DESTINY-Breast05 Safety Analysis Clarifies ILD and Radiation Risks With T-DXd in HER2+ Breast Cancer

An updated safety analysis of the phase 3 DESTINY-Breast05 trial (NCT04622319) showed no meaningful difference in interstitial lung disease (ILD) rates between concurrent and sequential radiation approaches in patients with HER2-positive breast cancer receiving fam-trastuzumab deruxtecan-nxki (T-DXd; Enhertu), distinguished drug-related interstitial lung disease (ILD) from radiation pneumonitis, and demonstrated that T-DXd rechallenge after resolved grade 1 ILD rarely triggered ILD recurrence, according to Charles E. Geyer, MD.

“The data show that either [concurrent or sequential radiation therapy administration] is reasonable, and investigators can figure out what works best for their patients,” Geyer said in an interview with OncLive®. “I tend to like the sequential [approach] because it gives patients a break from drug therapy, but that’s just my opinion.”

In the interview, Geyer discussed findings from the safety analysis, which were presented at the 2026 ASCO Annual Meeting (ASCO 2026). He also shared practical advice for identifying and managing toxicities in patients receiving T-DXd.

Geyer is division chief of Malignant Hematology and Medical Oncology in the Department of Medicine at the University of Pittsburgh School of Medicine, as well as a breast medical oncologist at the University of Pittsburgh Medical Center (UPMC) Magee Womens Hospital at the UPMC Hillman Cancer Center in Pittsburgh.

OncLive: What findings from DESTINY-Breast05 have been previously reported?

Geyer: DESTINY-Breast05 is an open-label, randomized trial that was conducted in patients with high-risk, HER2-positive early breast cancer who had received standard neoadjuvant chemotherapy with HER2-directed monoclonal antibodies, most commonly the dual antibodies trastuzumab and pertuzumab [Perjeta], and were found to either have residual disease in their axillary nodes when they had surgery, or had presented with locally advanced cancer and had disease in the breast, the nodes, or both. These were higher risk, non–pathologic complete response patients.

The way we came up with that cohort was by looking at our initial results from the phase 3 KATHERINE trial [NCT01772472] that established ado-trastuzumab emtansine [T-DM1; Kadcyla] for 14 cycles as standard of care for patients with any amount of residual invasive breast cancer after standard neoadjuvant therapy. When we first reported that result, we had 3-year invasive disease–free survival [IDFS] risk data, and we saw that of the patients who still had positive nodes, approximately 17% had recurred early, and in patients who had presented with locally advanced disease, approximately one-quarter had recurred. Even though T-DM1 cut their relative risk approximately in half, there was still an unacceptably high risk for those patients. This was a clear population that had an identifiable unmet need.

In DESTINY-Breast05, we randomly assigned patients in that population and compared 14 cycles of standard T-DM1 vs T-DXd. The study accrued 1635 patients globally over a 4-year period, and we reached the pre-specified number of events for the interim analysis with a median follow-up of 30 months.¹ The 3-year IDFS rate was improved from 83.7% [95% CI, 80.2%-86.7%] with T-DM1 [n = 817], which is what we had predicted, to 92.4% [95% CI, 89.7%-94.4%] with T-DXd [n = 818], which corresponded to a hazard ratio of 0.47 [95% CI, 0.34-0.66; P < .0001].

These were remarkable data. T-DM1 cut the risk by approximately 50% relative to trastuzumab. T-DXd cut it by another 53% relative to T-DM1. The results quickly led to regulatory approval in the patient population we enrolled in DESTINY-Breast05, but interestingly enough, the FDA broadened [this indication] to any patient with residual invasive breast cancer, getting back more to a KATHERINE population with eligibility based on the magnitude of benefit and the overall safety. The overall survival [OS] data were immature, but we were already seeing fewer deaths with T-DXd, 18 total deaths vs 29 deaths with T-DM1.

Safety-wise, we expected we were going to see ILD in patients receiving T-DXd, and the rates of ILD were higher [in that arm]. The rate of any-grade ILD was 9.6% with T-DXd compared with 1.6% with T-DM1. Fortunately, the majority of that ILD, 9.6% and 8.5%, respectively, was grade 1 or 2, and [these events] were largely reversible with discontinuation of T-DXd. But a note of caution was there were 2 deaths due to ILD.

What was the importance of conducting a safety analysis of DESTINY-Breast05 specifically looking at ILD and radiation pneumonitis?

