Effects of Supervised Exercise during Neoadjuvant Chemotherapy on Tumor Response in Patients with Breast Cancer (Neo-train): A Randomized Controlled Trial

Abstract

Purpose:

Preclinical studies have indicated that physical exercise may enhance chemotherapy efficacy. However, clinical trials are needed to investigate these findings. We hypothesized that an exercise program during neoadjuvant chemotherapy would improve tumor response in patients with breast cancer.

Patients and Methods:

Neo-train was a randomized controlled trial allocating patients with breast cancer to the usual care control (CON) group or the exercise (EX) group that received supervised high-intensity interval training and progressive resistance training 3 times weekly during 18 to 24 weeks of neoadjuvant chemotherapy. The two groups were compared on tumor size, assessed using magnetic resonance imaging as primary outcome, and secondary clinical/pathologic, biological, physical, and patient-reported outcomes.

Results:

From 2021 to 2023, 102 participants were randomly assigned to the EX (n = 50) or CON (n = 52) group. We found no between-group differences in median tumor size change from baseline to presurgery (EX vs. CON −3.0 mm [95% confidence interval (CI), −8.0 to 14.0]), the proportion with radiologic complete response [EX 65% vs. CON 56%; odds ratio 1.16 (95% CI, 0.39–3.91)], or pathologic complete response [EX 59% vs. CON 56%; odds ratio 1.03 (95% CI, 0.43–2.46)]. The exercise program was associated with higher relative dose intensity, fewer dose delays, shorter hospital stays, and increased cardiorespiratory fitness, muscle strength, and level of physical activity. We found no differences in tumor-infiltrating lymphocytes, body composition, health-related quality of life, anxiety, depression, psychological distress, or participation in rehabilitation between groups.

Conclusions:

Although the exercise program did not affect tumor size, the positive effects on chemotherapy completion and shorter hospitalizations suggest improved treatment tolerance.

Translational Relevance.

Neo-train was a randomized controlled trial testing a supervised exercise program among patients undergoing 18 to 24 weeks of neoadjuvant chemotherapy for breast cancer. Contrary to our hypothesis, the program with combined aerobic and resistance exercise did not significantly reduce tumor size as a primary outcome. However, it seemed to enhance physical health, improve chemotherapy completion rates, and shorten hospital stays, indicating better treatment tolerance. These findings support the potential of exercise as a valuable supplement to neoadjuvant chemotherapy to improve patient outcomes, meriting further investigation.

Introduction

Physical exercise is generally advised during and following cancer treatment to improve cardiorespiratory fitness, muscle strength, fatigue, anxiety, and depression (1, 2), which may enhance treatment tolerability. Preclinical studies have suggested that exercise inhibits primary tumor growth (3–5) and induces favorable changes in the tumor microenvironment such as increased stimulation of immune cells to infiltrate the tumor (6) and normalized vascularization, potentially enhancing the delivery and efficacy of chemotherapy (7). These findings have led to a growing interest in translating these benefits to clinical settings, yet the direct impact of exercise on tumor response during chemotherapy in humans remains a subject of investigation (8).

Neoadjuvant chemotherapy (NACT) is widely recommended for patients with early breast cancer (9). Maintaining the planned dose intensity is crucial for optimizing response and reducing recurrence (10). Achieving pathologic complete response (pCR), defined as the absence of residual disease in the breast and lymph nodes following NACT, is associated with improved disease-free survival (11). In patients with breast cancer, NACT offers a unique window for intervention to examine the added effect of exercise on tumor response over time and to study other clinically relevant measures to better understand the underlying mechanisms by which exercise may enhance treatment response.

In a randomized controlled trial (Neo-train), we tested the hypothesis that supervised exercise 3 times per week during NACT would improve tumor response in patients with breast cancer. We hereby report the results from the primary outcome tumor size measured using magnetic resonance imaging (MRI). In addition, we report on secondary outcomes including residual tumor after NACT, tumor-infiltrating lymphocytes (TIL), chemotherapy completion, hospital admissions, cardiorespiratory fitness, muscle strength, body composition, level of physical activity, participation in municipality-based rehabilitation, health-related quality of life (HRQL), anxiety, depression, and psychological distress. The large number of secondary outcomes was chosen to give a comprehensive understanding of the potential benefits of exercise on patients’ abilities to withstand intensive cancer treatment and the underlying mechanisms through which exercise might influence tumor biology and treatment efficacy.

Patients and Methods

Study design and participants

Neo-train was a two-arm, parallel randomized controlled trial, which was developed with patient involvement and pretested in a pilot study as reported in detail previously (12). Briefly, participants were randomly allocated to either the usual care control (CON) group or the exercise (EX) group which received supervised high-intensity interval training (HIIT) and progressive resistance training (PRT) three times a week throughout the duration of NACT (18–24 weeks). Participants were recruited from the outpatient clinic, Department of Clinical Oncology and Palliative Care, Zealand University Hospital, Naestved. This department provides all oncological treatment for patients with breast cancer living in Region Zealand, one of the five healthcare regions in Denmark, mainly comprising a low- to middle-income population of around 850,000 inhabitants.

The study was conducted in accordance with the Declaration of Helsinki and approved by the Committee on Health Research Ethics of Region Zealand (number SJ-827) and preregistered at clinicaltrials.gov (NCT04623554) on November 10, 2020. All participants signed written informed consent. The trial was reported according to the Consolidated Standards of Reporting Trials (CONSORT) for parallel group randomized trials (checklist available in Supplementary Material S1; ref. 13).

Eligibility criteria and recruitment

Inclusion criteria were female patients of ≥18 years of age who were newly diagnosed with histologically verified breast cancer and agreeing to start standard NACT according to the national Danish Breast Cancer Cooperative Group guidelines (14). Exclusion criteria were contraindications to MRI, severe physical or mental comorbidity preventing participation in exercise, inability to read and understand Danish, or the patient deemed not suitable for participation by the oncologist. Eligible patients were invited by a project team member either at the first consultation with an oncologist or on the day of the first administered dose of NACT. Patients who declined participation in Neo-train were asked if they would agree to complete the baseline questionnaire and allow data collection of clinical variables from medical records to contribute to the study as a nonparticipation (NON) group to assess generalizability.

