The Persephone Trial Reconsidered

Part 1: Should We Uniformly Give Only 6 Months of Adjuvant Trastuzumab?

Get Permission

Steven E. Vogl, MD

Steven E. Vogl, MD

The Persephone presentation by Helena Margaret Earl, MBBS, PhD, got a lot of publicity after a pre–ASCO Annual Meeting press release, suggesting that 6 months of adjuvant trastuzumab (Herceptin) is enough. The advice of experts has been that the evidence is inconclusive, but I have not heard a detailed explanation of why not to reduce the duration of trastuzumab on a routine basis in 2018, so I have decided to offer one. Part 1 of this column will review the details of the Persephone study, the results, and issues about the design and the analysis. Part 2, which begins on page 66, will place it in the context of 4 other large randomized trials of trastuzumab duration, none of which could establish noninferiority of a trastuzumab course shorter than 1 year.

The possibility that a very short 9-week course of trastuzumab is as effective as a longer course was raised in 2005 by the small 232-patient FinHER study.1 FinHER showed a reduction in breast cancer recurrences comparable to that achieved by a 1-year course of trastuzumab in 4 much larger studies.

The issue is not whether excluding inferiority to an absolute degree is appropriate, but whether the same degree of inferiority should apply to low- and high-risk patients.
— Steven E. Vogl, MD

Tweet this quote

Persephone is a British study randomly assigning 4,088 women with HER2-positive breast cancer to receive either 6 or 12 months of adjuvant trastuzu-mab. Randomization occurred anytime from diagnosis to the 6-month point. The randomization was stratified by hormone receptor status, chemotherapy program (which drugs), chemotherapy timing (neoadjuvant or adjuvant) and trastuzumab timing (concurrent with chemotherapy or all given after chemotherapy). It was not stratified by prognostic factors like tumor size and nodal involvement. 

Neoadjuvant chemotherapy was given to 15% of participants. Of the remainder, 59% had negative nodes and 69% had positive estrogen receptors in their cancers. The compliance rate was 90% with 6-month trastuzumab and 82% with 12-month trastuzumab. Median follow-up was 5.4 years. The current analysis was based on 512 relapses and 335 deaths. 

The primary analysis revealed a small decrease in invasive disease–free survival at 4 years with the shorter therapy, 89.4% compared to 89.8%. The hazard ratio (HR) was 1.07 (90% confidence interval [CI] = 0.93–1.24). The hazard ratio did not exceed the preset limit of 1.32, so Dr. Earl and her colleagues declared that 6 months of trastuzumab is not inferior to 12 months. 

The inferiority limits were based on a goal of showing with high confidence that 6 months of trastuzumab therapy is not greater than 3% worse in 4-year invasive disease–free survival than 12 months of trastuzumab. The investigators assumed that 4-year invasive disease–free survival would be 80% in the 12-month group and wanted to be sure it was not lower than 77.1% in the 6-month group. 

The use of an absolute difference as an endpoint in a noninferiority study is laudable, since this is what an intelligent patient would demand to know: “How much will I lose if I stop therapy after 6 months?” Its use across a wide range of risk makes one worry that a small absolute (but big relative) increase in events in the low-risk portion of the study population could dilute and mask a bigger absolute difference in higher-risk groups. The issue is not whether excluding inferiority to an absolute degree is appropriate, but whether the same degree of inferiority should apply to low- and high-risk patients. More about this later. 

A Predominantly Low-Risk Population

The 59% of adjuvant patients with negative nodes in Persephone is a much higher percentage than in all the studies that identified the adjuvant activity of trastuzumab. BCIRG 006 is the only first-generation adjuvant trastuzumab study to enroll a significant proportion of women without involved axillary nodes: 29% of the total population. These women did much better than all others in BCIRG 006: Five-year disease-free survival went from 85% to 93% when a year of trastuzumab was added to ACT (doxorubicin, cyclophosphamide, docetaxel) chemotherapy for node-negative patients and from 71% to 80% for node-positive patients.2 It seems that this result established a year of trastuzumab as the standard adjuvant therapy for HER2-positive node-negative breast cancer. 

