.7%)

[142/1075 vs. 192/1073],

0.67 (0.54-0.83), 4.7% (1.8%)

Notes: 1 Absolute improvements in DFS are the differences between rates of disease recurrence in the observation and treatment arms. Minimum absolute improvements derive from upper limits of 95% confidence intervals for calculated DFS relative risks. 2 Overall results for sequential treatment comprise the N9831 sequential (B) and HERA one year trastuzumab arms compared with respective control groups (HERA 2-year median follow-up results). 3 Overall results for concurrent treatment comprise N9831 concurrent (C), NSABP B31 trastuzumab and BCIRG 006 ACTH arms compared with respective control groups. These results are post-anthracyclines, and hence do not include concurrent treatment given in FinHer (pre-anthracyclines), nor the non-anthracycline (carboplatin + docetxel + trastuzumab) Arm TCH of BCIRG 006.

Table 2. Trials reporting outcomes for adjuvant trastuzumab against vs. standard treatment in HER2-positive early breast cancer

Sequential treatment trials, long duration (12 month)i regimens HERA Trial N9831 (Arm B)

Patient Numbers Observation: 1,693 Trastuzumab (1 yr): 1,694 Trastuzumab (2 yr):i 1,694 1 loading dose (8mg/kg) trastuzumab, then 6mg/kg every 3 weeks for one year or two years (17 or 35 infusions, respectively). Observation: 979 Trastuzumab: 985

Concurrent treatment trials, long duration (12 month) regimens ii N9831 (Arm C) BCIRG 006 B31

Observation: 979 Trastuzumab: 840 Observation: 1,073 Trastuzumab: 1,074 Observation: 1,024 Trastuzumab: 1,019

Concurrent treatment trials with short duration regimens iii FinHer Trial E2198

Observation: (116) vinorelbine – 58 docetaxel – 58 Trastuzumab: (116) Docetaxel – 54 Vinorelbine – 62 9 Trastuzumab infusions at 1 week intervals. First dose 4mg/kg (90min infusion), remaining doses 2mg/kg (30 min infusion)

Short duration : 115 patients Long duration: 112 patients

Intervention

1 loading dose (4mg/kg) trastuzumab, then 2mg/kg every week for 52 weeks

Short duration: 10 Trastuzumab infusions at 1 week intervals. Trastuzumab given in combination with paclitaxel– Loading dose 4mg/kg followed by 9 weeks 2mg/kg Long duration: As above but with further 52 weeks of Trastuzumab at 2mg/kg per week.

Timing of treatment

Sequential (after completion of all chemotherapy – anthracycline chemotherapyiv and taxane treatmentv)

Concurrent with taxane (paclitaxel), after completion of anthracycline chemotherapy

Concurrent with taxane (docetaxel) treatment, before anthracycline chemotherapy 36-mth median f/up: 0.42 (0.21-0.83)

Short duration: Concurrent with taxane treatment, before AC anthracycline chemotherapy2 Long duration: concurrent and sequential DFS at 5 yrs: 76% short duration 75% long duration Note that E2198 is a pilot study, not designed to report efficacyiii

Disease free survival Hazard Ratio (95% confidence interval)

12-mth median f/up: 0.54 (0.43-0.67) 23-mth median f/up: 0.64 (0.54-0.76)

0.87 (0.67-1.13)

0.55 (CI not reportedvi)

23-mth median f/up: 0.49 (0.37-0.79)

0.45 (CI not reportedvi)

Overall DFS HR Overall survival (95% CI)

0.70 (0.61-0.81) (2-yr HERA f/u) 12-mth median f/up: 0.76 (0.47-1.23) 0.85 (0.55-1.33) Not reported

36-mth median f/up: 0.61 (0.48-0.76) 0.53 (0.46-0.62) (3-yr BCIRG 006 f/u) 36-mth median f/up: 0.59 (0.42-0.85) Not reported (joint analysis with N9831 arm C = 0.67 (0.48-0.93))

0.42 (0.21.0.83) 36-mth median f/up: 0.41 (0.16-1.08)

Overall OS HR (95% CI)

23-mth median f/up: 0.66 (0.47-0.91) 0.72 (0.55-0.94)

(joint analysis with B31 = 0.67 (0.480.93))

N/A – placebo group data not reported OS at 5 yrs: 89% short duration 83% long duration Note that E2198 is a pilot study, not designed to report efficacyiii N/A – placebo group data not reported

0.63 (0.50-0.80)

0.41 (0.16-1.08)

No evidence is available on the outcomes of the 2 year trastuzumab treatment arm in the HERA trial. Note that there was also an arm to BCIRG 006 (‘arm TCH’) that consisted of 6 cycles of docetaxel and carboplatin with concurrent trastuzumab (i.e. no anthracycline chemotherapy). However, because this regimen is not comparable to the other regimens, these results are not presented in this table. For further information regarding BCIRG 006 see Appendix One: Minutes of the relevant clinical advisory committee meetings. Anthracycline-containing regimens comprise most of the available evidence and are currently most widely used and applicable to New Zealand as standard care chemotherapy. iii The E2198 study (Sledge et al, poster presentation at SABCS 2006) was not designed to test efficacy, and was not powered to determine equivalence, and results comparing the treatment arms to the control arm have yet to be reported. However, the results supported the efficacy of short duration concurrent trastuzumab therapy when administered before anthracycline containing chemotherapy, as demonstrated in the FinHer study (Appendix One: Minutes of the relevant clinical advisory committee meetings). iv Anthracycline containing chemotherapy regimens (FEC or FAC). v The HERA trial allowed several different chemotherapy regimens. vi Note that N9831 Arm C and trial B31 data were only published as a joint analysis (Romond, 2005) without stating the hazard ratios’ confidence intervals for the individual trials. Confidence limits for the disease recurrence HR for N9831 Arm B were stated in the 2005 conference presentation on the ASCO website.

Table 3. Event rates, absolute improvements and numbers needed to treat (NNTs) in adjuvant trastuzumab trials in HER2-positive early breast cancer, for disease free survival (DFS) and overall survival (OS)

Disease free survival median totals f/u n N pts evnts treatment group n N pts rate evnts observation group n N pts evnts rate hazard ratio HR

-95% +95 CL % CL

relative risk RR

ARR/improvement in DFS

(years)

-95% +95% from CL HR CL

disease recurrence free number needed survival (DFS) to treat (NNT) from RR -95% +95% obs tmt (from -95% CL, from RR -95% CL, from CL, from RR) from RR CL, from RR RR RR

sequential post-anthracyclines HERA 23-mth FU data 1.9 N9831 sequential arm 1.5 overall HR (95% CI) 1.8 concurrent post-anthracyclines B31 & N9831 arm C 2.0 BCIRG 006 arm AC-TH 3.0 overall HR (95% CI) 2.4 concurrent pre-anthracyclines FinHer 3.0 overall HR (95% CI) 3.0 concurrent non anthracycline BCIRG 006 arm TCH 3.0 overall HR (95% CI) 3.0 all regimens overall HR (95% CI)

539 220 759

3,401 1,964 5,365

218 103 321

1703 985 2688

13% 10% 12%

321 117 438

1698 19% 0.64 979 12% 0.87 2677 16% 0.70

0.54 0.76 0.67 1.13 0.61 0.81

0.68 0.87 0.73

0.58 0.68 0.64

0.79 1.12 0.83

6.8% 1.6% 4.9%

6.1% 1.5% 4.4%

3.9% -1.5% 2.7%

8.0% 81.1% 3.8% 88.0% 5.9% 83.6%

87.2% 89.5% 88.1%

85.0% 86.6% 86.3%

16 67 23

26 -68 37

395 320 715

3,351 2,147 5,498

134 128 262

1672 1074 2746

8% 12% 10%

261 192 453

1679 16% 0.48 1073 18% 0.61 2752 16% 0.53

0.39 0.59 0.48 0.76 0.46 0.62

0.52 0.67 0.58

0.42 0.54 0.50

0.63 0.82 0.67

8.1% 7.0% 7.7%

7.5% 6.0% 6.9%

5.8% 3.2% 5.5%

9.0% 84.5% 8.2% 82.1% 8.2% 83.5%

92.0% 88.1% 90.5%

90.2% 85.3% 89.0%

13 17 14

17 31 18

39 39

231 231

12 12

115 115

10% 10%

27 27

116 23% 0.42 116 23% 0.42

0.21 0.83 0.21 0.83

0.45 0.45

0.24 0.24

0.84 13.5% 0.84 13.5%

12.8% 12.8%

3.7% 3.7%

17.7% 76.7% 17.7% 76.7%

89.6% 89.6%

80.4% 80.4%

8 8

27 27

334 334

2,148 2,148

142 142

1075 1075

13% 13%

192 192

1073 18% 0.67 1073 18% 0.67

0.54 0.83 0.54 0.83

0.74 0.74

0.60 0.60

0.90 0.90

5.9% 5.9%

4.7% 4.7%

1.8% 1.8%

7.1% 82.1% 7.1% 82.1%

86.8% 86.8%

83.9% 83.9%

21 21

57 57

2.2 1,847 13,242

737

6624

11% 1110

6618 17% 0.62

0.57 0.68

0.66

0.61

0.72

6.3%

5.6%

4.6%

6.6% 83.2%

88.9%

87.9%

Overall survival median totals f/u (years) n N pts evnts sequential post-anthracyclines HERA 23-mth FU data 1.9 N9831 sequential arm 1.5 overall HR (95% CI) 1.9 concurrent post-anthracyclines B31 & N9831 arm C 2.0 BCIRG 006 arm AC-TH 3.0 overall HR (95% CI) 2.5 concurrent pre-anthracyclines FinHer 3.0 overall HR (95% CI) 3.0 all concurrent non anthracycline BCIRG 006 arm TCH 3.0 overall HR (95% CI) 3.0 all regimens overall HR (95% CI) treatment group n N pts rate evnts observation group n N pts evnts rate hazard ratio HR

-95% +95 CL % CL

relative risk RR

ARR/improvement in OS

overall survival (OS)

-95% +95% from CL CL HR

from RR -95% +95% obs CL, from CL, from RR RR 1.8% 4.3% 2.7% 0.5% 3.7% 2.0% 2.8% 94.7% 4.4% 95.6% 3.3% 95.0%

number needed to treat (NNT) tmt (from -95% CL, from RR -95% RR) from RR CL, from RR 96.5% 99.9% 97.8% 95.2% 99.3% 97.0% 54 23 37 191 27 51

149 44 193

3,401 1,964 5,365

59 1 60

1703 985 2688

3% 0% 2%

90 43 133

1698 979 2677

5% 0.66 4% 0.85 5% 0.72

0.47 0.91 0.55 1.33 0.55 0.94

0.65 0.02 0.45

0.47 0.00 0.33

0.90 0.17 0.61

1.8% 0.7% 1.4%

154 129 283

3,351 2,147 5,498

62 49 111

1672 1074 2746

4% 5% 4%

92 80 172

1679 1073 2752

5% 0.67 7% 0.59 6% 0.63

0.48 0.93 0.42 0.85 0.50 0.80

0.68 0.61 0.65

0.49 0.43 0.51

0.93 0.86 0.82

1.8% 3.1% 2.3%

1.8% 2.9% 2.2%

0.4% 1.0% 1.1%

2.8% 94.5% 4.2% 92.5% 3.0% 93.8%

96.3% 95.4% 96.0%

94.9% 93.6% 94.9%

56 35 45

251 99 87

20 20

231 231

6 6

115 115

5% 5%

14 14

116 12% 0.41 116 12% 0.41

0.16 1.08 0.16 1.08

0.43 0.43

0.17 0.17

1.09 1.09

7.1% 7.1%

6.9% 6.9%

-1.0% -1.0%

10.0% 87.9% 10.0% 87.9%

94.8% 94.8%

86.9% 86.9%

15 15

-97 -97

136 136

2,148 2,148

56 56

1075 1075

5% 5%

80 80

1073 1073

7% 0.66 7% 0.66

0.47 0.93 0.47 0.93

0.70 0.70

0.50 0.50

0.97 0.97

2.5% 2.5%

2.2% 2.2%

0.2% 0.2%

3.7% 92.5% 3.7% 92.5%

94.8% 94.8%

92.7% 92.7%

45 45

491 491

2.6

632 13,242

233

6624

399

6618

6% 0.66

0.56 0.77

0.58

0.50

0.68

2.0%

2.5%

1.9%

3.0% 94.0%

96.5%

95.9%

Figure 4. Hazard ratios for disease recurrence in sequential and concurrent trastuzumab regimens in RCTs of trastuzumab vs. standard care in the adjuvant treatment for HER2-positive early breast cancer