Randomization and blinding

After the baseline assessment, participants were randomly assigned 1:1 through concealed computer-based randomization in blocks of four to the CON group or EX group. Randomization was stratified by planned NACT regimens to account for their expected different influence on treatment response. As new NACT regimens were introduced into standard care during the study period, each was assigned a new stratification group, resulting in a total of 10 groups. Randomization lists were generated independently by a data manager and revealed in the randomization function of REDCap (version 13.7.14 LTS, RRID: SCR_003445; ref. 15). It was not possible to blind the participants, project staff, or care providers. However, the assessors (radiologists and pathologists) of the tumor-related outcomes, the data manager, and statisticians were blinded to the group allocation.

Usual care control

Participants in the CON group were not given any specific preoperative exercise instructions. They were advised to maintain their regular routines including taking part in municipality-based rehabilitation programs if referred through routine clinical practice.

Exercise program and screening for psychological distress

The EX group received the supervised exercise program concurrently with usual care. Details were described previously (12), including recommended elements from the Consensus on Exercise Reporting Template (CERT) and the Template for Intervention Description and Replication (TIDieR). Briefly, the exercise program was designed to provide exercise commencing at each participant’s initial physical fitness level and gradually intensifying the workload to accommodate any adaptation. In groups of up to three, participants exercised for 1 hour on Mondays, Wednesdays, and Fridays starting in the week following randomization and ending in the week before surgery (duration depended on the specific NACT regimen). Participants were instructed to attend sessions at least twice weekly but encouraged to attend all three sessions weekly. Sessions encompassed 30 minutes of HIIT on a stationary exercise bike and 30 minutes of PRT under the direct supervision of a physiotherapist. At the start of the study, the physiotherapists received 3 hours of theoretical and practical training on delivering the exercise program as outlined in the manual. This was followed by peer-to-peer training to reinforce fidelity to the protocol and to document any modifications in exercise duration or intensity, as well as any adverse events, in the exercise logs. HIIT included a 5-minute warm-up followed by four intervals of high-intensity cycling lasting 2 minutes each, separated by 4-minute periods of low-intensity cycling, and concluded with a 5-minute cooldown. Intervals of high intensity were primarily guided by reaching ≥85% of the maximum heart rate. However, in cases of issues with the activity tracker—such as when a participant forgot to wear the device—intensity was instead guided by a rating of ≥16 on the ratings of perceived exertion (Borg) scale (16). PRT involved three sets of 12 to 15 repetitions on pull down, chest press, and leg press, respectively. The prescribed load was based on an indirect one-repetition maximum (1RM) estimated from an eight-repetition maximum test at baseline, and subsequently starting with 65% of 1RM in weeks 1 to 3 followed by an increase in load of 10% from week 4 onward if the participant could complete ≥17 repetitions with good quality. Sessions took place at five physiotherapy departments across the region, at hospitals or a municipality rehabilitation center, with the aim to provide close-to-home exercise and minimize transportation time. Partial compensation for transportation costs was offered. A project team member contacted participants by telephone or face-to-face in the outpatient clinic at least four times (weeks 4, 7, 13, and 19) during the exercise program to increase motivation and discuss barriers for attendance and adherence.

During the study, participants in the EX group were screened for psychological distress using the Distress Thermometer (DT; ref. 17) at baseline, week 13, presurgery, and 3 months following surgery. The DT is a single-item tool with a 0 (no distress) to 10 (extreme distress) numeric rating scale (17). Participants scoring between 7 and 10, indicating moderate to severe psychological distress (18), were advised by a project team member to visit their nearest free, drop-in counseling center available at nine locations at Danish Cancer Society counseling centers across Region Zealand.

Outcomes, measurements, and sources

Primary outcome

MRI

The primary outcome was the change in tumor size in millimeters (mm) from pre-NACT to presurgery assessed using MRI. This was chosen to enable dynamic monitoring of treatment response concurrently with the exercise program. The MRI scan timing and frequency were individualized based on the NACT regimen and each participant’s treatment response. Initially, scans were evaluated by breast imaging radiologists as per standard care. Second, a senior breast imaging radiologist (S.G. Ceballos) reassessed all scans for consistency and registered the maximum diameter of the largest lesion in mm for trial purpose provided with supplementary remarks on fragmented shrinking patterns if the lesion’s diameter remained unchanged yet displaying a partial response with more dispersed islands of tumor cells, high apparent diffusion coefficient levels, or less intense contrast enhancement. In such cases, the unchanged diameter was registered, recognizing that it might underestimate the true response. Absence of contrast enhancement was registered as a tumor size of 0.0 mm [classified as radiologic complete response (rCR)].

Secondary outcomes

Secondary outcomes were measures of residual tumor after NACT, TILs, chemotherapy completion, hospital admissions, cardiorespiratory fitness, muscle strength, body composition, level of physical activity, HRQL, anxiety, depression, psychological distress, and participation in municipality-based rehabilitation. These outcomes were measured at baseline (the week before to 1–2 weeks after the first dose of NACT), during NACT (approximately week 13), and presurgery (approximately weeks 21–29). Residual tumor and TILs were measured at surgery (approximately weeks 21–30). Cardiorespiratory fitness, muscle strength, body composition, level of physical activity, HRQL, anxiety, depression, psychological distress, and participation in municipality-based rehabilitation were measured additionally at 3 months following surgery (approximately weeks 33–42; ref. 12).

Pathology reports and histopathologic analysis

Residual tumor was assessed as tumor size in mm from the surgical specimen and residual cancer burden (RCB) as per standard care based on national guidelines (14). RCB was classified as pCR (RCB-0), minimal (RCB-I), moderate (RCB-II), and extensive (RCB-III) residual disease. Stromal TILs from the pre-NACT diagnostic needle biopsy and post-NACT surgical specimen (in the latter in case of residual disease) were assessed both manually (expressed as percentage) and using open-source software QuPath version 0.4.3 (The University of Edinburgh) with the pretrained cell detection model StarDist (expressed as density per 10,000 square micrometers; refs. 19, 20). Manual assessment, carried out by an experienced pathologist (A.-V. Lænkholm) and a resident pathologist trained for the task (D. Rasic), was performed as suggested by the TIL working group 2014 and 2018 (further description in Supplementary Material S2; refs. 21, 22).

Medical records

Clinical data on tumor characteristics, performance status, comorbidity, and prescription medications were extracted at baseline. For chemotherapy completion, the date and dose of each chemotherapy drug administered, dose reductions, dose delays ≥3 days, early discontinuation, and the main reason for each dose modification were abstracted after NACT. Relative dose intensity (RDI) was defined as the percentage of delivered dose intensity (total administered mg/m² in actual number of weeks) to the prescribed dose intensity (total standard mg/m² in planned number of weeks). RDI was calculated separately for each drug, using the total administered dose, and then averaged to obtain a total RDI for each patient (23). RDI was dichotomized as <85% and ≥85% with low RDI (<85%) indicating lower dose than that recommended in the guidelines with expected poorer chemotherapy effectiveness and prognosis (24). Hospital admissions were defined as hospitalizations ≥24 hours due to toxicity-related reasons (not diagnostics) from baseline to presurgery. Total length of stay during NACT was registered as the sum of days for one or more hospitalizations.