Dr. Earl presented disease-free survival subgroup analyses for the stratification groups, with evidence that 12 months might be superior for those women who got neoadjuvant chemotherapy with trastuzumab (HR = 1.43) or concurrent adjuvant chemotherapy with trastuzumab (HR = 1.53). She did not present disease-free survival outcome data by primary tumor size or nodal status. 

Concurrent taxane and trastuzumab has been the standard therapy for this population in the United States since 2005, and by 2015, it was given to 71% of British women entered on Persephone. It is possible that the current Persephone analysis was able to exclude inferiority when four other large trials did not because most of the women who were entered early in the trial (and thus with the most mature results) were treated with a suboptimal schedule of trastuzumab and chemotherapy—if the efficacy is less for longer therapy when given only after chemotherapy, then the loss of efficacy with shorter therapy will also be less. 

Overall Survival Slightly Lower With Shorter Therapy, but…

Overall survival slightly favored longer trastuzumab—94.8% vs 93.8% at 4 years (HR = 1.14; 90% CI = 0.95–1.37). Here, the limiting hazard ratio to declare noninferiority was 1.62, and again longer therapy seemed to do a bit better among women who received neoadjuvant therapy or adjuvant concurrent therapy, plus those whose cancers had absent estrogen receptors. 

Overall survival so early in the course of follow-up could be a misleading endpoint: since many distant recurrences occur later than 5 years, and since survival after metastases has gotten as long as a median of 57 months (in the CLEOPATRA trial), many deaths from cancer will be late in follow-up. In contrast, a lot of the early deaths will not relate to metastatic breast cancer in low-risk populations, making the early survival rates more likely to appear similar, whereas the later analyses may show differences of importance to patients. 

Small Node-Negative Tumors Do Great With Paclitaxel and 1 Year of Trastuzumab 

The Dana-Farber APT trial, which defined the current therapy for small node-negative HER2-positive breast cancers, observed a 1% rate of distant metastases at 7 years, but a 2% rate of deaths without metastases.3 The APT study used 12 weekly doses of paclitaxel concurrent with the start of a year of trastuzumab. All the women entered in APT had negative nodes, and primaries had to be smaller than 3.1 cm (most were < 2.1 cm). Of 406 women entered and treated in this single-arm study, only 4 had developed distant metastases at a median follow-up of 7 years. Of these 4, only 1 had an estrogen receptor–negative primary, and only 1 had a primary > 2.0 cm (of 40 entered). All the distant metastases occurred between 27 and 63 months of follow-up. 

Since many distant recurrences occur later than 5 years, and since survival after metastases has gotten as long as a median of 57 months…, many deaths from cancer will be late in follow-up.
— Steven E. Vogl, MD

Tweet this quote

These superb outcomes suggest that de-escalating therapy by deleting AC chemotherapy did not impair results in this very low-risk population. Deleting the AC reduces myelosuppression, late-onset leukemia, alopecia (especially if scalp cooling is used with paclitaxel), and cardiac toxicity. The Dana-Farber group and its collaborators have completed accrual to a randomized study looking at ado-trastuzumab emtansine (Kadcyla, or T-DM1) vs APT in a similar low-risk population. This trial, called ATEMPT, is not powered to compare the efficacy of the two arms. The only other trial of adjuvant T-DM1 monotherapy—the international, multicenter Katherine trial—was recently reported by Roche in a press release (October 15, 2018). In the Katherine study, T-DM1 was superior to trastuzumab in prolonging invasive disease-free survival in a population with residual disease after neoadjuvant chemotherapy. T-DM1 is a very nice drug from the patient’s point of view regarding toxicity, but it is very expensive (about $9,800 per month, according to Forbes). 