Trastuzum ab treatm ent effects in early breast cancer - disease recurrence hazard ratio (95% CI) HERA (2-year median f/u) N9831 sequential arm 0 1 0 0.2 0.4 0.6 0.8 1 1.2

2 all sequential post-anthracyclines 3 B31 & N9831 concurrent arm 4 5

BCIRG 006 arm AC-TH (3-yr m f/u) 6 7 all concurrent post-anthracyclines ` 8 9 FinHer 10 11 all concurrent pre-anthracyclines 12 non-anthracycline, concurrent 13 (BCIRG006 arm TCH) 14 overall 15 16 17 18 all sequential all concurrent 12 month tmt 9-10 w eek tmt 19 20 21 22 23 24 25 26 favours treatment strong effect w eak effect favours controls

Figure 4A. Hazard ratios for overall mortality in sequential and concurrent trastuzumab regimens in RCTs of trastuzumab vs. standard care in the adjuvant treatment for HER2-positive early breast cancer

Trastuzum ab treatm ent effects in early breast cancer - overall m ortality HERA (2-year median f/u) N9831 sequential arm all sequential post-anthracyclines 0 0.2 hazard ratio (95% CI) 0.4 0.6 0.8 1 1.2 1.4

0 1 B31 & N9831 concurrent arm 2 BCIRG 006 arm AC-TH (3-yr m f/u)3 4 all concurrent post-anthracyclines5 6 FinHer 7 all concurrent pre-anthracyclines 8 9 non-anthracycline, concurrent 10 11 (BCIRG006 arm TCH) overall 12 13 14 15 favours treatment 16 strong effect

w eak effect

favours controls

Trial outcomes as reported Tables 1 to 3 and figures 4 to 5B (below) indicate that overall trastuzumab, when given in additional to standard chemotherpay, reduces the risk of both disease recurrence and overall deaths by ~1/3rd relative to standard chemotherapy along, with overall (pooled) hazard ratios (HRs) for the five RCTs of 0.62 (0.57-0.68) for disease recurrence and 0.66 (0.56-0.77) for overall mortality. This translates to overall absolute improvements in DFS of 5.6% (on top of 83% untreated free of disease recurrence/death over 2.2 years average median follow-up (f/u), number needed to treat (NNT) 18); and a 2.5% absolute improvement in overall survival (on top of 94% untreated survival by 2.2 years, NNT 40). All RCTs show statistically significant DFS gains except the sequential arm (Arm B) of N9831, and significant overall survival gains are demonstrated in the N9831 Arm C/B31, BCIRG 006 and HERA trials. The tables and figures also show the breadth of disparities in efficacy between the studies, i.e. differences in HRs between studies and regimens and differences in the ranges and breadths of confidence limits for those HRs: • For trials, hazard ratios for DFS vary between 0.42 (FinHer) to 0.87 (sequential arm of trial N9831), and confidence limits (precision) vary threefold in range (HERA 95% CI 0.54-0.76 vs. FinHer 0.21-0.83). These translate to ranges in DFS absolute improvements of 1.5% to 12.8% (NNT 8 by median 3 years for FinHer, 67 by median 1.5 years for N9831 sequential arm). These also translate to a range of minimum absolute improvements (from the upper 95% confidence limits of calculated relative risks (RR) applied to untreated disease recurrence rates) of -1.5% for the N9831 sequential arm (NNT -68 i.e. would harm every 68th patient by a median of 1.5 years) to 5.8% (B31 and N9831 concurrent arm, NNT 17 by median 2 years). For broad treatment regimens, overall (pooled) HRs for disease recurrence vary between 0.42 (0.21-0.83) for concurrent treatment pre-anthracycline, 0.53 (0.46-0.62) for concurrent treatment post-anthracycline and 0.70 (0.61-0.81) for sequential treatment post-anthracycline.

•

These values translate to absolute DFS improvements of 12.8%, 6.9% and 4.4% for concurrent pre-anthracycline (AC), concurrent post-AC and sequential post-AC respectively (NNTs 8 by median 3.0 years, 14 by 2.4 years and 23 by 1.8 years respectively). Minimum confident absolute DFS improvements for the broad regimens (from the RR upper 95% confidence limits) vary between 3.7% for concurrent treatment pre-anthracycline (NNT 27), 5.5% (18) for concurrent treatment post-anthracycline and 2.7% (37) for sequential treatment post-anthracycline. Figure 5. DFS in adjuvant trastuzumab trials in HER2-positive early breast cancer for untreated and treated groups

DFS without trastuzumab

Disease free survival in RCTs of trastuzumab for HER2 +ve early breast cancer

disease free survival (DFS) 0% 10% 20% 30% 40% 50% 60% absolute improvement in DFS from trastuzumab (including lower 95% confidence limit) 70% 80% 90% 100%

HERA 12 month arm (median followup 1.9 yrs) 12-month sequential post-anthracycline N9831 sequential arm (1.5 yrs)

B31 & N9831 concurrent arm (2.0 yrs) 12-month concurrent post-anthracycline

BCIRG 006 concurrent arm (3.0 yrs)

FinHer (3.0 yrs) 9-week concurrent pre-anthracycline

Figure 5A. DFS in adjuvant trastuzumab regimens for HER2-positive early breast cancer for untreated and treated groups

Disease free survival in trastuzumab regimens for HER2 +ve early breast cancer

disease free survival (DFS) 0% 10% 20% 30% 40% 50% 60%

DFS without trastuzumab absolute improvement in DFS from trastuzumab (including lower 95% confidence limit) 70% 80% 90% 100%

sequential post-anthracyclines 12mths (1.9 yrs*) 12-month sequential postanthracycline

concurrent post-anthracyclines 12mths (2.4 years*) 12-month concurrent post-anthracycline

concurrent pre-anthracyclines 9wks (3.0 years*) 9-week concurrent pre-anthracycline

*ITT patient population-weighted median years of follow-up

Figure 5B. Absolute improvements in DFS in adjuvant trastuzumab trials in HER2-positive early breast cancer, including minimum confident absolute improvements (from upper 95% confidence limits of hazard ratios)

Improvements in disease free survival in RCTs of trastuzumab for HER2 +ve early breast cancer

improvement in DFS -8% -6% -4% -2% 0% 2% 4% 6% 8% 10% 12% 14%

HERA 12 month arm (median follow-up 1.9 yrs) 12-month sequential post-anthracycline N9831 sequential arm (1.5 yrs) B31 & N9831 concurrent arm (2.0 yrs) 12-month concurrent post-anthracycline BCIRG 006 concurrent arm (3.0 yrs) FinHer (3.0 yrs) BCIRG 006 non-anthracycline arm (3.0 yrs)

9-week concurrent pre-anthracycline 12-month concurrent non-anthracycline likeliest improvement minimum improvement - lower 95% CL

sequential post-anthracyclines 12mths (1.9 yrs*) concurrent post-anthracyclines 12mths (2.4 years*) concurrent pre-anthracyclines 9wks (3.0 years*)

all regimens (2.2 yrs*)

*ITT patient population-weighted median years of follow-up

2. Increasing uncertainty around sequence and durations With data from the five RCTs, questions remain over the extent that sequential treatment prevents recurrence in the short and long term and the duration of this benefit. Firstly, the trials have a number of methodological issues that may affect their validity: • results from all five trials reported to date have been preliminary (interim)—all continue to follow-up patients, and none have met their preset target event accruals (e.g. HERA requires 951 primary endpoint events for final analysis4); it is difficult to assess the quality of two RCTs (B31 and N9831) because reporting has been limited to either a published joint analysis for B31 and the N9831 concurrent arm (Romond 20055)—with little disaggregation into the separate studies6 or description of key validity aspects of the separate studies—or a conference slideshow presentation for the N9831 sequential arm;7 results for BCIRG 006 too have been limited to conference slideshow presentations8, limiting assessiblity; all trials are open-label in trial design (unblinded); allocation concealment methodology is not adequately reported (except FinHer)—where inadequate or unclear allocation concealment has been associated with 30-40% larger estimates of treatment effects12; and reporting of compliance, contamination and co-intervention has been variable.

•

• • •

•

Further details on the quality of the trials are available in GATE13 appraisals undertaken by EPIQ at http://www.health.auckland.ac.nz/population-health/epidemiologybiostats/epiq/critical_appraisal_library/Herceptin.14 15 Secondly, the optimal schedule and duration of treatment for trastuzumab in the adjuvant treatment of early stage HER2-positive breast cancer cannot be determined from the current 16 evidence. This is not unusual for a new treatment with emerging evidence. Particular questions are raised about the durability of benefit, and the scheduling in relation to chemotherapy. Combinations of trastuzumab with taxane chemotherapy (i.e. concurrent treatment) seem to have a synergistic effect in both adjuvant treatment and treatment of metastatic disease.17 However, evidence is available that raises questions about the place of 12-month sequential therapy as the standard-of-care recommended by the supplier in its datasheet18. In New Zealand, the Medsafe-approved indication is currently restricted to the HERA regimen2 (12 months 19 sequential), unlike Australia , where sequential and concurrent (12 months or 9 weeks) is 20 approved and the US where 12 month concurrent treatment is approved. Non-publication of data for 12 month sequential treatment from N9831 PHARMAC (and other international bodies such as NICE in the UK) have been asked by the supplier to make decisions on funding trastuzumab on the basis of 12-month median follow-up data from the HERA study2 (i.e. the sequential treatment regimen). This was supported with longer-term follow-up data from BCIRG 006 and combined results of the NSABP B31 and NCCTG N9831 studies (all using concurrent treatment).

However, trial N9831, which had three arms, is the only study to have investigated both sequential and concurrent treatment with trastuzumab:

Arm A (control): 4 cycles of AC treatment followed by 12 weeks paclitaxel Arm B (post-anthracycline and sequential with trastuzumab treatment): 4 cycles of AC treatment followed by 12 weeks paclitaxel and then 52 weeks trastuzumab Arm C (post-anthracycline and concurrent with trastuzumab treatment): 4 cycles AC followed by 12 weeks paclitaxel with trastuzumab, and a further 52 (I think) weeks trastuzumab.

