Reduced-Frequency Endocrine Therapy and Challenges of Noninferiority Study Designs

In this issue of JAMA Oncology, Serrano et al present results of a phase 2b trial of postmenopausal women with early-stage breast cancer assessing change in serum estradiol levels and breast tumor Ki-67 after 4 to 6 weeks of 3-times-weekly or once-weekly exemestane reduced-frequency regimens vs the standard 25 mg daily.¹ The primary end point was the noninferiority of change in serum estradiol between the reduced-frequency and standard-frequency arms when estradiol was measured by solid-phase extraction followed by liquid chromatography–tandem mass spectrometry. The 3-times-weekly and once-weekly dosing regimens were selected based on prior observations of exemestane’s long half-life (27 hours) and at least partial suppression of aromatase activity 9 days after a single 25-mg dose.

For women who were compliant with prescribed dosing, the mean relative change in estradiol was −91% for the 3-times-weekly regimen vs −92% for daily exemestane.¹ Noninferiority was defined as having 97.5% confidence that the reduced-frequency regimens did not differ from the daily regimen in estradiol suppression by more than a relative 6%. Consequently, at least for women who were 80% compliant with their dosing, noninferiority was demonstrated for the 3-times-weekly regimen but not the once-weekly regimen (−69% mean change). Similar numerical changes were noted for the intent-to-treat (ITT) analyses. However, the 97.5% CI boundary could not be excluded, and thus the 3-times-weekly regimen could not be considered noninferior by the ITT analysis. No difference was observed among regimens in incidence or severity of adverse events, but treatment duration was likely too short to observe differences, particularly in musculoskeletal symptoms, one of the most frequent causes of aromatase inhibitor (AI) discontinuation.²

The authors suggest that the 3-times-weekly regimen should be studied further in longer-duration primary prevention and adjuvant therapy trials.¹ Should we take their advice, given the inability to demonstrate noninferiority in the ITT analysis of the primary efficacy biomarker or difference in adverse events between the standard and 3-times-weekly arms?

The underlying clinical question is whether there is a “sweet spot” of AI dose or frequency that maintains efficacy but minimizes adverse effects sufficiently to improve uptake and compliance, and avoid premature discontinuation. This is an important question because at least one-third of women started on AI therapy stop prematurely due to adverse effects—most frequently due to musculoskeletal symptoms.²

Suppression of serum estradiol in postmenopausal women is an acknowledged surrogate for suppression of tissue aromatase activity, reducing tissue estrogen levels and estrogen receptor activation. Standard-dose regimens of the 3 later-generation AIs (anastrozole, exemestane, and letrozole) all provide similar disease-free survival in the adjuvant setting, though letrozole may be the most potent inhibitor of aromatase.^3–5 However, there may be a level of systemic estrogen that when exceeded is associated with worse outcomes. Ingle et al,⁶ in a case-control study of postmenopausal women with early breast cancer, found that those with 0.5 pg/mL or higher serum estradiol after 6 months of adjuvant exemestane had a statistically significant 2.2-fold increase in early breast cancer–related events. Note that 0.5 pg/mL is below the 1-pg/mL limit of detection for the solid-phase extraction assay used for the primary analysis of the present study.¹ In total, 69% of women in the daily-dose arm and 65% in the 3-times-weekly arm, but only 17% in the once-weekly arm, had estradiol suppressed below the 1-pg/mL detection limit. Using an ultrasensitive liquid chromatography–tandem mass spectrometry method with an estradiol lower limit of detection of 0.22 pg/mL, the percentage of participants with estradiol suppression below detection limit was 78% for the daily regimen vs 47% for 3 times weekly but only 3% for the once-weekly dose. Change in tumor Ki-67 is another often-used surrogate end point in presurgical trials such as the one reported by Serrano et al. For the secondary end point of change in tumor Ki-67, there appeared to be a stepwise reduction in mean absolute Ki-67 between the standard daily (−7.5%), 3-times-weekly (−5%), and once-weekly (−4%) exemestane regimens, but there was no statistically significant difference among arms. Taken together, changes in both estrogen levels and Ki-67 suggest further study for the 3-times-weekly dose but not for the once-weekly dose despite the inability to demonstrate noninferiority for estradiol suppression in the ITT analysis or a reduction in adverse effects in the reduced-frequency arms.

The concept of noninferiority is often challenging for oncology clinicians who are more comfortable determining if a new treatment is superior to the old by disproving the null hypothesis. Noninferiority trials are usually initiated because a new treatment holds the promise of fewer adverse effects or is less expensive than the standard treatment. The acceptable margin of inferiority is usually determined by expert opinion (as was the case for the present trial¹). As assurance that the new treatment is no worse than the predefined acceptable margin, a confidence interval (usually 95%) must exclude the predefined noninferiority margin. It may be more difficult to demonstrate that a new treatment is noninferior to a standard treatment if there is marked individual variation in effect, the selected margin is too narrow, and/or the selected confidence interval is too high.⁷ Generally, for small noninferiority margins and high confidence intervals, a greater sample size is recommended. For this trial, the sample size may not have been sufficient to overcome the predefined noninferiority margin plus the 97.5% CI.¹ The primary analysis of superiority trials usually assesses data on an ITT basis, which is considered the most conservative. However, primary analysis of a noninferiority trial often targets the predefined compliant population as the most conservative. An ITT analysis could put the new treatment/schedule in a more favorable light if it were likely to be associated with fewer adverse effects and, thus, greater compliance. Generally, findings for both ITT and compliant participants are given.⁷ Considering the similar magnitude of change in estradiol levels for the daily and 3-times-weekly dosing plus the very tight confidence interval, there should not be too much concern that noninferiority was not demonstrated in the ITT analysis as it was for compliant participants and move to the next level of investigation for the 3-times-weekly dose.¹

It should be noted that prolonged holidays for women on AIs are also being investigated in adjuvant treatment trials but with an aim of reversing endocrine resistance as well as reducing long-term adverse effects. The SOLE trial of continuous vs intermittent letrozole (9 months on, 3 months off a daily dose) showed similar efficacy and adverse effects in both arms.⁸ The reduced frequency (3 times weekly) approach, if associated with fewer adverse effects, is likely to be more attractive to patients and deserves further study. Furthermore, the ability to readily monitor estradiol levels would allow conversion to a daily dose if suppression is not optimal.