Physical tests and bioimpedance analysis

Cardiorespiratory fitness was measured as indirectly estimated maximum oxygen consumption (VO₂max) using a progressive cycle ergometer test (Monark Ergomedic 828E; ref. 25) and reported as absolute VO₂max (L/minute) and relative VO₂max (mL/minute/kg). During the test, the heart rate was documented using a wrist-worn activity tracker (vivosmart 4, Garmin International Inc.), and participants were instructed to continue the test until reaching ≥16 on the Borg scale (16) or feeling high discomfort or pain. Muscle strength was assessed as an indirect 1RM on pull down (BH L110 Lat Pulley, BH Fitness, United Kingdom) and leg press (Artis Leg Press, Technogym) in kg and as maximal isometric handgrip strength in kg using a hand dynamometer (DHD-1 Digital Hand Dynamometer, Saehan Corporation). Body composition was measured using bioelectrical impedance analysis (seca medical Body Composition Analyzer 515, seca GmbH & Co.) prior to physical tests, obtaining information on fat mass, fat-free mass, skeletal muscle mass, and body mass index. Tests followed a structured testing manual by a trained project team member who recorded results in electronic testing logs.

Questionnaires

At baseline, participants completed electronic questionnaires assessing cohabitation status, education level, work market affiliation, and smoking history. Level of physical activity was assessed at baseline, week 13, presurgery, and at 3 months following surgery using two study-specific items. The items included information on weekly time spent in (i) moderate to vigorous activity (e.g., running, gymnastics, and ball games), with response options no activity, <30 minutes, 30–60 minutes, 60–90 minutes, and 90–120 minutes, later dichotomized as <1 hour/week vs. ≥1 hour/week and (ii) light activity (e.g., walking, cycling, and gardening), with response options no activity, <30 minutes, 30–60 minutes, 60–90 minutes, 90–150 minutes, 150–300 minutes, and ≥300 minutes, later dichotomized as <150 minutes/week and ≥150 minutes/week. To capture more information on exercise behavior, participation in municipality-based rehabilitation with group exercise and engagement in leisure time exercise (e.g., structured exercise done independently in the gym) were also reported at presurgery and 3 months following surgery (yes/no). HRQL was assessed on the Trial Outcome Index-Physical/Functional/Breast (TOI-PFB) summary score of the Functional Assessment of Cancer Therapy—Breast Cancer scale (26) ranging from 0 to 96 with higher scores indicating better breast cancer–specific HRQL. Anxiety was measured on the Generalized Anxiety Disorder-7 scale (GAD-7; ref. 27) ranging from 0 to 21 and with cut point of 10 indicating moderate to severe anxiety. Depression was assessed on the Patient Health Questionnaire-9 scale (PHQ-9; ref. 28) ranging from 0 to 27 with cutpoint of 10 indicating moderate to severe depression. Psychological distress was assessed on the DT (17) as described above. Participation in counseling was assessed at presurgery and 3 months following surgery and recorded as having attended counseling sessions (yes/no).

Exercise logs

Electronic exercise logs kept by physiotherapists at the exercise locations provided data on attendance and adherence. Attendance rates were calculated for a dual purpose: (i) raw attendance rates to describe actual exercise attendance and (ii) adjusted attendance rates to describe the rate of attended sessions of possible sessions (excluding cancellations by physiotherapy departments, same-day chemotherapy administration, or colliding hospital appointments). Because the exercise prescription was two times weekly, but three sessions weekly were offered and encouraged, we calculated attendance rates for both two and three possible weekly sessions for transparency and descriptive purposes. Adherence, i.e., the extent to which participants followed the prescribed HIIT duration, HIIT intensity, and PRT load, was calculated only for attended sessions. A cutoff of 75% was considered satisfactory as defined in detail previously (12).

Adverse events

(Serious) adverse events, as defined previously (12), and any complication to peripherally inserted central catheters for infusion of chemotherapy were extracted from the medical records and the testing logs (both study groups) and exercise logs (EX group). Adverse events related to NACT were not systematically collected as these were monitored as part of routine clinical care.

Sample size

A sample size calculation was performed for the primary outcome using a two-part mixture model as described in detail previously (12). A sample size of 90 (45 in each group) was needed to obtain a power of 80% with a 5% significance level to detect (i) an odds ratio (OR) of 1.5 of a tumor size of 0.0 mm and (ii) a 38% reduction in tumor size for those with residual tumor on the MRI before surgery in the EX group compared with the CON group. We initially planned to include 100 participants (allowing 10% dropout), which was adjusted to 120 participants (allowing 25% dropout) after the pilot study (12). Based on the actual dropout rate (4%), assessed after including 98 participants, we followed the original target of 100 participants.

Statistical analysis

All randomized participants were included in the primary analysis in line with the intention-to-treat principle. Participants who dropped out contributed data up to the point of dropout. The primary outcome was reported as median tumor size in mm, change of median tumor size from baseline to presurgery, and the proportion of participants achieving rCR by study group. In addition, a logistic regression model was used to estimate the OR of rCR (primary outcome) and RCB-0 (pCR; secondary outcome) between study groups. The model was adjusted for log-transformed tumor size at baseline and breast cancer subtype according to estrogen receptor (ER) and human epidermal growth factor receptor 2 (HER2) status: ER+ HER2normal, ER+ HER2positive, ER− HER2positive, and ER− HER2normal. Adjustment for breast cancer subtype was used to represent treatment options instead of the randomization stratification factor, NACT regimens, because of the high number of strata with small subgroups. Nonparametric bootstrap was used for the construction of 95% confidence intervals (CI). This choice of analysis differed slightly from the originally planned two-part mixture model as used for the sample size calculation (12). Due to the zero-inflated nature of the data with much higher response rates than expected, only the logistic regression model was used and the planned linear regression of the subgroup with residual tumor was replaced by median tumor size in mm including all participants. A sensitivity analysis, estimating the OR of rCR and RCB-0 (pCR), was performed in which all dropouts were conservatively classified as nonresponders to assess the robustness of the primary findings of treatment response. In addition, per protocol analyses for the primary outcome were performed as described above based on having achieved attendance and adherence ≥75% for two or three weekly exercise sessions, respectively.