Although not initially provided to PHARMAC and others, separate data from each of the arms of the N9831 from an unplanned interim analysis were presented at the American Society of Clinical Oncology (ASCO) conference in 2005 (at http://www.asco.org/ac/1,1003,_12-002511-00_180034-00_19-005815-00_21-001,00.asp7). The separation of the data for sequential treatment from the concurrent treatment, albeit not yet formally published, is important as it raises questions about optimal treatment scheduling in relation to chemotherapy. At 1.5 years’ median follow-up, N9831 Arm C, one year of trastuzumab administered concurrently with taxane chemotherapy, demonstrated a large and significant benefit in diseasefree survival (DFS) over usual care (control Arm A) (HR 0.48, 95% CI 0.39-0.59). However, N9831 Arm B (one year sequential trastuzumab therapy) did not show a statistically significant improvement in DFS compared with usual care (control Arm A) (HR 0.87, 95% CI 0.67-1.13) (see Figure 6A). Figure 6A. Disease free survival in trial N9831, sequential trastuzumab vs. control arms

Perez EA. Further Analysis of NCCTG-N9831. Slide presentation at ASCO annual meeting 2005, available online at http://www.asco.org/ac/1,1003,_12-002511-00_18-0034-00_19-005815-00_21-001,00.asp, accessed January 2007.

Importantly, Arm C (12 months’ concurrent trastuzumab) showed a statistically significant improvement in DFS over Arm B (12 months’ sequential trastuzumab) standard chemotherapy (HR 0.64, 95% CI 0.46- 0.91)) (see Figure 6B). Figure 6B. Disease free survival in trial N9831, concurrent vs. sequential trastuzumab arms

The sequential regimen in Arm B of N9831 is essentially the same 12-month regimen as examined in the HERA study (i.e. the regimen indicated in New Zealand (Medsafe-approved NZ datasheet21)). The conflicting results from these two studies raise questions about the efficacy of 12 months sequential trastuzumab. The evidence available for N9831 Arm B is confined to a conference presentation of the interim analysis (median f/up 1.5years) and has not been formally published. The results for the sequential arm were not statistically significant, and may have therefore been subjected to publication bias, although there may be other reasons for non-publication.vii At its May 2006 meeting, the Pharmacology and Therapeutics Advisory Committee (PTAC, PHARMAC’s independent clinical advisory body http://www.pharmac.govt.nz/ptac.asp) considered the above unpublished disease-free survival data for N9831, in the form of MS PowerPoint slides of the conference presentation. The PTAC minute states that members noted that sequential trastuzumab treatment (Arm B) was not statistically superior to non-trastuzumab treatment (Arm A), but that concurrent trastuzumab treatment (Arm C) resulted in a significant improvement in disease-free survival compared with Arm B. Members considered that although these data were preliminary, they raised concerns about the optimal dosing schedule of trastuzumab treatment. PTAC requested that the supplier provide full data from the N9831 trial. Non-publication of what are said to be interim data reduces confidence in the results; however, earlier 1.5-year data for Arm C (also interim) were used in the high-profile published Romond paper5 (NEJM 2005) when they were pooled with trial B31 data, and in fact were published individually in the on-line appendix to the Romond paper (6 Figure 1). The de facto publication of the Arm C results lends legitimacy to the Arm B results, particularly when the key reason stated in that paper for excluding Arm B was simply because trastuzumab was not given concurrently with paclitaxel5 (and hence was not comparable with trial B31 in joint analysis), when in turn the

vii

Publication bias emerges when published trials do not represent all trials undertaken, usually because statistically significant results tend to be submitted and published more frequently than indeterminate results.

differences between concurrent treatments in N9831 (Arm C) and B31 were appreciable. [The N9831 Arm C and B31 trials in the Romond interim analysis differed in patient eligibility (high risk negative node status), methods of randomisation allocation, taxane regimens, anthracyclines, sequencing with radiotherapy, sequencing with hormonal therapy, and aromatase inhibitor types and when they started to be used in the trials.22] Another key reason stated for treating the Arm C results differently,that it was planned to be jointly analysed with B31 as approved by the FDA, is unclear, given that the joint analysis may not have been pre-planned.23 In general terms, it has been observed that failing to publish means, at best, that ineffective treatments are widely used in patients and, at worst, can lead to unnecessary illness and even death if the reported risks of harms are underestimated.24 25 26 The authors of the joint N9831C/B31 trials considered that further follow- up of groups B and C in trial N9831 would be necessary for an adequate evaluation of the efficacy of concurrent as compared with sequential administration of trastuzumab (Romond et al 2005). Professor Ian Smith and colleagues in the publication of the HERA 23-month median follow-up data made similar observations, stating: “The first [major question regarding adjuvant trastuzumab] is whether trastuzumab started concurrently with taxane chemotherapy (as in the USA trials) is better than trastuzumab starting sequentially after completion of chemotherapy (as here). The NCCTG N9831 trial addresses this issue—this trial includes a third group given sequential trastuzumab. Preliminary data suggest that sequential treatment might be less effective than concurrent treatment [Perez slideshow ASCO 2005] but this was an unplanned comparison with low statistical power, and longer follow-up is needed for confirmation.” (Smith et al. Lancet 2007).

Initial HERA results overstate effects on disease free survival Because the HERA study of sequential treatment came to a different result than the sequential arm of N9831, the conflicting information from these studies raises questions about the extent of effect and duration of response to sequential trastuzumab. This becomes more relevant in light of the 23 month median follow-up data from HERA; these reported improvements in overall survival to now be statistically significant27, but at the same time showed reduced benefit in disease-free survival over time. Specifically, in HERA by 23 months median time the risk reduction in DFS for patients who had been treated with trastuzumab had reduced from what had been reported in the 12 month median f/u HERA publication; the HERA disease-free survival 12 month median f/u hazard ratio was 0.54 (95% CI 0.43-0.67) (Piccart-Gebhart et al 20052); 23 month median f/u HR 0.64 (0.54-0.76) (Smith et al 20074). These translated to an 8.3% absolute difference in DFS at the 12 month median follow-up, reducing to 6.1% at 23 months, this reduction being statistically significant28 (detailed later). Note that this apparent decrease in the effect on DFS over time in the HERA trial does not appear to be due to the appreciable contamination that occurred in the study (where patients in the comparator arm were offered trastuzumab treatment once the 12-month DFS results emerged, causing a substantial loss of patients in the control arm in a non-randomised fashion). The opportunity to cross-over occurred relatively late, and the hazard ratios for both intention-to-treat analysis (ITTA) of all patients and a censored analysis (where cross-over patient data were removed) were virtually identical for DFS (0.64 ITTA, 0.63 censored analysis).4 This

concordance of hazard ratios suggests a genuine effect (rather than artefact due to comparator group patients latterly receiving trastuzumab treatment). Importantly, this cross-over has limited the ability of the HERA study to examine longer-term outcomes for 12 months’ sequential trastuzumab compared to standard chemotherapy, because the standard chemotherapy arm has largely been lost, possibly to a point where longer-term comparison questions can never be resolved. The impact of timing of analysis in HERA, with the apparent waning of treatment effect over time, can be seen in the following graph (Figure 7) of disease free survival over time (Figure 7; note the use of abbreviated scale for DFS). Figure 7. Comparison 12-month and 23-month median follow-up DFS in HERA

Disease-free survival in HERA: 12-month interim vs. 23-month follow-up data

100%

95%

disease-free survival (abbreviated scale)

90%

HERA 12-month median follow-up HERA interim 12-month analysis, tmt

85%

80%

HERA interim 12-month analysis, cntrl

75%

HERA 23-month median follow-up, tmt

70% difference between 12-month and 23month data, affecting hazard ratios

HERA 23-month median follow-up, cntrl

65%

60%

55%

50% 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

years since randomisation

Impact of N9831 Arm B and the updated HERA results on disease free survival benefits of 12 month sequential treatment The non-inclusion of the Arm B N9831 data and use of the interim HERA results (limited to 12 months median follow-up) means that initial estimates of the effectiveness of trastuzumab in the sequential setting may have been overstated. Including the available interim N9831 Arm B data and then updating the HERA results for the 23-month median data means that the effects of sequential treatment on disease progression reduce by one third—the overall hazard ratio increases from 0.54 (95% CI 0.43-0.67) for the HERA 12-month median follow-up data to 0.70 (0.61-0.81) for the pooled HERA median 23-month follow-up and N9831 Arm B data. The potential impact on the overall effectiveness (HR) for 12 month sequential treatment of the missing N9831 Arm B vs. A comparison may be even larger, considering the potential added events accruing since the time of its initial analysis. The N9831 Arm B vs. A events were analysed at around April 2005, compared with May 20064 for the 23-month median follow-up for HERA trial. Had the N9831 B vs. A results been re-analysed and then presented at the same time

as were HERA 23-month median follow-up data at the 2006 ASCO conference, the additional numbers of events would influence the weight given to N9831 Arm B overall—similar to the impact of time on event numbers in the HERA follow-ups, and as suggested in the following graphs (Figures 8 and 9) that model possible disease events (numerical) and disease free survival (comparative) over time. Figure 8. Timecourse of disease recurrence events in HERA and N9831 Arm B vs. A

Timecourse of evidence for sequential trastuzumab in early breast cancer: disease recurrence in HERA and N9831-B trials*

350 *empirical and anticipated

HERA = 539 events by 2.0 years median f/u (May 2006)

300 ? effects of cross-over of N9831 Arm A and B patients to Arm C (Arm B patients were eligible to cross-over from 26/01/05, Arm A from 26/04/04)

HERA 1-year median follow-up

cumulative no. disease events

250

N9831 arm B vs. A events, which await publication. At May 2006 (2 yrs 7 mths) these should have accounted for ~41% of all HERA/N9831 events at that time

200

150

N9831 B vs. A = 340 events projected to any censoring by May 2006 (which would be 2.6 years median f/u)

100

HERA, tmt HERA, cntrl HERA interim 12-month analysis, tmt

HERA interim 12-month analysis, cntrl N9831 arm B, analysed to date, tmt

0 Apr/2001 Jun/2001 Aug/2001 Oct/2001 Dec/2001 Feb/2002 Apr/2002 Jun/2002 Aug/2002 Oct/2002 Dec/2002 Feb/2003 Apr/2003 Jun/2003 Aug/2003 Oct/2003 Dec/2003 Jan/2004 Mar/2004 May/2004 Jul/2004 Sep/2004 Nov/2004 Jan/2005 Mar/2005 May/2005 Jul/2005 Sep/2005 Nov/2005 Jan/2006 Mar/2006 May/2006

N9831 arm B, analysed to date, cntrl N9831 arm B, projected to May 2006, tmt N9831 arm B, projected to May 2006, cntrl

date (month/year)

Figure 9. Timecourse of DFS in HERA and N9831 Arm B vs. A

Timecourse of evidence for sequential trastuzumab in early breast cancer: disease-free survival in HERA and N9831-B trials*

*empirical and anticipated 100% 95% 90% 85% disease-free survival 80% 75% 70% 65% 60% 55% 50% Apr/2001 Jun/2001 Aug/2001 Oct/2001 Dec/2001 Feb/2002 Apr/2002 Jun/2002 Aug/2002 Oct/2002 Dec/2002 Feb/2003 Apr/2003 Jun/2003 Aug/2003 Oct/2003 Dec/2003 Jan/2004 Mar/2004 May/2004 Jul/2004 Sep/2004 Nov/2004 Jan/2005 Mar/2005 May/2005 Jul/2005 Sep/2005 Nov/2005 Jan/2006 Mar/2006 May/2006

N9831 arm B, projected to May 2006, tmt N9831 arm B, projected to May 2006, cntrl HERA 23-month median follow-up, tmt HERA 23-month median follow-up, cntrl N9831 arm B, analysed to date, tmt N9831 arm B, analysed to date, cntrl

HERA 1-year median follow-up

date (month/year)