In the United States, at least, the APT trial established the standard adjuvant therapy for smaller node-negative HER2-positive breast cancer as 12 weekly doses of paclitaxel beginning concurrently with 1 year of trastuzumab. I would be reluctant to reduce this to 6 months of trastuzumab for women who tolerate trastuzumab well in the first 6 months, unless we have data proving that by doing so, we do not double or triple the rate of distant metastases (ie, to 2% or 3%)! Lead APT investigator Sara Tolaney, MD, MPH, and her colleagues were able to delete the AC from ACTH (doxorubicin, cyclophosphamide, paclitaxel, and trastuzumab). One could argue that we need data specific to the regimen before we halve the duration of trastuzumab with only 12 weeks of a single chemotherapy drug. 

Why Reduce Trastuzumab Duration: The Three Cs

There are three reasons why we might want to use a shorter course of trastuzumab: cost, convenience, and cardiac toxicity. In the current environment in the United Sates, trastuzumab is expensive, but it’s not in the class of new monoclonal antibodies like daratumumab (Darzalex, $16,000 per week). However, for developing countries, trastuzumab is expensive indeed. 

Shorter courses of trastuzumab clearly have activity, are less cardiotoxic, and may be preferred if money for long courses is not available. The convenience issue is minor if the toxicities are mild and the benefits for longer courses are real, though in situations where travel is difficult or infusion resources scarce, the extra 9 to 14 infusions of trastuzumab could be a major issue. 

Cardiac toxicity is the big reason to shorten trastuzumab therapy. It is clear from Persephone and the other randomized studies that ejection fraction decrements are less common with, and recovery of ejection fraction is quicker after, shorter courses.4 However, long-term clinical toxicity after AC × 4 and a full year of trastuzumab is rare in well-done clinical trials, and little late cardiac toxicity has been observed clinically.5 Risk factors are prior hypertension or heart disease, older age, and anthracycline use. 

In the PHARE trial comparing 12 to 6 months of adjuvant trastuzumab in 3,380 women, the number of women with congestive heart failure during follow-up was 11 and 9, respectively (0.65% and 0.53%, a nonsignificant difference).6 Cardiac dysfunction on echocardiogram was largely reversible in PHARE, but the incidence fell from 5.9% for 12 months of trastuzumab to 3.4% for 6 months. 

The Italian Short-HER study compared 52 and 9 weeks of trastuzumab in 1,254 women, all of whom received anthracyclines. The incidence of cardiac events (including major reductions in ejection fraction) decreased from 13.1% to 4.3% when trastuzumab therapy was drastically shortened to 9 weeks. Grade 3 and 4 events decreased from 2.9% to 1.5%.7 

All patients on the SOLD trial received 225 mg/m2 of epirubicin in 3 doses after 9 weeks of docetaxel and trastuzumab. The incidence of any cardiac event increased from 2.0% with 9 weeks of trastuzumab to 3.9% with 52 weeks, and the incidence of congestive heart failure requiring intervention increased from 1.9% to 3.3%.8 In just about every trial, longer trastuzumab increased cardiac toxicity, but cardiac deaths remain rare. 

BCIRG 006 contained a nonanthracycline chemotherapy arm of docetaxel and carboplatin that began concurrently with a year of trastuzumab (TCH). Patients on this arm had much less cardiac toxicity (on echocardiogram testing) than those given the standard AC followed by paclitaxel and trastuzumab (ACTH). The overall survival at 10 years was 3% lower on the TCH arm than the ACTH arm, but the study was not powered to compare these two “experimental” arms with one another. I note that many people, including the BCIRG 006 study chairman, do it anyway!9 

A number of pharmacologic approaches to reducing anthracycline cardiac toxicity have been tried, with mixed results and no clear-cut winner.10 The simplest and least expensive way of preventing anthracycline cardiac toxicity (aside from not giving the drug) is to infuse the drug continuously over 48 to 96 hours rather than give an intravenous bolus.11 This requires only an infusion pump, a port-a-cath, and the correct sort of bent needle. It is inexpensive and well tolerated. I have safely done this in my practice when I thought cardiac function was borderline for safe doxorubicin bolus administration. Prolonged doxorubicin infusions and pegylated liposomal doxorubicin (which has different toxicities than a simple doxorubicin infusion but is less cardiotoxic) have each never been formally studied in large adjuvant breast cancer trials. 