Secondary outcomes were compared between study groups at a single assessment time using the Fisher exact test (categorical variables) or nonparametric tests (numerical variables). The McNemar test was used to test for within-group changes over time for dichotomous traits. Repeated measurements of numerical outcomes over multiple assessment points (including baseline) were analyzed using a linear mixed model with fixed effects of group, time, and their interaction. Due to randomization, levels at baseline were constrained to be equal between study groups. For each outcome, an appropriate covariance structure was used to model dependence within subject. Implementations in the R packages “emmeans,” “nlme,” “lmer,” and “lmerTest” were used to extract estimates, 95% CIs, and P values from the models. A significance level of 5% was used for interpreting results. However, due to the large number of tests reported for secondary outcomes, we emphasize the exploratory nature of the study and the risk of false-positive results.

Results

Among 278 patients screened for eligibility between June 23, 2021, and August 31, 2023, 200 eligible patients were informed about possibility to participate. Of these, 102 (51%) patients signed consent and were randomly assigned to the CON group (n = 52) and EX group (n = 50), with even distribution between 10 stratification groups of NACT regimen (A–J; Fig. 1). Data collection for the 3-month follow-up ended on June 24, 2024. Baseline characteristics were well balanced between groups, except for work market affiliation (Table 1). For the EX and CON groups combined compared with the NON group, there were no statistically significant differences in sociodemographics, lifestyle, HRQL, psychologic health, and clinical data at baseline, except for the NON group having more patients on age retirement (P value 0.025) and with lymph node involvement (P value 0.0095; Supplementary Table S1). The representativeness of study participants is further contextualized in Supplementary Table S2. The timing of baseline assessment and randomization was similar between study groups, with most participants tested after the first NACT dose [CON: 88%, median +7 days (interquartile range [IQR], +4 to +9 days); EX: 92%, median +7 days (IQR, +4 to +8 days)]. The timing of presurgery assessments [CON: median −6 days (IQR, −2 to −10 days); EX: median −6 days (IQR, −4 to −10 days)] and 3-month follow-up after surgery [CON: median +97 days (IQR, +90 to +107 days); EX: median +101 days (IQR, +96 to +110 days)] were also considered comparable.

Figure 1.

CONSORT flow diagram of primary outcome. Participants were stratified at randomization by NACT regimen divided into 10 groups (A–J) based on chemotherapy agents and dose intervals. CARB, carboplatin once every 3 weeks on day 1 of paclitaxel; ddEC, dose-dense epirubicin and cyclophosphamide once every 2 weeks; EC, epirubicin and cyclophosphamide once every 3 weeks; HER2, HER2 blockade; PAC, paclitaxel once weekly on days 1, 8, and 15; PBR, pembrolizumab once every 3 weeks on day 1 of paclitaxel.

Table 1.

Clinical and sociodemographic characteristics at baseline by study group in the Neo-train randomized controlled trial.

Characteristic	CON group (n = 52), n (%)	EX group (n = 50), n (%)
Menopausal status	​	​
Premenopausal	18 (35)	23 (46)
Postmenopausal	34 (65)	27 (54)
Histologic subtype	​	​
Ductal carcinoma	48 (92)	48 (96)
Lobular carcinoma	3 (6)	1 (2)
Other	1 (2)	1 (2)
Receptor status	​	​
ER+ HER2normal	10 (19)	6 (12)
ER+ HER2positive	13 (25)	14 (28)
ER− HER2positive	14 (27)	12 (24)
ER− HER2normal (triple negative)	15 (29)	18 (36)
Lymph node metastases	​	​
None	29 (56)	27 (54)
Verified by biopsy	23 (44)	23 (46)
Age at randomization, years (median, IQR)	52.5 (45.0–61.0)	50.5 (42.0–57.7)
≤40	10 (19)	11 (22)
41–50	11 (21)	14 (28)
51–69	26 (50)	20 (40)
≥70	5 (10)	5 (10)
BMI (kg/m2, median, IQR)	26.5 (22.9–30.7)	26.9 (23.6–30.8)
Underweight and normal weight (<25)	19 (38)	19 (36)
Overweight (25–29.9)	17 (34)	17 (33)
Obese (≥30)	14 (28)	16 (31)
Performance status	​	​
0	49 (98)	49 (94)
1	1 (2)	3 (6)
Comorbidity (chronic diseases)	​	​
None	46 (89)	40 (80)
1	6 (11)	8 (16)
≥2	0 (0)	2 (4)
Long-term medications	​	​
None	34 (65)	35 (70)
1–4 different medications	17 (33)	13 (26)
≥5 different medications (polypharmacy)	1 (2)	2 (4)
Smoking history	​	​
Never smoked	27 (52)	26 (52)
Previous smoker	17 (33)	21 (42)
Current smoker	8 (15)	3 (6)
Moderate to vigorous activity in a regular week	​	​
<1 hour/week	35 (67)	34 (68)
≥1 hour/week	17 (33)	16 (32)
Light activity in a regular week	​	​
<150 minute/week	32 (61)	28 (56)
≥150 minute/week	20 (39)	22 (44)
Education levela	​	​
Short	7 (14)	8 (16)
Medium	18 (35)	15 (30)
Long	27 (52)	27 (54)
Work market affiliation	​	​
Working/self-employed	34 (65)	43 (86)
Sick leave/wage-subsidized	5 (10)	1 (2)
Disability pension	3 (6)	2 (4)
Age retirement	8 (15)	4 (8)
Student	1 (2)	0 (0)
Other	1 (2)	0 (0)
Cohabitation status	​	​
Married or living with a partner	37 (71)	38 (76)
Living alone	15 (29)	12 (24)
Had children	45 (87)	44 (88)

^aEducation level: short (mandatory school), medium (secondary high school or vocational education), and long (further education).

Tumor response by MRI

Both study groups had a statistically significant reduction in tumor size from pre-NACT to presurgery: the CON group [baseline median 30.5 mm (95% CI, 28.0–40.0) to presurgery median 0.0 mm (95% CI, 0.0–9.0); change −23.0 mm (95% CI, −39.0 to −22.5)] and the EX group [baseline median 27.0 mm (95% CI, 21.0–36.0) to presurgery median 0.0 mm (95% CI, 0.0–0.0); change −20.0 mm (95% CI, −35.5 to −21.0)], with no between-group differences in EX versus CON [baseline −3.5 mm (95% CI, −14.0 to 5.0); presurgery 0.0 mm (95% CI, −9.0 to 0.0); change −3.0 mm (95% CI, −8.0 to 14.0)]. A total of 56% (95% CI, 42%–69%) in the CON group and 65% (95% CI, 53%–80%) in the EX group had rCR, with no between-group difference adjusted for baseline tumor size and breast cancer subtype [OR 1.16 (95% CI, 0.39–3.91) in EX vs. CON; Fig. 2A; Supplementary Table S3]. The sensitivity analysis, and per protocol analyses based on a 75% level of attendance and adherence, yielded similar conclusions (results not shown). Figure 2B shows variability in the timing and frequency of MRI scans and tumor response for each participant. In both groups, a few participants had a tumor shrinkage pattern of unchanged diameter but exhibited partial response with intratumoral fragmentation.