However, the results for N9831 Arm B were an unplanned interim analysis with low statistical power, and longer follow-up is needed for confirmation.5 PTAC (February 2007) considered that there was now likely to be longer-term follow-up of outcomes (disease free survival and mortality) in this study, and asked for all the updated data from all three arms of the trial be made available. Such data should be available, the authors of the joint B31/N9831 Arm C analysis (Romond et al.) commented that the group C vs. B comparison requires longer follow-up and that such further follow-up is necessary.5

The impact of sequencing on the effectiveness of 12 month treatment Trial N9831 was also designed to directly address the efficacy of trastuzumab initiated concurrently with paclitaxel versus sequentially after completion of paclitaxel (Arm C vs. Arm B). Despite an overall small number of events, the N9831 trial showed a statistically significant benefit of concurrent therapy over sequential therapy for disease progression (HR 0.64, 2P logrank 0.0114), although this finding is qualifiedviii (Figure 6B). Figure 6B, repeated. Disease free survival in trial N9831, concurrent vs. sequential trastuzumab arms

The median duration of trastuzumab treatment was likely to be longer in the concurrent arm (C) than the sequential arm (B). This was due to the different timing of trastuzumab sequencing, with the sequential arm commencing trastuzumab 12 weeks later. Hence patients in the concurrent arm were exposed to higher cumulative doses of trastuzumab at the time of the interim analysis—which may have contributed to both the concurrent arm’s apparent improved disease progression and higher cardiotoxicity.

viii

However the concurrent arm of N9831 (Arm C) had a higher incidence of cardiovascular events than the sequential arm (Arm B), with similar caveats to the efficacy dataviii, described later. The Arm C vs. Arm B comparison in N9831 represents an internally valid direct head-to-head comparison of concurrent vs. sequential treatment regimens, and these early data29 are the only available to directly compare the two regimens. These results showed statistically significant improvement in DFS with concurrent over sequential treatment. According to the Romond et al. joint analysis of trials B31 and N9831, after reviewing the results of the first joint interim efficacy analysis, the data monitoring committee overseeing trial N9831 requested an unplanned comparison of groups B and C and subsequently recommended disclosure of the results. Romond et al considered that though early, the comparison suggested delayed (i.e. sequential) administration of trastuzumab may be less effective than concurrent administration.5 Another large study of concurrent treatment, BCIRG 006, also demonstrated a similar benefit post-anthracycline in DFS over median 3 years of follow-up (HR 0.61 (0.48-0.76) for Arm ACTH). Further evidence, if less direct, suggesting improved efficacy of concurrent over sequential treatment can be seen in the results for both disease free survival and overall survival for all trials, and pooled results for regimens, shown earlier in tables 1 to 3 and figures 4 to 5B. These indicate overall (pooled) HRs for disease recurrence of 0.53 (0.46-0.62) for concurrent treatment postanthracycline and 0.42 (0.21-0.83) for concurrent treatment pre-anthracycline, compared with (less efficacious) 0.70 (0.61-0.81) for sequential treatment post-anthracycline, Similarly, HRs for overall mortality were 0.63 (0.50-0.80) and 0.41 (0.16-1.08) for concurrent treatment post- and pre-anthracycline respectively, compared with 0.72 (0.55-0.94) for sequential treatment postanthracycline.

The impact of sequencing on the cardiotoxicity of 12 month treatment All three studies of concurrent treatment (BCIRG 006, N9831 and B31) used a regimen of trastuzumab administered concurrently with a taxane in addition to extended (up to 12 months) treatment on completion of anthracycline-containing chemotherapy. Anthracycline treatment has a dose-related cardiotoxic effect on the heart, and as a result 15-20% of patients did not meet the cardiac criteria required for trastuzumab treatment initiation where trastuzumab was initiated after patients received cardiotoxic anthracycline treatment.30 The concurrent treatment regimens used in N9831 (Arm C) and NSABP B31 (i.e. anthracycline chemotherapy followed by 12 months trastuzumab started concurrently with a taxane) demonstrated a higher risk of cardiotoxicity than the sequential treatment arms (HERA and Arm B of N9831). The pooled analysis of N9831 and NSABP B315 reported that 14.2% of trastuzumab treated patients discontinued treatment before 52 weeks because of a confirmed asymptomatic decline in left ventricular ejection fraction (LVEF). Another 4.7% discontinued because of symptoms of congestive heart failure (CHF) or another adverse cardiac effect. In addition, 30.5% of patients required at least one trastuzumab dose delay in because of a decrease in LVEF or cardiac symptoms. The cumulative incidence for New York Heart Association (NYHA) class 3 or 4 CHF/death from cardiac causes at 3 years was 0.8% and 0% respectively in the control group, and 4.1% and 2.9% in the trastuzumab group. The above information suggests that, although more efficacious than 12 month sequential treatment (overall HR for DFS = 0.49 for concurrent compared with HR = 0.66 for sequential), the 12 month concurrent regimen (post anthracycline) is less safe for patients because of the associated risk of cardiotoxicity. However, it should be noted that patients in HERA were randomised into the study after completion of their chemotherapy; therefore patients with anthracycline-induced cardiac toxicity would not have been enrolled in the study and therefore were not captured in the denominators. Hence the HERA trial population can generally be considered to be at reduced risk of further cardiotoxicity than the B31/N9831 or BCIRG 006 trial populations, which may have contributed to the lower rates for cardiotoxicity seen in HERA.

Durability of response To reiterate, the durability of response of sequential therapy has not been demonstrated. Extended follow-up of the HERA study population is unlikely to quantify the sequence’s long-term benefits and risks, due to the partial loss of the control arm (with potential bias from non-randomisation) . As discussed earlier, the 2-year follow-up HERA results showed reduced benefit in disease-free survival, when the interim analysis (Piccart-Gebhart et al 2005) reported a disease-free survival hazard ratio over 12-months’ median follow-up of 0.54 (95% CI 0.43-0.67), but the 2-year median follow-up data (Smith et al 2007) showed a HR of 0.64 (0.54-0.76). The 23-month follow-up report for the HERA study included period-specific hazards by time, showing a change in disease recurrence over time, with convergence of hazards and overlapping of their confidence intervals beyond ~18 months – see Figure 10 below: Figure 10. Annualised DFS hazard rates, HERA one-year trastuzumab treatment vs. observation arms

Source: Smith et al 2007

Ideally the reduction in DFS benefits over time in the HERA trial should be confirmed by testing proportional hazards by time.ix 31 However, in the absence of any available such analysis of individual patient data, the reduction in DFS benefits over time reported in the HERA trial (the 0.54 HR becoming 0.64)2 4 is statistically significant on testing for interaction32 33 by the two time periods (p=0.02).28 The effects of the discord between the cumulative hazard ratio reported in the 23-month HERA f/u and period-specific effects from events occurring after the 12-month interim f/u can be seen in the following graph (Figure 11). This demonstrates a greater waning of effects with periodspecific relative risk. The similarities between the published 12-month median f/u hazard ratio (HR 0.54) and the calculated relative risk for the same time period (RR 0.58) give some comfort around the above tesing for statistical interaction. Figure 11. Cumulative hazard ratios and period-specific relative risks in HERA by time

Testing for proportional hazards over time will be to determine whether the effect of treatment is constant or varies significantly over time

Comparison of trends in hazard ratios and relative risks in HERA by publication (time period)

1.4

12 month median f/u, hazard ratio 23 month median f/u, hazard ratio (cumulative) 12 month median f/u, relative risk 23 month median f/u, relative risk (period-specific, beyond 12 months)

1.4

1.2

1.0 hazard ratio (HR)

1.0 relative risk (RR)

0.8

0.6

23-month median follow-up (Smith et al. Lancet 2007) 12-month median follow-up (Piccart-Gebhart et al. NEJM 2005)

0.6

0.4

0.2

0.0

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48

0.0 months median follow-up

For sequential regimens, the data therefore suggest convergence in period-specific rates perhaps by 3 years, i.e. trastuzumab no longer has any effects on disease events. For concurrent regimens, the data suggest a mixture of continuing divergence and some convergence by 4 years—see the following graphs in Figure 12 that compare death or relapse rates per year for trastuzumab versus standard treatment in the five trials. Figure 12. Estimated annual prevalence rates in trastuzumab trials, from visual abstraction of available disease free survival curves

Sequential 12 month regimens

HERA, Rx 12% HERA, cntrl

12% NCCGT N9831-B, Rx NCCGT N9831-B, cntrl

10%

0% Year1 Year2 Year3 Year4

HERA 12-month regimen Concurrent 12 month regimens

N9831 Arm B vs. Arm A

NCCGT N9831-C & NASBP B-31, Rx NCCGT N9831-C & NASBP B-31, cntrl 12% 10% 8% 6% 4% 2% 0% Year1 Year2 Year3 Year4

BCIRG-006, AC-TH arm BCIRG-006, cntrl 12% 10% 8% 6% 4% 2% 0% Year1 Year2

BCIRG-006, TCH arm

Year3

Year4

Joint analysis N9831 arm C vs. arm A, B31 (Romond) Concurrent 9 week regimens

FinHer, Rx FinHer, cntrl

BCIRG 006 AC-TH

12%

10%

0% Year1 Year2 Year3 Year4

FinHer

These results must however be treated with caution, given the small numbers of patients and events as time progresses (and hence instability with greater uncertainty in the later years). Using the above data, the following figure shows changes in hazard ratios over time for relevant trials, calculated from the published or reported DFS survival graphs. This suggests apparent waning of effects with the HERA trial (with statistical heterogeneity by time period28), also for trial N9831 Arm B (sequential)x and the BCIRG 006 trial (concurrent) with significant timerelated statistical interaction (p=0.01)34, but not for other concurrent regimens (B31/N9831 Arm C joint analysisx, FinHerx) (Figure 13).

Tests for statistical interaction by time period are available for HERA and BCIRG 006, as both these studies have available event numbers obtainable from interim analyses. No such tests for statistical interaction by time period are performable for trials N9831, B31 and FinHer, as these studies have reported their interim results only once. Ideally such testing should be performed using individual patient data and testing proportional hazards by time.

Figure 13. Efficacy of trastuzumab over time (period-specific hazard ratios for DFS)

Trastuzumab treatment effects over time

sequential post AC: HERA

1.6

sequential post AC: NCCGT N9831-B concurrent post AC: NCCGT N9831-C & NASBP B-31 favours controls

period-specific hazard ratio for disease events (hazards extrapolated from DFS graphs in published data or conference presentations)

1.4

concurrent post AC: BCIRG-006 AC-TH concurrent pre AC: FinHer

1.2

concurrent pre AC: BCIRG-006 TCH (non-anthracycline)

HERA

1.0

0.8

favours treatment weaker effect

0.6

0.4

0.2

FinHer

0.0 Year 1 Year 2 Year 3 Year 4 time from randomisation (years)

favours treatment stronger effect

3. Concurrent treatment pre-anthracyclines (FinHer and E2198) FinHer FinHer was an open-label RCT, comparing docetaxel with vinorelbine for the adjuvant treatment of early stage breast cancer (n=1010), where women with HER2-positive cancers were also randomised to receive nine weeks concomitant treatment with trastuzumab or no such treatment (n=232). Examining trastuzumab in FinHer is in effect a nested RCT for HER2-positive patients [comparing trastuzumab (with docetaxel or vinorelbine) vs. no trastuzumab (but still with docetaxel or vinorelbine)], sited within the wider RCT [comparing docetaxel with vinorelbine (with randomisation to trastuzumab or no trastuzumab for those patients who were HER2+ve)]. Hence the trastuzumab component of the trial has randomised HER 2 positive patients to trastuzumab or no trastuzumab and to docetaxel or vinorelbine (see diagram).

no.patients in FinHer subgroups HER2+ve trastuzumab no trastuzumab HER2 -ve trastuzumab no trastuzumab total