ILD was a well-recognized and potentially fatal adverse effect [AE] associated with T-DXd in the metastatic setting at the time we were designing DESTINY-Breast05, and we were bringing that drug into a high-risk but curative patient population. It was important to incorporate appropriate monitoring for symptoms and findings consistent with ILD, and to have conservative measures to hold or discontinue T-DXd when drug-related ILD was identified. We gave investigators the discretion to administer radiation therapy prior to starting study medication or concurrently with the study medication. We have detailed information available on the relative incidence of ILD and radiation pneumonitis with the 2 study approaches, as well as information on possible differences in efficacy outcomes [between the arms].

The ASCO 2026 abstract went into that in greater depth than we were able to present in the original presentation at the 2025 ESMO Congress. An important aspect of the trial that is good to be aware of is that we had a panel of experts on the recognition and management of ILD, both imagers and pulmonologists, who helped develop the plan that included serial safety imaging in the trial with low-dose, non-contrast CT scans to monitor for the early onset of ILD, [which is] essentially asymptomatic, when we thought that interventions would be most likely to be effective and produce full recovery without sequelae. The schedule on the trial was that scans were obtained at baseline, after the initial 6 weeks of therapy, and every 12 weeks until therapy was completed; then we had a 40-day scan after the last cycle to get a final look for possible AEs.

On the trial, we had strict criteria that required T-DXd to be discontinued for any grade 2, 3, or 4 ILD; any symptomatic drug-related ILD required discontinuation of the medication. Importantly, with grade 1, which is asymptomatic ILD found on imaging, we held T-DXd and recommended steroid therapy, then watched to see how long it took for that change noted on CT scans to resolve. If it resolved quickly within 28 days, patients could be rechallenged without a dose reduction. If it took longer, getting out as far as 125 days, you could rechallenge, but then you should drop the dose. If the patient still hadn’t recovered at 126 days, then T-DXd was discontinued. Physician discretion was always allowed, but these were the broad guidelines and parameters that governed the safety portion of the study.

Another factor we had to think about was the reality that these patients with high-risk breast cancer are generally going to require post-mastectomy regional nodal radiation therapy; especially if they opt for breast conservation, [they should receive] breast radiation therapy. If we get subsequent imaging, we’ll see little focal areas of radiation pneumonitis. We knew we would be seeing those as well in the patients who had radiation therapy.

We all kept our eye out for symptoms like dry cough and new-onset dyspnea that make us wonder about radiation pneumonitis, which would be grade 2, but we knew we needed to manage the grade 1 radiation pneumonitis differently than the grade 1 ILD associated with the drugs. On the trial, in these asymptomatic cases of grade 1 radiation pneumonitis, which you can differentiate from drug-related pneumonitis primarily because it’s within the radiation field of the patient’s treatment, we did not require holding T-DXd. However, if patients developed symptoms, then we recommended holding the drug until the symptoms resolved, and then they could resume it.

We did not require radiation pneumonitis to resolve before resuming T-DXd. We did require drug-induced ILD changes to resolve. We also made sure to provide a summary of how we managed ILD and radiation pneumonitis in the supplemental appendix of the primary publication of the DESTINY-Breast05 trial in The New England Journal of Medicine. That’s a good source to have handy if you are treating a patient and run into toxicity.

What were some of the key findings from the updated DESTINY-Breast05 safety analysis?

At the initial presentation in the presidential symposium at ESMO 2025, we could only provide top-line results of the safety analysis. We wanted to provide more details at a subsequent meeting, and that’s what Michael Untch, MD, [of the Breast Cancer Center, HELIOS Klinikum Berlin-Buch GmbH in Berlin, Germany], presented at an oral session at ASCO 2026. We allowed investigators to decide if and when they were going to do radiation therapy, [as well as whether they wanted to] administer it prior to the study medication or co-administer radiation therapy and the study drug.

We had 2 cohorts of patients: patients treated sequentially, and those treated concurrently; we were interested to see whether they had different AE profiles. Much of what Untch presented was the granular data on those differences, that also looked at when ILD occurred, how often did it completely resolve? How often did [ILD] improve, but not quite resolve?

Regarding the timing of the radiation therapy, there were no real differences in the rates or maximal grades of ILD between the 2 approaches. We saw a single case of grade 5 ILD occur with each approach.² We had 2 deaths on the study, and as it happens, 1 of these patients got treated sequentially, and 1 got treated concurrently.

With the concurrent approach [n = 438], we had 42 cases of ILD develop, which amounted to 9.6% of the patients in that cohort. With initial radiation therapy followed by treatment [n = 319], we had 34 cases, or 10.7% of the patients in that cohort, so similar [rates]. There were more patients treated concurrently than sequentially, so you have to look at the percentages. There’s not a statistical difference of any magnitude there, and the clinical difference is not all that important either.

Of the 42 concurrent cases, most patients, 29, recovered. Five patients were improving, but still had some changes. Seven patients did not recover, and we had 1 death. Most of the patients with this approach recovered or were on their way to recovering at the last data cutoff.