Figure 2.

A, Median tumor size change in mm with IQR by MRI from before NACT to presurgery and proportion of participants with rCR at MRI 1 to MRI 4, by study group. B, Tumor size change in mm by MRI for each participant in weeks from before NACT to presurgery. Left, CON group; right, EX group. In 16 participants (CON n = 9, EX n = 7), the lesion’s diameter was unchanged, but partial response was observed because of more dispersed islands of tumor cells, high apparent diffusion coefficient levels, or less intense contrast enhancement. C, Residual tumor size in mm from the surgical specimen by study group (Fisher exact test, P value 0.82). D, RCB by study group: RCB-0 (pCR), RCB-I (minimal disease), RCB-II (moderate disease), and RCB-III (extensive disease). E, Manually assessed stromal TIL score in percentage in pre-NACT needle biopsy and post-NACT surgical specimen for each participant. Left, CON group; right, EX group. F, Percentage of participants hospitalized during NACT by study group (Fisher exact test, P value 0.068). G, Total length of hospital stay during NACT in days by study group (Fisher exact test, P value 0.019*). *Statistically significant P value <0.05.

Residual tumor, RCB, and TILs

There was no difference in residual tumor size from the surgical specimen between study groups, with about 60% of participants having no measurable tumor and only 9% in the EX group versus 12% in the CON group having residual tumor over 30 mm (P value 0.82, Fig. 2C). Fifty-seven percent (56/98 with available data from surgery) had RCB-0 (pCR), with 59% (95% CI, 45%–73%) in the EX group versus 56% (95% CI, 42%–69%) in the CON group and no between-group difference when adjusted for baseline tumor size and breast cancer subtype [OR 1.03 (95% CI, 0.43–2.46; Fig. 2D)]. Similar results were found from the sensitivity analysis (results not shown) and for RCB stratified by breast cancer subtype (Supplementary Fig. S1). Twenty-nine percent in total, with a similar proportion per study group, had disagreement between the final MRI result and pathologic evaluation (Supplementary Table S4). Figure 2E shows the manually assessed stromal TIL score for each participant. As presented in Supplementary Table S5, there was no difference between study groups at baseline for stromal TILs in percent by manual assessment (P value 0.73). Among participants with residual tumor available for TIL analysis following NACT (n = 42), there was a statistically significantly higher mean score in the CON group [mean TILs 35.1% (95% CI, 24.3%–46.0%)] compared with the EX group [mean TILs 15.2% (95% CI, 3.8%–26.6%; P value 0.014)]. The computer image analysis showed no difference in TIL density per 10,000 square micrometers by study group at both time points (Supplementary Table S5). Across participants, there was no association between manual TIL score in percent prior to NACT and achieving pCR (Supplementary Fig. S2).

Chemotherapy completion and hospital admissions

Participants in the EX group received a statistically significantly higher proportion of the prescribed chemotherapy (median RDI 94%) compared with the CON group (median RDI 88%; P value 0.048). There was no difference in the number of participants achieving RDI ≥85% by study group. Dose modifications were frequent with statistically significantly fewer participants having dose delays ≥3 days in the EX group (48%) compared with the CON group (69%; P value 0.040), although there were no differences in the proportions experiencing dose reductions or early discontinuation between study groups. The most common reasons for dose modifications were infections/fever of unknown cause, neuropathy, and hematologic toxicities (Table 2). During NACT, there was no difference in the proportion being hospitalized between study groups (P value 0.068; Fig. 2F). The EX group had a statistically significantly shorter total length of hospital stay during NACT compared with the CON group (P value 0.019, Fig. 2G). The most common reason for being hospitalized was infections/fever of unknown cause. Other reasons were dehydration, nausea/vomiting, diarrhea/obstipation, and infected peripherally inserted central catheter with deep vein thrombosis.

Table 2.

Chemotherapy completion, surgery, and adjuvant therapy by study group in the Neo-train randomized controlled trial.

	CON group	EX group	P valueb
	(n = 52)	(n = 46)a
Chemotherapy completion	​	​	​
RDI in % (median, IQR)	88 (79–94)	94 (80–98)	0.048c
≥85% [n (%)]	30 (58)	31 (67)	0.40
<85% [n (%)]	22 (42)	15 (33)	​
Dose reductions [n (%)]	34 (65)	27 (59)	0.54
Dose delays ≥3 days [n (%)]	36 (69)	22 (48)	0.040c
Total dose delays in days [n (%)]	​	​	0.082
0 days	16 (31)	24 (52)	​
3–14 days	25 (48)	17 (37)	​
≥15 days	11 (21)	5 (11)	​
Discontinuation [n (%)]	16 (31)	10 (22)	0.36
Reasons for toxicity-related modificationsd [n (%)]	​	​	​
Infections/fever of unknown cause	39 (31)	19 (25)	​
Neuropathy	29 (23)	19 (25)	​
Hematologic toxicities	20 (16)	20 (26)	​
Fatigue	19 (15)	3 (4)	​
Diarrhea	5 (4)	9 (12)	​
Nausea/vomiting	5 (4)	2 (3)	​
Liver toxicity	3 (2)	0 (0)	​
Kidney toxicity	0 (0)	2 (3)	​
Othere	7 (6)	3 (4)	​
Surgery [n (%)]	​	​	​
Breast-conserving surgery	23 (44)	22 (48)	​
Mastectomy	29 (56)	24 (52)	​
Sentinel node	50 (96)	44 (96)	​
Axillary dissection	12 (23)	7 (15)	​
Radiotherapy [n (%)]	36 (69)	35 (76)	​
Adjuvant chemotherapy [n (%)]	8 (15)	8 (17)	​
Adjuvant endocrine therapy [n (%)]	21 (40)	19 (41)	​

^aFour participants in the EX group withdrew consent during NACT and has missing clinical data.

^b P values estimated using the Wilcoxon rank-sum test for continuous measures and the Fisher exact test for categorical measures.