* 1 excluded from analysis **1 excluded from analysis because had HER2+ve CAB

docetaxel 54 58 0 390 (389)** 502 (501)

vinorelbine 62 (61)* 58 0 388 508 (507)

total

106 (105) 116 0 778 (777)

1010 (1008)

Results to date, as reported in the NEJM in 2006, were from a pre-planned early interim analysis, pending final analysis at five years median follow-up (or 150 events overall, whichever occurring earlier). Data from final analysis may be available from mid-2007 (the early efficacy analysis (NEJM 2006) was for median follow-up three years and took place in May 2005) (Joensuu et al NEJM 20069).35 The shorter duration concurrent regimen for adjuvant trastuzumab in the FinHer study was based on two theories: 1. trastuzumab acts synergistically when given concurrently with chemotherapy (indicated by data from the metastatic disease setting; note that this synergism hypothesis has subsequently been supported by the trial N9831 comparison of sequential vs. concurrent7); and 2. cardiotoxicity may be better managed when trastuzumab treatment is given concurrently 36 with taxanes but prior to cardiotoxic anthracycline chemotherapy. The efficacy results of the FinHer trial to date suggest that administration of trastuzumab for nineweeks concurrently with a taxane is effective in the treatment of HER2-positive early stage breast cancer. Trastuzumab was administered weekly concurrently either with three-weekly docetaxel or weekly vinorelbine, followed by three cycles of FEC (5-fluorouracil/ epirubicin/ cyclophosphamide) in each arm. Docetaxel was associated with improved recurrence-free survival as compared to vinorelbine (hazard ratio 0.58, 95% CI 0.40 to 0.85), and the addition of trastuzumab to chemotherapy was associated with improved recurrence-free survival as compared to the same chemotherapy administered without trastuzumab (hazard ratio 0.42, 95% CI 0.21 to 0.83). During a median follow-up time of 3 years reported to date, these treatments have not been associated with detectable cardiac toxicity, although caution is required with this interpretation.xi

The low cardiotoxicity observed in FinHer could also be explained by the relatively low cumulative dose of anthracycline chemotherapy (180 mg/m2 epirubicin while the maximum tolerated cumulative dose of epirubicin is of

The PTAC considered that the number of patients treated in the FinHer study was not insignificant and that the data from that trial were valuable. Possible synergism of docetaxel with trastuzumab The main aim of FinHer was to determine which treatment was the most effective—vinorelbine or docetaxel—in early stage breast cancer (with and without HER2 amplification). In patients that did and did not have HER2 amplification (i.e. all patients), docetaxel had better efficacy in terms of disease progression—the HR for DFS favoured docetaxel over vinorelbine (HR 0.58 (0.40.85)). The total patient numbers initiating treatment in this study were 507 (vinorelbine) and 502 (docetaxel). Overall survival (OS) favoured docetaxel, but this was not statistically significant (HR 0.66 (0.38-1.17)). The HER2-positive trastuzumab treated patients group did not undergo a prospectively-defined subgroup analysis to examine its effects according to underlying docetaxel or vinorelbine chemotherapy. However, in response to concerns raised about mixing results for docetaxel with vinorelbine for the trastuzumab-treated HER2-positive patients, informal post-hoc analysis of these patients who were treated with docetaxel and trastuzumab reveals persisting statistically significant efficacy for this subgroup, despite a halving patient numbers. Such analysis indicates a relative risk for disease recurrence of 0.27 (0.08-0.90)xii—where docetaxel + FEC + trastuzumab vs. docetaxel +FEC (no trastuzumab) = 3/54 vs. 12/58 events/patients = 5.6% vs. 20.7%. This compares with a worse and statistically non-significant result for vinorelbine + FEC + trastuzumab patients in FinHer (RR for trastuzumab vs. no trastuzumab of 0.57 (0.27-1.20)).37 It is acknowledged that this is a small, retrospectively-defined post-hoc subgroup analysis with no evidence of statistical heterogeneity on formal testing,37 and as such the results should be treated with caution. However, although such results cannot statistically be validly extrapolated to estimate expected results in clinical practice, they do support the hypothesis, along with in vitro 38 evidence, that docetaxel and trastuzumab have a synergistic effect when they are used together as per the FinHer regimen and that the real-life risk reductions are likely to be more, if anything, than seen with sequential 12 month treatment.

Reliability and validity of the FinHer trial and nine-week concurrent treatment The data from FinHer show that that the combination of docetaxel plus 9-week concurrent trastuzumab is effective and well tolerated in the treatment of HER2-positive early breast cancer. However, concerns have been raised that the evidence for 9-week treatment is less reliable because of small patient numbers showing effects only on disease free survival in only one clinical trial (the FinHer study). In other words, the subgroup of the FinHer study that evaluated adjuvant trastuzumab was small, with 232 patients randomized to receive or not to receive trastuzumab, overall survival (OS) reported to date has not been statistically significant (HR 0.41), and the data have not been formally confirmed by other trials. These three issues are addressed as follows:

720mg/m2), and perhaps by less sensitive LVEF testing/thresholds. In the B31/N9831 studies doxorubicin was administered at a cumulative dose of 240 mg/m2 while its maximum tolerated cumulative dose is only 500 mg/m2. Epirubicin is generally presented as a less cardiotoxic agent than doxorubucin. Indeed, as stated in the FinHer paper, the small size and the short duration of the follow-up are limitations of the study and the optimal duration of adjuvant trastuzumab therapy is not known and may be clarified only in further randomized trials. (KCE2006)

xii

chi-square 0.02 (df=1), Fisher 2-tail exact probability 0.03

Patient numbers In terms of patient numbers in the FinHer trial (the wide 95% CI , 0.21-0.83, which reflected the degree of uncertainty from small numbers), the trial gave results that were statistically significant despite its smaller size. This reflects the strong efficacy of short duration concurrent treatment in this setting (more than halving disease recurrence), so that fewer patients were needed to confirm such a strong effect. Because this trastuzumab regimen was so effective, HER2-positive patients studied in FinHer (assigned to trastuzumab or no trastuzumab) could have been as few as 145 for the results to still be statistically significant (calculated using binomial methods39 from the central estimate relative risk).40 Concerns about the reliability of the FinHer data because of small patient numbers ignore the large effects (strong efficacy) seen in the study, and represent a statistical misunderstanding that study populations/sample sizes (denominators) drive variability of effect; in fact it is event numbers (numerators) that are the more important.41 42 xiii Large treatment effects—likely to be more clinically worthwhile—but with wider confidence intervals (greater imprecision) should not be ignored essentially because of less power (this is similar to where ‘absence of evidence is not evidence of absence’43). Conversely, more precise treatment effects from larger samples (with narrower confidence intervals) but with smaller treatment effects are likely to have less impact clinically (i.e. greater confidence but a lesser effect on outcomes); such evidence should not override less precise evidence as of right. Such concerns about the patient numbers in FinHer44 seem analogous to post-hoc power calculations, where in fact once results are available, a trial yields a treatment effect and confidence interval for the results; the power of the trial is expressed in that confidence interval, and hence the power is no longer a meaningful concern.41 The first modern RCT—streptomycin for tuberculosis, undertaken in 1948—had only 107 patients45, and is a classic example of a big effect not needing big patient numbers. Recently, a phase III RCT involving 208 patients (TAnDEM) was the basis of the approval in the European Union of trastuzumab for use in combination with aromatase inhibitors for treatment of postmenopausal patients with HER2 and hormone receptor-co-positive metastatic breast cancer46—being fewer patients than in the FinHer trastuzumab nested RCT. In addition, similar scrutiny to study size and effects can also apply to the HERA and other sequential data (i.e. all the information available for the sequential 12 month regimen, being the regimen strongly advocated for49). The hazard ratios for all 12 month sequential data combined (HERA and N9831 Arm B) and for FinHer are highly similar across all ranges, suggesting few differences despite FinHer’s smaller numbers. Combining HERA 2-year follow-up and N9831 Arm B data reduces effectiveness by 1/3rd to reach a hazard ratio for 12 months’ sequential

xiii

When examining variability of effects, the influence of numbers of events rather than numbers of patients can be seen by examining the standard formulae for variance of relative risk for discrete dichotomous (binary) outcomes (Rothman 1986):

95% CI for RR =

where

RR (relative risk) = (n1/N1)/(n2/N2), Z (standardised normal variate (with µ =0, σ = 1) = 1.96 for a 95% CI n1 = no. events in experimental group, n2 = no. events in observational group, N1 = no. subjects in experimental group, N2 = no. subjects in observational group n1/ N1 = incidence rate for experimental group, n2/ N2 = incidence rate for observational group

treatment of 0.70 (0.61-0.81); the upper confidence limit for this hazard ratio (i.e. the minimum extent that disease recurrence can confidently be expected to reduce) is similar to FinHer’s: • • Historically (and the basis for funding decisions internationally for 12 months’ sequential trastuzumab treatment), HERA’s interim hazard ratio (HR) was 0.54 (95% CI 0.43-0.67); However, the combined updated HERA/N9831-B HR increases the hazard ratio for sequential treatment to 0.70 (0.61-0.81), thus a relative hazard reduction (RHR) of 30% (95% CI 19%-39%); This compares with the FinHer HR of 0.42 (0.21-0.83), being a RHR of 58% (17%79%).

•

Hence, the most realistic estimates suggest the FinHer regimen should be at least as effective as sequential regimens (RHR 58% vs. 30%). More importantly, at the very worst—i.e. the minimum extent that disease recurrence can confidently be expected to reduce—and even accounting for its smaller number of patients, the FinHer results were nearly as effective as sequential regimens (17% vs. 19%)—see graphs below (figures 14 and 15). Figure 14.

Disease-free survival over time in trials of trastuzumab in early HER2+ve breast cancer, trastuzumab vs. no trastuzumab in 12 month sequential and 9 week concurrent regimens

% disease-free survival 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

no trastuzumab 0 years minimum confident improvement with trastuzumab minimum to likeliest confident improvement with trastuzumab likeliest to maximum confident improvement with trastuzumab 2 years - combines HERA data (23 months median f/u) and a 6 month extrapolation of N9831 arm B data (where the median f/u reported to date was 18 months) minimum extent that DFS can confidently be expected to increase (calculated using the upper 95% confidence limits for hazard ratios for disease recurrence)

median follow-up time in RCTs

1 year

2 years - HERA and N9831 arm B

3 years - FinHer

Figure 15

Trastuzum ab treatm ent effects in early-stage breast cancer: pre-anthracycline concurrent vs. sequential regim ens hazard ratio (95% CI) for disease events 0.0 0 1 (HERA 12-month follow -up - interim) 2 HERA 23-month follow -up N9831 arm B, 1.5 yr interim 3 4 5 0.2 0.4 0.6 0.8 1.0 1.2

6 HERA 23-month f/u & N9831 arm B 7 FinHer 8 9

stro 10 nger effect weaker effect favo urs co ntro ls

As noted previously, PTAC considered that the number of patients treated in the FinHer study was not insignificant and that the data from the FinHer trial were valuable. Overall survival As stated above, the FinHer results have not confirmed benefits in terms of statistically significant improvement in overall survival. At three years median follow up, the overall survival (OS) benefit in the trastuzumab treatment arm compared with those that did not receive trastuzumab, although strong (HR 0.41), was not statistically significant (95% CI 0.16-1.08). However, this non-significant overall survival result may result from the combination of the small sample size and short follow-up at the time of analysis; as noted above, the trial results as reported in the NEJM in 2006 were a pre-planned early interim analysis, pending final analysis at five years median follow-up (or 150 events overall, whichever occurring earlier).47 xiv Given the trend to a strong effect on OS (the 0.41 median HR), it is possible to speculate that this trend in FinHer may have become statistically significant within 3 years and 8 months median follow-up (calculated using binomial methods using constant relative risk of 0.43), i.e. by around January 2004. Although this may be an underestimate (given for instance the improved baseline survival seen in out years elsewhere), it still seems possible that, based on the trend in OS seen in the early efficacy FinHer results, a statistically significant improvement in OS may become evident in the final 5-year median follow-up analysis of FinHer expected later in 2007.48

xiv

Data from final analysis of FinHer may be available from mid-2007 (where the early efficacy analysis (NEJM 2005) was for when median follow-up exceed three years and was able to take place in May 2005) (Joensuu et al NEJM 2006).