It seems fair to say that cardiotoxicity is generally mild, mostly reversible, and sometimes a problem, but it is less severe without anthracycline and with shorter durations of trastuzumab. For each patient, we need to balance the absolute cardiac risks against the absolute survival benefits of anthracycline chemotherapy and longer-duration trastuzumab. 

Persephone Features Tended to Make Activity of Study Arms Appear More Similar

Table 1 lists features of the Persephone design and analysis that might have led the investigators to conclude the shorter course of trastuzumab is noninferior, although shorter treatment duration could prove inferior in a very large trial done with the proper endpoint. The goal is not to suggest we ignore the results or consider them invalid, but to suggest that we look at them carefully and in more detail, than what we heard presented at the 2018 ASCO Annual Meeting. 

One possible design criticism is that patients were entered before they reached the 6-month point, so many patients in the 12-month arm never even got to month 7. This is partially corrected by the 6-month landmark analysis that Dr. Earl presented in Chicago (which excluded women who did not complete 6 months of trastuzumab). The landmark analysis gave results very similar to the primary analysis. 

British physicians, especially early in the trial, liked to begin trastuzumab after all chemotherapy had been given. This was done for 53% of women on the trial and probably makes trastuzumab less active12—so the 12-month group got more of the drug but started the drug at a suboptimal time. In fact, Dr. Earl noted more benefit from prolonged trastuzumab when the drug was given during chemotherapy! 

TABLE 1: Features of The Persephone Trial That Favored Noninferiority

  • Randomization before the 6-month point—patients about to refuse therapy would not be filtered out in a landmark analysis; they would be if re-consented at 6 months
  • Inclusion of local breast and regional recurrences and deaths without recurrence in the endpoint
  • Study population skewed to those with a better prognosis (node-negative with small primary)
  • Majority of subjects treated with suboptimal sequential schedule

The combination of a low-risk population and the use of the invasive disease–free survival endpoint (rather than a distant metastasis–only endpoint) is particularly likely to drive the two arms toward noninferiority or equivalence. Local recurrences in the previously involved breast or chest wall and nodal recurrences in the axilla, supraclavicular, or internal mammary sites might be influenced by trastuzumab, but they do not have the same grave, uniformly lethal outcome as distant metastases. One would expect variation in surgical technique and radiation to have more influence on ipsilateral breast and regional recurrence than the duration of systemic trastuzumab.

Especially in low-risk populations, many early deaths (tragic though they are) will be from causes other than metastatic breast cancer. One would not expect an extra 6 months of trastuzumab to prevent these deaths. In the APT trial, 8 of 23 disease-free survival events at 7 years were deaths without recurrence, accounting for 2% of entered patients. Dr. Tolaney thus observed twice as many deaths without recurrence as distant metastases at a 7-year follow-up. That means the incidence of distant metastasis would be one-third the incidence of events if coding “distant metastasis or death” (the reciprocal of metastasis-free survival) as the measured outcome. 

New cancer arising in a nonbreast site and new cancer in the contralateral breast, included as a disease-free survival event by the STEEP consortium definition,13 were excluded in the Persephone analysis (Helena Margaret Earl, MBBS, PhD, personal communication, September 6, 2018). Using the STEEP disease-free survival endpoint (as was done in other trastuzumab duration studies), with which about one-third of events are not influenced by the therapy whose duration is being studied, dilutes any difference in distant metastases (the ultimate killer) that may be lurking in the data. Using distant recurrence–free interval (as defined in STEEP) makes much more sense. 