^cStatistically significant P value ≤0.05.

^dThe main reason for each chemotherapy modification (dose reductions, dose delays ≥3 days, and discontinuation) was counted. Each participant could have more than one dose modification.

^eOther: skin toxicities, edema, headache, and hearing loss.

Cardiorespiratory fitness, muscle strength, body composition, and level of physical activity

Both absolute and relative VO₂max increased statistically significantly from baseline to 3 months following surgery in the EX group, whereas it decreased in the CON group (both P values <0.001 comparing changes between groups; Fig. 3A and Supplementary Table S6 for complete overview of estimates and 95% CIs). For muscle strength, there was a statistically significantly larger increase in the EX group compared with the CON group from baseline to 3 months following surgery (both P values ≤0.001 comparing changes between groups; Fig. 3B and C; Supplementary Table S6). For handgrip strength and measures of body composition, there were no differences between groups at any time point (Fig. 3D; Supplementary Table S6). For self-reported level of physical activity, a statistically significantly higher proportion of participants in the EX group reported ≥1 hour/week of moderate to vigorous activity in week 13 (P value <0.001) and presurgery (P value 0.0014) compared with the CON group, with no differences at 3 months following surgery. For light activity, conversely, a statistically significantly higher proportion of participants in the CON group reported ≥150 minutes/week at presurgery (P value 0.047) compared with the EX group, with no differences at the other time points. In addition, a statistically significantly higher proportion of the CON group reported having engaged in leisure time exercise during NACT (41%) compared with the EX group (16%; P value 0.010), with no difference at 3 months following surgery. For self-reported participation in municipality-based rehabilitation activities with exercise, there were no differences between study groups assessed prior to surgery or 3 months following surgery (Supplementary Table S7).

Figure 3.

Estimated means with 95% CIs from baseline to 3 months following surgery by study group extracted from a linear mixed model. A, Absolute VO₂max in L/minute from a progressive cycle ergometer test. B, Muscle strength in arms in kg from 1RM on pull down. C, Muscle strength in legs in kg from 1RM on leg press. D, Skeletal muscle mass (total body) in kg from bioimpedance analysis. E, HRQL. F, Anxiety. G, Depression. H, Psychological distress.

HRQL, anxiety, depression, and psychological distress

The mean baseline score for HRQL was 64.00 (SD 11.15) and 68.00 (SD 12.80) for the CON and EX groups, respectively, indicating the mid-range level on the scale from 0 (lowest) to 96 (highest) points. Furthermore, baseline mean levels for anxiety, depression, and psychological distress were all below the defined cutpoint for moderate to severe symptoms. Overall, scores remained stable or improved over time with no between-group differences from baseline to 3 months following surgery at any time point (Fig. 3E–H; Supplementary Table S8 for complete overview of estimates and 95% CIs). Analyses dichotomized by symptom severity comparing moderate/severe cases between study groups over time confirmed these findings (results not shown).

Intervention delivery

The median duration of the exercise program was 19 weeks (IQR, 16–22 weeks) with a median of 36 attended exercise sessions (IQR, 21–42 sessions). The adjusted attendance relative to the prescribed two exercise sessions a week was 75% (IQR, 54%–85%), whereas when relative to the encouraged three exercise sessions a week, it was 65% (IQR, 42%–77%). Ten (20%) participants discontinued the exercise program after having attended ≤10 sessions. The median adherence to the HIIT duration, HIIT intensity, and PRT volume was high, reaching 100%, 86%, and 91%, respectively (Table 3). In total, 64% in the EX group experienced moderate to severe psychological distress at least once, all of whom were advised to seek counseling at least once. In the EX group, 49% reported having attended counseling one or more times between baseline and 3 months following surgery (Table 3). In the CON group, 50% reported having experienced moderate to severe distress and despite not having been advised or guided to seek counseling by a project team member, 45% had received this on an alternative initiative (results not shown).

Table 3.

Intervention delivery of the prescribed exercise program and screening for psychological distress in the EX group of the Neo-train randomized controlled trial.

	Total, n	Median (IQR) or n (%)
Exercise program (median, IQR)	50	​
Exercise duration in weeks	​	19 (16–22)
Attended exercise sessions	​	36 (21–42)
Raw attendance in %a	​	​
3 sessions/week	​	58 (36–69)
Adjusted attendance in %b	​	​
2 sessions/week	​	75 (54–85)
3 sessions/week	​	65 (42–77)
Adherence in %	​	​
HIIT durationc	​	100 (98–100)
HIIT intensityd	​	86 (66–94)
PRTe	​	91 (67–98)
Screening for psychological distress [n (%)]	50	​
Scored moderate to severe psychological distress at least oncef	​	32 (64)
Advised to seek counseling by the project team member at least onceg	​	32 (64)
Attended counseling at least onceh	​	22 (49)
Type of counseling attendedh	22	​
Danish Cancer Society counseling centersi	​	7 (32)
Other counseling, e.g., private practice psychologist	​	9 (41)
Both of the above	​	6 (27)

^aNumber of attended sessions divided by the number of prescribed sessions.

^bNumber of attended sessions divided by the number of possible sessions (i.e., excluding canceled sessions by the physiotherapy departments or same-day treatment/examination at the hospital).

^cCompleted minimum 20 minutes of cycling per session.

^dCompleted four HIIT intervals with ≥16 on the ratings of perceived exertion (Borg) scale and ≥85% maximum heart rate per session.

^eCompleted three sets of 12–15 repetitions on pull down, chest press, and leg press, respectively, at ≥65% of baseline 1RM per session.

^fMeasured as the score of 7–10 on the DT at baseline, week 13, presurgery, and/or 3 months following surgery.

^gAt baseline (after randomization), week 13, presurgery, and/or 3 months following surgery.

^hSelf-reported by questionnaire at presurgery and/or 3 months following surgery. Forty-five participants responded.

ⁱFree, drop-in counseling available at nine locations across Region Zealand, Denmark.

Adverse events

No serious adverse events were observed. Adverse events were reported in three participants from the EX group: vasovagal syncope during HIIT (n = 1) and back strain after leg press (n = 2), for which reason one participant exited the supervised sessions but continued in the study by using the activity tracker. Most participants had peripherally inserted central catheters (74%) with only few complications observed in both study groups: thromboembolic events (CON n = 2; EX n = 3) and infections requiring antibiotics (CON n = 3; EX n = 3).