sequential

In addition, the lack of overall survival benefit to date with FinHer is similar to the early interim analysis of the HERA study at 12 months median follow-up—the basis of New Zealand’s current provisional registration for early stage trastuzumab18—which also did not show statistically significant survival improvements.2 Studies of trastuzumab in the early breast cancer setting have now repeatedly demonstrated that statistically significant improvements in DFS do correlate with statistically significant improvements in overall survival.4 5 8 Therefore, it seems reasonable to consider that the statistically significant improvements in DFS seen with the 9 week FinHer regimen could represent a surrogate marker for likely improvements in overall survival. 12 months’ sequential treatment was widely advocated in 2005/06 well before the availability of the HERA 23-month follow-up data that demonstrated statistically significant effects in OS. Instead, the OS effect seen for 12 month concurrent treatment in the Romond joint analysis of trials B31 and N9831 Arm C vs. A (NEJM 2005) was applied to the 12 month sequential setting. 49 However, HERA and the two concurrent trials (Romond) used quite different dosing/sequencing, and only 26% in HERA received taxanes and 6% received no anthracyclines, meaning direct comparison of those results with the Romond trials is problematic.22 Similarly, the significant effects on overall survival from the joint analysis of trials B31 and N9831 Arm C (Romond) could be applied to the 9 week concurrent regimen. For ease of reference, the relationships between regimens can be seen in the following diagram:

duration of treatment 9 weeks FinHer 12 months N9831 C vs A, B31

sequence of treatment

concurrent with anthracycline

sequential to anthracycline

HERA, N9831 B vs A

= the direction that evidence of statistically significant overall survival in Romond 2005 was extrapolated or could be extrapolated to other settings

Given these observations and assumptions, it therefore seems reasonable to consider that the statistically significant improvements in DFS seen with the 9 week FinHer regimen could represent a surrogate marker for likely improvements in overall survival, whilst awaiting further longer-term results from FinHer. Data from other clinical trials (E2198) The results for short-term trastuzumab treatment in FinHer have not been replicated and published in a similar trial setting. However, although significant effects have been reported in only the one trial, data supporting the concept of short-term concurrent trastuzumab treatment is available from a second study. The results of ECOG E219810, a trial of short-duration trastuzumab therapy given concurrently with paclitaxel, were presented as a poster at San Antonio Breast Cancer Symposium (SABCS) 2006. This pilot study, which assessed cardiac safety as the primary outcome, compared short duration trastuzumab (10 weeks) given concurrently with paclitaxel prior to anthracycline treatment, with the same treatment plus an additional 52 weeks trastuzumab after completion of anthracycline treatment. The 5-year followup reported similar clinical outcomes from the short duration concurrent regimen as with extended (12 month) trastuzumab treatment.

In detail, E2198 was a small randomized adjuvant trial of adjuvant chemotherapy in early stage breast cancer that examined the cardiac effects of HP (trastuzumab [H] plus 3 weekly administered paclitaxel [P]) followed by AC (doxorubicin [A] plus cyclophosphamide [C]) among 234 breast cancer patients with HER2-positive stage II disease. The patients were randomly allocated to receive either HP175 q3w X 4 followed by AC q3w x 4 (Arm B; HPx3ACx4), or to the same regimen followed by H for 52 weeks (Arm C; HPx3ACx4H for 52 wks). The median follow-up time at reporting was 64 months. Disease-free survival at 5 years was similar for Arms B and C (76% vs. 73%, 32/115 vs. 19/112 respectively, P= 0.55), and there was no difference in overall survival between the study arms (5-year survival was 88% for Arm C (22/115) and 83% for Arm B (12/112), P=0.29). Seven study participants were diagnosed with congestive heart failure (Arm B, n=3; Arm C, n=4), all within 3 years from randomisation. Note that the poster presentation did not describe outcomes for the standard treatment arm (control Arm A). Although the study did not set out to determine efficacy, and was not powered to determine equivalence, the results lend weight to the case for efficacy of short duration concurrent trastuzumab therapy before anthracycline containing chemotherapy. The PTAC considered, inter alia, that the results of ECOG E2198 supported the efficacy of short duration concurrent trastuzumab therapy when administered before anthracycline containing chemotherapy, as demonstrated in the FinHer study.

4. The SOLD study The Synergy or Long Duration (SOLD) study plans to assess the incremental efficacy and risks of adding an extended period of sequential trastuzumab to a short course of concurrent therapy (prior to anthracyclines)xv, to help determine optimum treatment length and sequence for the use of trastuzumab in early stage HER2-postive breast cancer. The trial, planned to enroll 3000 patients internationally, will compare two arms: • nine weeks’ trastuzumab, concurrent with chemotherapy (the FinHer treatment regimen), compared with • the same FinHer treatment regimen plus an additional 42 weeks of trastuzumab (see Figure 16 below). Figure 16. SOLD Study Schema, with detail

The Synergy Or Long Duration (SOLD) Study (Arm A)

A randomised phase III study comparing trastuzumab plus docetaxel (HT) followed by 5-FU, epirubicin, and cyclophosphamide (FEC) (H3-weekly T×3?FE 75C×3) to the same regimen followed by single-agent trastuzumab (H3-weekly T×3?FE 75C×3?H 3-weekly×14) as adjuvant treatments for breast cancer. Radiotherapy is administered at the close of the studies.

Week H8

3 H6

6 H6

10 FEC

13 FEC

16 FEC

D100

D100 Radiotherapy

Continued

60 E N D

The Synergy Or Long Duration (SOLD) Study (Arm B)

Week H8 D100 1 2 3 H6 D100 4 5 6 H6 D100 Radiotherapy 7 8 9 10 FEC 11 12 13 FEC 14 15 16 FEC 17 18 19 H8 20 21 22 H6 23 24 25 H6 26 27 28 H6 29 30

Continued

31 H6

34 H6

37 H6

40 H6

43 H6

46 H6

49 H6

52 H6

55 H6

58 H6

60 E N D

FEC D100 H8 H6

5-FU, epirubicin, and cyclophosphamide 2 Docetaxel 100mg/m Herceptin 8 mg/kg Herceptin 6 mg/kg

xv The SOLD study proposes two adjuvant regimens with different duration of trastuzumab administration are compared in the treatment of early HER2-positive breast cancer. Trastuzumab will be administered in both arms for 9 weeks in combination with docetaxel to exploit the putative synergism between these drugs and to ensure effective adjuvant therapy to all study participants regardless of the result of random allocation. All patients will also receive 3 cycles of anthracycline-containing chemotherapy, and those who have hormone receptor-positive breast cancer will be treated with adjuvant hormonal therapy for a minimum of 5 years. (Joensuu—SOLD protocol)

5. Detailed schema regimens used in RCTs of trastuzumab in early breast cancer (excluding BCIRG 006)

The HERA Study

Cycle Week Diag 1 2 3 Mast or Lump Radiotherapy 4 5 6 7 8 9 10 11 12 1 13 F500 E100 or A60 C500 14 15 2 16 F500 E100 or A60 C500 17 18 3 19 F500 E100 or A60 C500 20 21 4 22 F500 E100 or A60 C500 23 24 1 25 C600 M40 or F600 26 27 2 28 29 30 D100 or P175

D100 C600 or M40 or

P175 F600

3 Continued 31 D100 or P175 or 32 33

4 33 35 36

1 37 H8 38 39

2 40 H6 40 42

3 43 H6 44 45

4 46 H6 47 48

5 49 H6 50 51

6 52 H6 53 54

7 55 H6 56 57

8 58 H6 59 60

C600 D100 M40 or F600 P175

9 Continued 61 H6 62 63

10 64 H6 65 66

11 67 H6 68 69

12 70 H6 71 72

13 73 H6 74 75

14 76 H6 77 78

15 79 H6 80 81

16 82 H6 83 84

17 85 H6 86 87

18 88 H6 E N D

Mastectomy or breast conservative surgery, radio and chemotherapy must be completed before starting Herceptin treatment Herceptin was administered after anthracyclines Duration of the treatment is 88 weeks (26 day care) F500 F600 E100 A60 C500 C600 5-Fluorouracil 500 mg/m2 every 3 weeks 5-Fluorouracil 600 mg/m2 twice every 4 weeks Epirubicin 100 mg/m2 every 3 weeks Doxorubicin 60mg/m2 every 3 weeks Cyclophosphamide 500 mg/m2 every 3 weeks Cyclophosphamide 600 mg/m2 twice every 4 weeks

M40 D100 P175 H8 H6

Methotrexate 40 mg/m2 twice every 4 weeks Docetaxel 100mg/m2 every 3 weeks Paclitaxel 175 mg/m2 every 3 weeks Herceptin 8 mg/kg Herceptin 6 mg/kg every 3 weeks

The FinHER Study

Cycle Week Diag 1 2 3 Mast or Lump 4 H4 1 5 H2 2 6 H2 3 7 H2 4 8 H2 5 9 6 10 H2 7 11 H2 8 12 H2 9 13 H2 1 14 F600 E60 C600 15 16 2 17 F600 E60 C600 18 19 3 20 F600 E60 C600 Radiotherapy 21 22 23 24 25 26 27 E N D

D100-80

Herceptin is administered immediately before anthracyclines Duration of the treatment is 27 weeks (12 day care) Radiotherapy is administered at the end of the study F600 E60 C600 5-Fluorouracil 600 mg/m2 every 3 weeks Epirubicin 60 mg/m2 every 3 weeks Cyclophosphamide 600 mg/m2 every 3 weeks

2 D100-80 Docetaxel 100-80mg/m every 3 weeks H4 Herceptin 4 mg/kg H2 Herceptin 2 mg/kg weekly

The B31-N9831(Arm A) Studies (placebo arms - no trastuzumab)

Cycle Week Diag 1 2 3 Mast or Lump 4 1 5 A60 C600 6 7 2 8 A60 C600 9 10 3 11 A60 C600 12 13 4 14 A60 C600 15 16 1 17 2 18 3 19 4 20 5 21 6 22 7 23 8 24 9 25 10 26 11 27 12 28

P80 P80 P80 P80 P80 P80 P80 P80 P80 P80 P80 P80

13 Continued 29

14 30

15 31

16 32

17 33

18 34

19 35

20 36

21 37

22 38

23 39

24 40

25 41

26 42

27 43

28 44

29 45

30 46

31 47

32 48

33 49

34 50

35 51

36 52

37 53

38 54

39 55

40 56

Radiotherapy

41 Continued 57

42 58

43 59

44 60

45 61

46 62

47 63

48 64

49 65

50 66

51 67

52 68 E N D

Duration of the treatment is 68 weeks (56 day care)

B31-N9831 (Arm B) Studies (trastuzumab sequential to taxane)

Cycle Week Diag 1 2 3 Mast or Lump 4 1 5 A60 C600 6 7 2 8 A60 C600 9 10 3 11 A60 C600 12 13 4 14 A60 C600 15 16 1 17 2 18 3 19 4 20 5 21 6 22 7 23 8 24 9 25 10 26 11 27 12 28