Among the sequential cases, these numbers were similar. Twenty patients recovered, 5 patients were recovering, 8 patients had not recovered, and there was 1 death, Again, there wasn’t a big difference in terms of how patients were recovering.

Regarding how long it took to recover, the median duration of ILD was 77.0 days with the sequential approach and 67.0 days with the concurrent approach. There does not appear to be a clear advantage to either the sequential approach or the concurrent approach. Regarding the overall incidence of the most severe grade and the recovery from ILD, if you look at the efficacy, the concurrent [IDFS] HR for T-DXd relative to T-DM1 was 0.35 [95% CI, 0.19-0.64]. With the sequential approach, it was 0.55 [95% CI, 0.35-0.85].¹ With the concurrent approach, the confidence intervals were overlapping, so there’s not a statistical difference there, but it is interesting that there was that difference in terms of the HR that seemed tilted a bit toward the sequential approach. However, the data show that either approach is reasonable. Regarding radiation pneumonitis, the trial design did give us an opportunity to identify how often patients had asymptomatic radiation pneumonitis associated with T-DM1.

The rates of symptomatic radiation pneumonitis in KATHERINE were low, but we did [this DESTINY-Breast05 analysis] in both patients who received T-DM1 and those who received T-DXd, so the information was there and was of interest. With the sequential approach, the rate of any-grade radiation pneumonitis was 34.5% with T-DXd, and with the concurrent approach, it was a bit lower, at 29.2%.² With T-DM1, the corresponding rates were 37.4% and 26.7%, so there did not seem to be higher levels of grade 1 radiation pneumonitis in patients who received T-DXd vs T-DM1, which is reassuring.

Not surprisingly, complete recovery with resolution of imaging findings occurred less often with radiation pneumonitis than with ILD. Clinically, we often see that for years later, patients have a bit of scarring in the lungs where they were in the radiation field, so this wasn’t surprising and provides evidence that with early recognition and early treatment, drug-induced ILD should be reversible in most patients. This heightens the need to be attentive to that.

One other thing we were able to look at that was reassuring was that in the trial, if grade 1 ILD recovered on steroids during withholding of T-DXd, meaning that the patient’s CT scan results normalized, you could rechallenge the patient. We have information on how the patients did when we decided to rechallenge them after their grade 1 T-DXd–related pneumonitis resolved. In the sequential arm, no patients who were rechallenged developed recurrent ILD. In the concurrent arm, only 2 patients who were rechallenged developed recurrent ILD, so rechallenge appears to be an appropriate consideration for high-risk patients. This speaks to the need to, if ILD starts to develop, treat it and give the patient a rest, because you can probably rechallenge and get them on medication again without recurrence. This is useful information to have in clinic.

The other interesting finding was that the risk of ILD was slightly higher in patients who had moderate renal impairment. For those with normal renal function [n = 506], it was 9.7%, but for those with moderate renal impairment [n = 42], it was 14.3%. We also saw that ILD rates were higher among Japanese patients vs those from the rest of the world. This subgroup [n = 87] had an all-grade ILD rate of 14.9% compared with 8.9% for the rest of the world [n = 719].

How are you incorporating CT scans into your practice for ILD/radiation pneumonitis identification?

The data support the use of serial CT scans to try to identify grade 1 asymptomatic ILD, which appears to have optimized recovery, with T-DXd still providing a benefit. The T-DXd label does not require that. It’s left to physician discretion, but I am going to continue incorporating that approach in my patients. It’s important, if you order CT scans, to spend time talking with your radiologist.

You can use low-dose, non-contrast CT scans. You need to understand that you’re looking for radiation pneumonitis, so that if you just order a CT scan, you’re going to get a lot of noise on that.

Supported in part by AstraZeneca and Daiichi Sankyo, content independently developed by OncLive.

References

1. Geyer CE, Park YH, Shao Z, et al. Trastuzumab deruxtecan (T-DXd) vs trastuzumab emtansine (T-DM1) in patients (pts) with high-risk human epidermal growth factor receptor 2–positive (HER2+) primary breast cancer (BC) with residual invasive disease after neoadjuvant therapy (tx): interim analysis of DESTINY-Breast05. Ann Oncol. 2025;36(suppl 2):S1556-S1557. doi:10.1016/j.annonc.2025.09.021
2. Untch M, Shao Z, Huang CS, et al. Secondary safety analysis of trastuzumab deruxtecan (T-DXd) vs trastuzumab emtansine (T-DM1) in DESTINY-Breast05: clinical and demographic risk factors of interstitial lung disease (ILD) and radiation pneumonitis (RP). J Clin Oncol. 2026;44(suppl 16):516. doi:10.1200/JCO.2026.44.16_suppl.516