Discussion

To the best of our knowledge, this was the first published randomized controlled trial to investigate if supervised HIIT and PRT during NACT could improve tumor response by MRI in patients with breast cancer as a primary outcome. Contrary to our hypothesis, we observed no difference in tumor reduction between study groups. MRI is regarded as the most accurate breast imaging technique for monitoring tumor response during NACT (29). Nevertheless, response evaluation by MRI is challenging because of diverse tumor shrinking patterns influenced by complex factors such as necrosis, fibrous tissue, and molecular subtypes. The accuracy of MRI to detect pCR showed the highest pooled sensitivity in ER− HER2normal (triple negative) and ER− HER2positive subtypes and the lowest in ER+ HER2normal in a recent meta-analysis based on 26 studies with 4,497 individuals (30). In nearly a third of our sample (29%), equally distributed by study group, there was disagreement between presurgery MRI and pathologic evaluation, highlighting this challenge. An alternative primary outcome could have been pCR, as it may provide a more accurate measure of treatment success and residual disease. This is currently being investigated in the ongoing Neo-ACT trial with home-based exercise supported by a mobile application during NACT in patients with breast cancer compared with usual care (31). In our trial, 57% of participants who received NACT between 2021 and 2023 achieved pCR. This rate is notably higher than the 35% pCR rate we reported in our recent study on NACT completion rates and tumor response in usual care from 2017 to 2019 (32). Similar findings were reported in an international cohort of 5,161 patients receiving NACT for breast cancer between 1994 and 2019 (pCR 33%; ref. 11). Treatment regimens for NACT for early breast cancer are rapidly evolving, and response rates are improving. With improvements in the primary treatment, the room for improvement for an additional effect by exercise becomes narrower. This may also support a more targeted approach to testing exercise interventions as also suggested by Courneya and Booth (33)—for example, among patients with ER+ breast cancer, who typically have lower pCR rates and may respond differently to adjuvant exercise interventions (34). In our trial, both study groups achieved a median tumor size of 0.0 mm prior to surgery, and it is unknown whether an effect of the exercise program on tumor reduction, if present, was too small to be detected. Secondary analyses from the recent LEANer trial found a significantly higher pCR rate in women randomized to a home-based exercise and nutrition program during NACT, compared with usual care (53% vs. 28%, P value 0.037; ref. 35). Notably, the control group’s pCR rate (28%) was much lower than that reported in our trial, and in a subgroup analysis (n = 72), such a low response rate could be due to random chance.

Following NACT, we observed a higher mean concentration of stromal TILs in the CON group compared with the EX group, which contrasts to our hypothesis that exercise would stimulate immune cells (3, 6), leading to a higher concentration of TILs in the EX group. However, this finding was derived from a subgroup analysis of participants with residual tumor (n = 42) and did not align consistently with our findings from the automated image analysis of TIL density. We found no association between stromal TIL score in biopsy and achieving pCR, most likely due to the limited sample size, which again contrasts with a much larger pooled analysis of 3,771 patients treated with NACT, in which increased TIL concentration was found to predict response to NACT across all molecular subtypes (36).

Of clinical importance, we found significantly improved chemotherapy completion in the EX group, with a higher median RDI and fewer dose delays compared with the CON group. This aligns with findings from two trials on supervised exercise during adjuvant chemotherapy in patients with breast cancer (37) but contrasts with two more recent trials showing no difference in RDI between exercise and usual care in similar patients (35, 38). A substantial proportion in both our study groups (CON 42% and EX 33%) had an RDI <85%, which is higher than that in previous studies (32, 35). All participants in our trial received long NACT regimens (18–24 weeks) currently offered, and 60% were prescribed dose-dense regimens, which may have increased the need for toxicity-related dose modifications compared with earlier studies. Interestingly, we found significantly more participants in the EX group having no hospital stay and fewer having a short total length of stay of 1 to 7 days compared with the CON group. This aligns with a recent meta-analysis, primarily on patients with hematologic cancers but including two trials on breast cancer, showing that structured exercise during chemotherapy, radiotherapy, or stem cell transplant reduced the length of hospital stay and risk of hospital admissions compared with usual care, although based on low-quality evidence (39). The findings of improved chemotherapy completion and reduced total hospital stay during NACT in our trial suggest that exercise may enhance treatment tolerance, allowing patients to endure chemotherapy more effectively while minimizing hospital time. Our data are, however, too limited to draw definite conclusions.

Our exercise program achieved median adjusted attendance rates of 65% to 75% for two to three weekly sessions, with high adherence to both HIIT and PRT, which is comparable with those reported in two recent trials of supervised exercise during 16 to 18 weeks of chemotherapy in patients with breast cancer (38, 40). Nevertheless, only 25% (n = 13) had attendance and adherence ≥75% based on three weekly sessions resulting in three of four participants not completing the amount of exercise that we originally designed in the trial. Toxicity and competing obligations were the main reasons for not attending sessions. This highlights the difficulty of finding an exercise prescription that patients are able to attend and adhere to during intensive and time-consuming treatments. With a rural region as site for recruitment, the flow of participants during the 2-year period was limited and the need to provide exercise locations close to home caused difficulties in offering group-based sessions to all participants. This may have affected attendance negatively. We identified few adverse events consistent with previous high-intensity exercise programs during cancer treatment (41). A highly relevant finding for clinical practice was that we observed no increased complications related to having peripherally inserted central catheters while performing heavy-load resistance training.

Our exercise program led to increased estimated VO₂max levels and muscle strength, which aligns with findings from previous trials during cancer treatment (41–43). Baseline levels of relative VO₂max (mean 22.0 mL/minute/kg) in our trial were comparable with those in women with breast cancer following adjuvant chemotherapy and lower than those in healthy, sedentary women (mean 29.7 mL/minute/kg) across 27 trials and observational studies (44). Our study population was somewhat inactive as only a third of participants reported engaging in at least 1 hour of moderate to vigorous activity in a regular week at baseline. In the EX group, the proportion of time spent on moderate to vigorous activity increased significantly, which together with increased cardiorespiratory fitness and muscle strength supports the effect of the exercise intervention. Self-reported participation in group-based exercise rehabilitation in the municipality and leisure time exercise revealed that 34% and 41% in the CON group had engaged in these activities during NACT, respectively. The lack of improvement in cardiorespiratory fitness and muscle strength in the CON group indicated, however, that these activities were likely to be of lower intensity or shorter duration than our exercise program. As level of physical activity was self-reported, these results may be subject to bias. We found minimal changes in body composition from baseline to 3 months following surgery, in accordance with a recent study in patients with breast cancer (45), and with no consistent differences between study groups. A recent systematic review identified only few studies measuring the effects of exercise on body composition during NACT and using diverse measurement methods, which challenges comparison (43).