P80 P80 P80 P80 P80 P80 P80 P80 P80 P80 P80 P80

1 Continued 29

2 30

3 31

4 32

5 33

6 34

7 35

8 36

9 37

10 38

11 39

12 40

13 41

14 42

15 43

16 44

17 45

18 46

19 47

20 48

21 49

22 50

23 51

24 52

25 53

26 54

27 55

28 56

Radiotherapy

29 Continued 57

30 58

31 59

32 60

33 61

34 62

35 63

36 64

37 65

38 66

39 67

40 68

41 69

42 70

43 71

44 72

45 73

46 74

47 75

48 76

49 77

50 78

51 79

52 80 E

N D

Herceptin is administered immediately after anthracyclines Duration of the treatment is 68 weeks? (56 day care)

The B31-N9831 (Arm C) Studies (trastuzumab concurent with taxane)

Cycle Week Diag 1 2 3 Mast or Lump 4 1 5 A60 C600 6 7 2 8 A60 C600 9 10 3 11 A60 C600 12 13 4 14 A60 C600 15 16 1 17 2 18 3 19 4 20 5 21 6 22 7 23 8 24 9 25 10 26 11 27 12 28

P80 P80 P80 P80 P80 P80 P80 P80 P80 P80 P80 P80 H4 H2 H2 H2 H2 H2 H2 H2 H2 H2 H2 H2

13 Continued 29

14 30

15 31

16 32

17 33

18 34

19 35

20 36

21 37

22 38

23 39

24 40

25 41

26 42

27 43

28 44

29 45

30 46

31 47

32 48

33 49

34 50

35 51

36 52

37 53

38 54

39 55

40 56

Radiotherapy

41 Continued 57

42 58

43 59

44 60

45 61

46 62

47 63

48 64

49 65

50 66

51 67

52 68 E

N D

Herceptin is administered immediately after anthracyclines Duration of the treatment is 68 weeks (56 day care) A60 C600 P80 Doxorubicin 60mg/m2 every 3 weeks Cyclophosphamide 600 mg/m2 every 3 weeks Paclitaxel 80 mg/m2 weekly

H4 H2

Herceptin 4 mg/kg Herceptin 2 mg/kg weekly

References and endnotes

1 RCTs identified by: Belgian Health Care Knowledge Centre (KCE). Trastuzumab in early stage breast cancer. KCE reports vol. 34C, 2006. http://kce.fgov.be/index_en.aspx?ID=0&SGREF=5211&CREF=7198 2

Piccart-Gebhart M.J. Procter M, Leyland-Jones B, et al. Trastuzumab after Adjuvant Chemotherapy in HER2-Positive Breast Cancer. N Engl J Med 2005;353(16):1659-1672. http://content.nejm.org/cgi/content/full/353/16/1659

On-line Supplementary Appendix to: Piccart-Gebhart MJ, Procter M, Leyland-Jones B, et al. Trastuzumab after adjuvant chemotherapy in HER2-positive breast cancer. N Engl J Med 2005;353:1659-72. http://content.nejm.org/cgi/data/353/16/1659/DC1/1

Smith I, Procter M, Gelber RD, Guillaume S, Feyereislova A, et al. 2 year follow up of trastuzumab after adjuvant chemotherapy in HER2-positive breast cancer: a randomised controlled trial. Lancet 2007;369:29-36. http://www.thelancet.com/journals/lancet/article/PIIS0140673607600282/fulltext

Romond, EH, Perez EA, Bryant J, et al. Trastuzumab plus adjuvant chemotherapy for operable HER-2 positive breast cancer. N Engl J Med 2005;353(16):1659-1672. http://content.nejm.org/cgi/content/full/353/16/1673

6 On-line Supplementary Appendix to: Romond EH, Perez EA, Bryant J, et al. Trastuzumab plus adjuvant chemotherapy for operable HER2-positive breast cancer. N Engl J Med 2005;353:167384,.http://content.nejm.org/cgi/data/353/16/1673/DC1/1 7

Perez EA. Further Analysis of NCCTG-N9831. Slide presentation ASCO annual meeting 2005, available online at http://www.asco.org/ac/1,1003,_12-002511-00_18-0034-00_19-005815-00_21-001,00.asp

8 Slamon D, Eiermann W, Robert N, Pienkowski T, Martin M, et al, on behalf of the BCIRG 006 Investigators. Phase III Trial Comparing AC-T with AC-TH and with TCH in the Adjuvant Treatment of HER2 positive Early Breast Cancer Patients: Second Interim Efficacy Analysis. Slide presentation ASCO annual meeting 2006, available online at http://www.bcirg.org/NR/rdonlyres/eqkdodg2dy7t557o7s6uvj7ytpe6gcfg5gmh2ely6hnhh5pjlabz3nd6jddlnao7qoikej3e dohsijyiisfvp367uuc/BCIRG 006+2nd+Interim+Analysis.pdf 9

Joensuu H, Kellokumpu-Lehtinen PL, Bono P, Alanko T, Kataja V, et al; FinHer Study Investigators. Adjuvant docetaxel or vinorelbine with or without trastuzumab for breast cancer. N Engl J Med. 2006;354(8):809-20. http://content.nejm.org/cgi/content/full/354/8/809

Sledge GW, O’Niell A, Thor AD, Kahanic SP, Zander SP, Davidson NE. Adjuvant trastuzumab: long term results of E2198. Poster presentation 2075, SABCS 2006 http://www.abstracts2view.com/sabcs06/view.php?nu=SABCS06L_561

Numbers of patients in the study arms reported in the HERA trial differ between the first (12 month median followup) and second (23 month median follow-up) interim publications of the trial (Piccart-Gebhart et al. NEJM 2005, Smith et al. Lancet 2007)

12 Schulz K F. Assessing allocation concealment and blinding in randomised controlled trials. Why bother? Evidence Based Medicine 2000; 5:36-37. 13 Jackson R, Ameratunga S, Broad J, Connor J, Lethaby A, Robb G, Wells S, Glasziou P, Heneghan C. The GATE frame: critical appraisal with pictures. Evid Based Med. 2006 Apr;11(2):35-8. http://ebm.bmj.com/cgi/content/full/11/2/35.

S Wells. Summary of GATE critical appraisals of trastuzumab trials. EPIQ, University of Auckland, May 2007. Commissioned by PHARMAC. Further critical appraisal of the five relevant clinical trials (HERA, B31, N9831, BCIRG006, and FinHer) using the full Graphic Appraisal Tool for Epidemiology (GATE) framework (http://ebm.bmj.com/cgi/content/full/11/2/35), to summarise and assess the strength and quality of all the available relevant RCTs. The analysis was restricted to critically appraising the five individual RCTs as published or otherwise reported; it was not intended to integrate the epidemiological evidence with patient preferences, policy issues or clinical considerations (as occurs with page 4 of the GATE instrument), nor provide a meta-analysis or systematic review or other formal policy advice. EPIQ obtained both specialist clinical epidemiological peer review and specialist oncologist content review for each of the GATE appraisals. The GATE frame is a visual framework, designed for critical appraisal, developed by EPIQ. GATE emerged out of the Evidence-Based Medicine (EBM) Working Group users’ guides for appraising for evidence practice (the 28-article series published in JAMA by Drs Sackett, Oxman, Guyatt, Cook, Naylor, etc. (http://www.cche.net/usersguides/main.asp). The framework graphically represents the generic structure of

epidemiological studies and substantively helps systematise critical appraisal in a comprehensive but intuitive way. Details on the GATE framework are available at the EPIQ website (www.epiq.co.nz) and Jackson et al http://ebm.bmj.com/cgi/content/full/11/2/35.

Hind D, Pilgrim H, Ward S. Questions about adjuvant trastuzumab still remain. Lancet 2007; 369:3-5. http://www.thelancet.com/journals/lancet/article/PIIS014067360760004X/fulltext

Hypothetically, all (or almost all) breast cancer cells need to be eradicated for cure. Concomitant administration of the most effective agents available is an obvious strategy to achieve complete eradication of all subclinical cancer. Based on in vitro data, the combination of trastuzumab and docetaxel may be one of the most synergistic ones of all the trastuzumab combinations available to date. These in vitro findings are supported by the clinical data suggesting high activity of the docetaxel plus trastuzumab regimen in the adjuvant, preoperative systemic, and metastatic setting. Long duration of adjuvant administration of single-agent trastuzumab might also result in cancer cell eradication and gradual death of dormant cancer cell populations, although the bulk of evidence suggests that trastuzumab administered in combination with chemotherapy is more effective that trastuzumab given as a single agent. (Joensuu—SOLD protocol)

New Zealand datasheet (Medsafe): Herceptin® Trastuzumab 150 mg and 440 mg powder for concentrate for solution for infusion, http://www.medsafe.govt.nz/profs/Datasheet/h/Herceptininf.htm

Australian datasheet: HERCEPTIN® PRODUCT INFORMATION, dated 21 April 2006. http://www.pbs.gov.au/html/healthpro/search/results?term=herceptin&publication=&form-type=simple

US datasheet: Initial US Approval: September 1998, Revision Date: November 2006. http://www.gene.com/gene/products/information/oncology/herceptin/insert.jsp

21 New Zealand datasheet (Medsafe): Herceptin® Trastuzumab 150 mg and 440 mg powder for concentrate for solution for infusion, http://www.medsafe.govt.nz/profs/Datasheet/h/Herceptininf.htm 22

Herceptin and early breast cancer: a moment for caution. Lancet. 2005 Nov 12;366(9498):1673. http://www.thelancet.com/journals/lancet/article/PIIS0140673605676702/fulltext ‘Although the [Romond 2005] “joint analysis was developed and analysed” by both trial teams, it is unclear whether this synthesis was planned in advance of the start of both trials. The report merely notes that the US FDA and National Cancer Institute approved the joint analysis plan, which may reflect the expectation that neither trial alone would demonstrate a positive result’ (Lancet 2005). Geddes J, Szatmari P, Streiner D. The worm turns: publication bias and trial registers revisited. Evid Based Ment Health. 2004 Nov;7(4):98-9. http://ebmh.bmj.com/cgi/content/full/7/4/98

25 26 27 24 23

Chalmers I. Underreporting research is scientific misconduct. JAMA 1990;263:1405-1408 Antes G, Chalmers I. Under-reporting of clinical trials is unethical. Lancet 2003;361:978-979.

HERA’s previous 12-month median f/u results for overall survival (Piccart-Gebhart et al 2005) did not reach statistical significance. Advocates for the HERA regimen (Hortobagyi 2005) relied instead on extrapolating from the Romond overall survival results.