It turns out that 4 other randomized trials of more than 300 patients addressing this question have been reported, each of which has failed to exclude noninferiority for shorter-term trastuzumab. Persephone is thus an outlier in concluding noninferiority. One of them (PHARE) suggests that a year of trastuzumab compared with 6 months of trastuzumab produces an approximately 4% further reduction in distant recurrence at 3 years among women with node-positive disease or those with larger primaries. I review these studies, recommend areas for research, and draw conclusions for current practice in part 2 of this column (page 66 of this issue). ■

DISCLOSURE: Dr. Vogl reported no conflicts of interest. 

At Microphone 1 is an occasional column written by Steven E. Vogl, MD, of the Bronx, New York. When he is not in his clinic, Dr. Vogl can generally be found at major oncology meetings and often at the microphone, where he stands ready with critical questions for presenters of new data.

The opinions expressed in this column are those of the author. If you would like to share your opinion on this or another topic, please write to


1. Joensuu H, Kellokumpu-Lehtinen PL, Bono P, et al: Adjuvant docetaxel or vinorelbine with or without trastuzumab for breast cancer. N Engl J Med 354:809-820, 2006.

2. Slamon D, Eirmann W, Robert N, et al: Adjuvant trastuzumab in HER2-positive breast cancer. N Engl J Med 365:1273-1283, 2011.

3. Tolaney SM, Barry WT, Guo H, et al: Seven-year follow-up of adjuvant paclitaxel and trastuzumab for node-negative, HER2-positive breast cancer. 2017 ASCO Annual Meeting. Abstract 511

4. Earl HM, Vallier AL, Dunn J, et al: Trastuzumab-associated cardiac events in the Persephone trial. Br J Cancer 115:1462-1470, 2016.

5. Advani PP, Ballman KV, Dockter TJ, et al: Long-term cardiac safety analysis of NCCTG N9831 (Alliance) adjuvant trastuzumab trial. J Clin Oncol 34:581-587, 2015.

6. Pivot X, Suter T, Nabholtz JM, et al: Cardiac toxicity events in the PHARE trial, an adjuvant trastuzumab randomized phase III study. Eur J Cancer 51:1660-1666, 2015.

7. Conte P, Frassoldati A, Bisagni G, et al: 9 weeks vs 1 year adjuvant trastuzumab in combination with chemotherapy: Final results of the phase III randomized Short-HER study. Ann Oncol. September 13, 2018 (early release online).

8. Joensuu H, Fraser J, Wildiers H, et al: Effect of adjuvant trastuzumab for a duration of 9 weeks vs 1 year with concomitant chemotherapy for early human epidermal growth factor receptor 2-positive breast cancer: The SOLD randomized clinical trial. JAMA Oncol 4:1199-1206, 2018.

9. Vogl SE: Adjuvant doxorubicin for HER2-positive breast cancer. Clin Oncol News, March 1, 2016. 

10. Payne DL, Nohria A: Prevention of chemotherapy induced cardiomyopathy. Curr Heart Fail Rep 14:398-403, 2017.

11. Legha SS, Benjamin RS, Mackay B, et al: Reduction of doxorubicin cardiotoxicity by prolonged continuous intravenous infusion. Ann Intern Med 96:133-139, 1982.

12. Perez EA, Suman VJ, Davidson NE, et al: Sequential versus concurrent trastuzumab in adjuvant chemotherapy for breast cancer. J Clin Oncol 29:4491-4497, 2011

13. Hudis CA, Barlow WE, Costantino JP, et al: Proposal for standardized definitions for efficacy end points in adjuvant breast cancer trials: The STEEP system. J Clin Oncol 25:2127-2132, 2007.

Related Articles

The Persephone Trial Reconsidered: Part 2: Persephone Is an Outlier—Four Other Large Trials Could Not Establish Noninferiority

Steven E. Vogl, MD

Steven E. Vogl, MD

Persephone is a 4,088-patient trial that Helena Margaret Earl, MBBS, PhD, reported at the 2018 ASCO Annual Meeting as establishing that 6 months of trastuzumab (Herceptin) is not inferior to 12 months in 4-year survival without invasive or local regional recurrence or...