Surprisingly, we found no beneficial effects of the exercise intervention on any of the HRQL or psychologic outcomes. This contrasts with a meta-analysis on 34 trials showing that exercise, and supervised sessions in particular, improved HRQL in patients with cancer (46) and other studies demonstrating exercise to reduce levels of anxiety and depression (47). Notably, our small sample size and generally low scores in both study groups throughout the study indicate that participants, regardless of group allocation, had modest symptoms, contrary to what could be expected in patients who receive chemotherapy. Subgroup analysis stratifying by high symptom burden at baseline could have been useful in exploring differences in responses to the intervention. A similar proportion of participants in both study groups (CON 45% and EX 49%) reported having attended at least one counseling session between baseline and 3 months following surgery. Given that 64% of the EX group was given the extra advice to seek counseling, one might expect higher attendance compared with the CON group, especially with free services available in close proximity to the hospital through the Danish Cancer Society. Some participants may have felt too overwhelmed to seek counseling, or the available options may not have aligned with their needs and expectations.

When developing supportive strategies that might affect treatment outcomes, it is crucial to ensure that these do not increase inequalities in cancer care. This is particularly important for lifestyle interventions such as engaging in exercise, which depend on individual motivation influenced by socioeconomic factors, social support, exercise habits, and general health (48, 49). Designing interventions that appeal broadly may help ensure benefits reach diverse populations and promote equitable treatment outcomes. Through patient involvement and pilot testing, our goal was to create an exercise intervention also appealing to individuals who might be more reluctant to participate, such as patients of older age, having short education, or who were not regularly active before their cancer diagnosis. Our baseline data showed enrollment of participants with overweight, obesity, low self-reported physical activity, and low cardiorespiratory fitness levels. Our data also indicated that nonparticipants were comparable with trial participants, suggesting that our study population was representative of the broader cohort of patients with breast cancer receiving NACT at a university hospital covering a region in Denmark and eligible for an exercise trial.

Limitations of our trial include the small sample size of 102 participants from a single site, the heterogeneous study population with different breast cancer subtypes, treatment regimens, and physical fitness levels and that our study was not powered to detect differences in the many secondary outcomes, chosen to provide a broad insight into the effects of the Neo-train program, but which should be interpreted cautiously. To allow recruitment without any effect on treatment start date, baseline assessments were allowed both before and after the initiation of NACT. Most participants were tested after receiving the first dose, with a balanced distribution between study groups. However, this timing may have affected the measures of physical and psychosocial health. Lastly, a limitation of the cycle ergometer test was that tests terminated because of discomfort or pain may not reflect true maximal effort, potentially affecting the accuracy of the estimated VO₂max. Important strengths are the blinded assessment of radiological and pathologic assessments of tumor response, measurement of several clinical, physical, and patient-reported outcomes in combination with biological data, and transparent documentation of the exercise program’s design, attendance, and adherence.

In conclusion, among women with breast cancer, a majority being overweight or having obesity and low cardiorespiratory fitness levels, a supervised high-intensity exercise program during the entire duration of NACT (18–24 weeks) did not enhance tumor reduction by MRI, nor did it improve pCR as compared with usual care. It is unclear whether this is due to the exercise prescription itself, low attendance, the diverse participant group, varying NACT regimens, efficient effect of chemotherapy constituting a floor effect, insufficient power to detect differences, or a true lack of effect. The findings of fewer chemotherapy dose delays, higher RDI, and shorter total length of hospital stay during NACT in the EX group, although secondary outcomes, are highly clinically relevant results pointing toward possible improved treatment tolerance.

Supplementary Material

Supplementary Material S1

CONSORT checklist

Supplementary Material S2

Methods on tumour infiltrating lymphocytes

Supplementary Figure S1

Residual Cancer Burden by breast cancer subtype

Supplementary Figure S2

Tumour infiltrating lymphocytes and Residual Cancer Burden

Supplementary Table S1

Baseline characteristics in participants vs non-participants

Supplementary Table S2

Representativeness of study participants

Supplementary Table S3

Tumour size and radiologic complete response by magnetic resonance imaging

Supplementary Table S4

Radiologic complete response vs pathologic complete response

Supplementary Table S5

Analyses of tumour infiltrating lymphocytes

Supplementary Table S6

Cardiorespiratory fitness, muscle strength and body composition

Supplementary Table S7

Level of physical activity and participation in municipality-based rehabilitation

Supplementary Table S8

Health-related quality of life, anxiety, depression and psychological distress

Data Availability

Due to patient privacy restrictions, the raw data generated in this study are only available for collaborative research upon reasonable request to the corresponding author, subject to approval by the relevant ethics committee.

Authors’ Disclosures

E. Kjeldsted reports grants from Helsefonden, Dansk Kræftforskningsfond, Dagmar Marshall Foundation, Aase and Ejnar Danielsens Foundation, and Fabrikant Einar Willumsens Mindelegat during the conduct of the study. A.-V. Lænkholm reports personal fees from AstraZeneca, MSD, and Daiichi Sankyo; other support from Daiichi Sankyo and AstraZeneca; and grants from MSD and AstraZeneca outside the submitted work. J. Gehl reports grants from Helsefonden, Kræftens Bekæmpelse, Dagmar Marshall Foundation, Aase and Ejnar Danielsens Foundation, and Fabrikant Einar Willumsens Mindelegat during the conduct of the study. No disclosures were reported by the other authors.

Authors’ Contributions

E. Kjeldsted: Conceptualization, resources, funding acquisition, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. G. Ammitzbøll: Conceptualization, resources, investigation, methodology, project administration, writing–review and editing. A.-V. Lænkholm: Resources, investigation, methodology, writing–review and editing. D. Rasic: Resources, investigation, methodology, writing–review and editing. S.G. Ceballos: Resources, investigation, methodology, writing–review and editing. L.B. Jørgensen: Methodology, writing–review and editing. S.T. Skou: Methodology, writing–review and editing. R.D. Bojesen: Methodology, writing–review and editing. A. Lodin: Resources, investigation, methodology, writing–review and editing. A. Tolver: Formal analysis, visualization, methodology, writing–review and editing. S. Rosthøj: Methodology, writing–review and editing. S. Jack: Methodology, writing–review and editing. J. Gehl: Conceptualization, resources, supervision, investigation, visualization, methodology, project administration, writing–review and editing. S.O. Dalton: Conceptualization, supervision, funding acquisition, methodology, project administration, writing–review and editing.