28 PHARMAC analysis of the two HERA publications (Piccart-Gebhart 2005, Smith 2007). The 23-month median HERA f/u (Smith 2007) comprises cumulative hazards over the entire 23 months, with 218+321=539 first disease recurrence events – in turn comprising events and hazards reported as accruing by the first 12 months median f/u (Piccart-Gebhart) (127+220=347 events) and then the complement of events for the remaining time period (539347=192), i.e. .events occurring in the time since the first analysis. Analysis therefore involves subtracting events recorded as accruing in the first time period (12 months median follow-up) from the cumulative events recorded as accruing over the whole time period analysed to date (23 month median follow-up). This derives numbers of events occurring in the time since the first analysis, in order to calculate period-specific rates and relative risks. In turn this allows testing for statistical interaction (comparing the two time periods) using standard binomial methods:

HERA - 23-month vs. 12 month median follow-ups numbers trastuzum std tmt ab Input data: 12 mth f/u n events 0-12 mths N patients % 23 mth f/u n events 0-23 mths N patients cumulative % calculations 0-12 mths n events N patients % std error of % 13-23 mths n events =23mth f/u N patients minus 12m % std error of % total

difference i.e. tmt effect (ARR) 347 3387

calculated relative risk (RR) published RR HR (95% CI) -95% CI +95% CI

RR reduction (RRR)

127 1694 7.5% 218 1703 12.8%

220 1693 13.0% 321 1698 18.9%

5.5% 0.54 (0.43-0.67) 539 3401 6.1% 0.64 (0.54-0.76)

127 1694 7.5% 0.6% 91 1703 5.3% 0.5%

220 1693 13.0% 0.8% 101 1698 5.9% 0.6%

347 3387 5.5% 0.8% 192 3401 0.6% 0.3% 4.9% 4.9% 3.3% 0.8% 5.915 0.02 +ve evidence of heterogeneity 6.5% 0.90 0.68 1.18 10.2% 0.54 0.58 0.47 0.71 42.3%

difference in treatment effects approximate 95% CI for difference standard error for difference in treatment effects ratio of difference/SE p-value

29 The N9831 Arm C vs. Arm B data await further analysis and publication, as was accorded the Arm C vs. Arm A comparison when this was combined with data from another study in the Romond 2005 joint analysis publication. 30 Belgian Health Care Knowledge Centre (KCE). Trastuzumab in early stage breast cancer. KCE reports vol. 34C, 2006. http://kce.fgov.be/index_en.aspx?ID=0&SGREF=5211&CREF=7198. pp 51, 72.

Testing for proportional hazards over time will be to determine whether the effect of treatment is constant or varies significantly over time.

Matthews JNS, Altman DG. Interaction 3: How to examine heterogeneity. BMJ 1996;313:862. http://www.bmj.com/cgi/content/full/313/7061/862.

Altman DG, Bland JM. Interaction revisited: the difference between two estimates. BMJ 2003;326:219. http://www.bmj.com/cgi/content/full/326/7382/219.

34 PHARMAC analysis of the two conference slide presentations for BCIRG 006, subtracting events recorded as accruing in the first time period (23 months median follow-up) from the cumulative events recorded as accruing over the whole time period analysed to date (36 month median follow-up) to derive numbers of events occurring in the time since the first analysis, then testing for interaction using standard binomial methods (Matthews & Altman 1996, Altman & Bland 2003):

BCIRG 006 concurrent anthracyclines - 36-month vs. 23 month median follow-ups difference numbers calculated relative risk (RR) trastuzum std tmt ab Input data: 23 mth f/u n events 0-23 mths N patients % 36 mth f/u n events 0-36 mths N patients cumulative % calculations 0-12 mths n events N patients % std error of % 24-36 mths n events =36mth f/u N patients minus 23m % std error of % total i.e. tmt effect (ARR) 224 2147 6.5% 0.49 320 2147 6.0% 0.61 published RR HR (95% CI) -95% CI +95% CI

RR reduction (RRR)

77 1074 7.2% 128 1074 11.9%

147 1073 13.7% 192 1073 17.9%

77 1074 7.2% 0.8% 51 1074 4.7% 0.6%

147 1073 13.7% 1.0% 45 1073 4.2% 0.6%

224 2147 6.5% 0.49 1.1% 96 2147 -0.6% 1.13 0.77 1.68 0.3% (uses absolute value (+0.6%) for ARR (-0.6%)) 7.1% 7.1% 4.9% 1.1% 6.363 0.01 +ve evidence of heterogeneity 9.3% -13.2% 0.52 0.40 0.68 47.7%

difference in treatment effects approximate 95% CI for difference standard error for difference in treatment effects ratio of difference/SE p-value

The FinHer study was designed to have a power of 0.80 to detect an increase in five-year recurrence-free survival from 70 percent to 80 percent in the docetaxel-plus-FEC group as compared with the vinorelbine-plus-FEC group (with use of a two-sided test at a significance level of 0.05); approximately 150 events were required for this purpose. The study’s designers estimated at the time that 30% of the participants would have breast cancer with HER2/neu amplification and that the study would be able to detect a difference in their five-year recurrence-free survival of 50 percent to 67 percent at a power of 0.80 when approximately 1000 patients were enrolled. Protocol-defined safety analyses took place in March 2001, September 2001, and December 2002. The protocol specified that safety and early efficacy analyses were to be carried out when the median follow-up time exceeded three years; this point was reached in May 2005, hence the NEJM paper published in February 2005. The final analysis is scheduled to be performed when 150 events have occurred or the median follow-up time exceeds five years. For the primary variable, a p-value of less than 0.029 was considered to indicate significance, in order to maintain an overall type 1 error of 0.05 for the interim and final analysis.

PHARMAC analysis of the FinHer publication:

RR reduction (RRR)

FinHer HER2 +ve pts - docetaxel +/- trastuzumab vs vinorelbine +/- trastuzumab numbers difference calculated relative risk (RR) trastuzum no total ab trastuzum ab n events 3 12 N patients 54 58 % 5.6% 20.7% std error of % 3.1% 5.3% 9 61 14.8% 4.5% 15 58 25.9% 5.7% i.e. tmt effect (ARR) 15 112 15.1% 6.8% 24 119 11.1% 5.8% 4.0% 4.0% -13.4% 21.5% 8.9% 0.453 0.50 no evidence of heterogeneity 0.57 0.27 1.20 0.27 0.08 0.90 RR -95% CI +95% CI

docetaxel + FEC

73.1%

vinorelbine n events + FEC N patients % std error of %

43.0%

difference in treatment effects approximate 95% CI for difference standard error for difference in treatment effects ratio of difference/SE p-value

FinHer disease recurrence event rates in patients with Her2 +ve breast cancer: trastuzumab vs. vinorelbine, by adjuvant chemotherapy (docetaxel, vinorelbine)

30.0%

(Illustrative) relative risks with 95% Cis for FinHer disease events

relative risk (with 95% CI) 0.00 0 0.20 0.40 0.60 0.80 1.00 1.20 1.40

25.0%

HER 2+ve using docetaxel + FEC (trastuzumab vs. no trastuzumab) for illustrative purposes only - unplanned post hoc subgroup analysis with no evidence of statistical heterogeneity 2

20.0%

rate

15.0%

trastuzumab

all HER 2+ve: trastuzumab vs. no trastuzumab

no trastuzumab

10.0%

HER 2+ve using docetaxel + FEC (trastuzumab vs. no trastuzumab)

5.0%

(HER 2+ve, vinorelbine + FEC (trast vs. no trast)) all HER 2+ve, trastuzumab vs. no trastuzumab

HER2 +ve, docetaxel vs. vinorelbine (1/2 of patients also randomised to trastuzumab)

0.0%

D vs V overall

HER2 -ve, D vs. V all, D vs. V

docetaxel + FEC

vinorelbine + FEC

all, HER2 +ve

38 Pegram MD, Lopez A, Konecny G, Slamon DJ. Trastuzumab and chemotherapeutics: drug interactions and synergies. Semin Oncol. 2000;27(6 Suppl 11):21-5 39 40

Rothman KJ. Modern epidemiology. 1st edition. Boston, MA: Little, Brown, 1986

By contrast, in order for HERA to achieve its HR of 0.64 at the 23 month median follow-up, a minimum of around 560-570 patents would have been required. HERA did so with its 3,401 patients, because it was powered to detect changes in overall survival, hence its narrower CI for DFS, albeit with lessened efficacy for point estimates than seen with FinHer.

41 Schulz KF, Grimes DA. Sample size calculations in randomised trials: mandatory and mystical. Lancet. 2005 Apr 915;365(9467):1348-53. http://www.thelancet.com/journals/lancet/article/PIIS0140673605610343/fulltext 42 Sackett DL, Cook DJ. Can we learn anything from small trials? Ann N Y Acad Sci. 1993 Dec 31;703:25-31; discussion 31-2. 43

Altman DG, Bland JM. Absence of evidence is not evidence of absence. BMJ. 1995 Aug 19;311(7003):485. http://www.bmj.com/cgi/content/full/311/7003/485 Concerns about FinHer’s small number of patients seem to relate to the power (1–β) of the study. Power is the probability of detecting a statistically significant difference when a difference of a given magnitude really exists (analogous to sensitivity or ‘true-positives’). Power derives from β error, type II error (β) being the probability of not detecting a statistically significant difference when a difference of a given magnitude in reality exists—i.e. the chance of a false-negative result; mathematically, power is the complement of β (1-β) and represents the probability of avoiding a false-negative conclusion. However, rather than just power, sample sizes depend on four components, being

type I error (α), power (1-β), the event rate in the control group, and the treatment effect (or analogously the event rate in the treatment group) (Schulz & Grimes 2005). Some small trials are so definitively positive that they are sufficient to identify the best therapy (Sackett & Cook DJ 1993). It has been observed that preoccupation with sample size overshadows the more pertinent concerns of elimination of bias, and that unbiased trials with imprecise results trump no results at all (Schulz & Grimes 2005): “Trials should be methodologically strong, thus eliminating bias. Unfortunately, the adequate-power mantra frequently overwhelms discussion on other methodological aspects, e.g. inadequate randomisation usually yields biased results which cannot be salvaged even if a huge sample size generates great precision. By contrast, if investigators design and implement a trial properly, that trial essentially yields an unbiased estimate of effect, even if it has lower power (and precision).”

45 Yoshioka A. Use of randomisation in the Medical Research Council's clinical trial of streptomycin in pulmonary tuberculosis in the 1940s. BMJ. 1998 Oct 31;317(7167):1220-3. http://www.bmj.com/cgi/content/full/317/7167/

Roche media release 3 May 2006. Herceptin approved in Europe for use in combination with an aromatase inhibitor for the treatment of patients with HER2 and hormone receptor-co-positive metastatic breast cancer. http://www.roche.com/med-cor-2007-05-03

Data from final analysis may be available from mid-2007 (where the early efficacy analysis (NEJM 2005) was for when median follow-up exceed three years and was able to take place in May 2005) (Joensuu et al NEJM 2006).

48 The impact of the timing of analysis and hence the potential for extra events to accrue and hence improve precision can be seen by examination of the 23 month HERA results compared with the 12 month HERA results (survival analysis using published hazards (Smith 2007) and numbers of patients at risk (Piccart-Gebhart 2005, Smith 2007)). Similar patterns of increased events would be expected with the FinHer final analysis—both for DFS and overall survival.

Disease events HERA: 12-month interim vs. 23-month median follow-up data

survival analysis using published hazards (Smith 2007) and nos. patients at risk (Piccart-Gebhart 2005, Smith 2007)

350

300

no. first events of disease recurrence or death

HERA 12-month median follow-up

250

HERA interim 12-month analysis, tmt

200

HERA interim 12-month analysis, cntrl

HERA 23-month follow-up, tmt

150

HERA 2-year follow-up, cntrl

100 differences between 12-month and 2-year data affect their hazard ratios 50

0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

years since randomisation

Hortobagyi GN. Trastuzumab in the treatment of breast cancer. N Engl J Med. 2005 Oct 20;353(16):1734-6.

Metadata

Title

Appendix 4 to NZMJ Herceptin - Clinical#2

Abstract

Appendix Four. Clinical effectiveness Contents 1. Treatment regimens and available trial data ….. 1 Treatment regimes in randomised controlled trials… 1 Trial outcomes as reported….. 10 2. Increasing uncertainty around sequence and durations …. 13 Non-publication of data for 12…

Page 1

Note

This text has been extracted from the source PDF document.

Also available as plain text.

Please contact webmaster to discuss alternative format options.