One hundred sixteen older adults (13/103 M/F, 70.5 ± 4.6 years, 165.3 ± 6.2 cm, and 66.6 ± 12.6 kg) volunteered for this study. Sixty-five of the participants were healthy older adults, whereas fifty-one of the participants were osteoporosis patients. Osteoporosis patients were included if they had verified osteoporosis (− 2.5 ≥ T-score ≥  − 3.5) and were in medical treatment for osteoporosis, whereas healthy participants were included if they had a T-score >  − 2.5 and did not receive medical treatment for osteoporosis. All participants were excluded if they were younger than 65 years old, had previous neurological, musculoskeletal (besides osteoporosis) or mental illnesses, or participated in medical trials.

All participants were given a detailed verbal explanation of the experiment and provided written informed consent. The study was conducted in accordance with the Helsinki Declaration, approved by the local ethics committee (N-20180065). This study is a part of a clinical trial [26], investigating the effect of dance as a fall-preventing exercise program. The clinical trial was designed to have 80% power to detect at least a 10% reduction in falls per person-years.

All participants had a single test session, lasting approximately three hours. Upon arrival, anthropometric data and history of falls and fall-related fractures (“no fall,” “fall – no fracture,” and “fall – fracture”) were acquired. Bone mineral density (BMD) for lumbar spine (L1-L4), femoral neck, and total hip, as well as body composition, were then examined using a dual-energy X-ray absorptiometry (DXA) scanner. Pain was assessed using a 10-cm visual analogue scale (VAS). Hereafter, postural sway was quantified on a force plate, during quiet stance in four conditions, eyes open and eyes closed on firm surface, eyes open, and eyes closed on a foam surface. Participants had their gait tested during a 25-m walking test, with and without a cognitive challenging dual task. Lastly, the participants functional fitness and postural stability were quantified using the 30-s chair stand test, 2-min step test, 8-Foot Up-and-Go test, and the Mini-BESTest.

Bone mineral density was measured to secure that all participants followed the inclusion and exclusion criteria. The participants were scanned using a Horizon A (Horizon A, Hologic, Marlborough, MA, USA) DXA scanner. The BMD was measured in g/cm2 and computed in a T-score. The radiation from the scanner was less than 100 microSv. Body composition was measured in grams and converted to a bodyfat percentage.

Pain intensity was assessed using a 10-cm visual analogue scale (VAS), graded for every cm, with the description “no pain,” “moderate pain,” and “worst imaginable pain” at 0, 5, and 10 cm, respectively. The pain intensity was measured with one decimal precision. When assessing pain, participants were asked to rate the average rating of pain they felt in their everyday life. If participants had a VAS score higher than 0, they were asked to draw the painful sites on a body chart. The drawings on the body chart were converted into a percentage measure, describing how large the painful area was, compared to the whole body (pain area%).

To quantify the static balance, participant stood on a force plate (Plux Biosignals S.A., Arruda don Vinhos, Portugal) during four conditions: (i) eyes open on a firm surface, (ii) eyes closed on a firm surface, (iii) eyes open on a soft foam surface, and (iv) eyes close on a soft foam surface. Participants were asked to stand in a normal stance, feet with a shoulder width apart and arms resting at their side. In the soft surface condition, a 48 × 40 × 6 cm Airex® Balance-pad (Airex, Sins, Switzerland) were placed on the force plate. All conditions were applied three times in a randomized order, each with a 30-s duration. Ground reaction forces were recorded at 1 kHZ (Opensignals, Plux Biosignals S.A, Arruda dos Vinhos, Portugal) and filtered using a second order Butterworth filter (15 Hz low pass frequency). Center of pressure (CoP) sway velocity and sway area was calculated, with area extracted via principal component analysis and 95% confidence interval for ellipse calculation [27]. A mean of all 12 tests (four conditions with three trials of each) was calculated for velocity and area of the sway, returning a single measure for sway velocity and for sway area. Dynamic balance was assessed using the Mini BESTest [28], which is a test-battery consisting of 14 tests, each rated from 0 to 2, with a total test score ranging from 0 to 28.

To quantify step length, step time, and gait velocity, participants walked at a self-selected pace in a 25-m path, wearing 17 wireless inertial sensors (Awinda, Xsens Technologies B.V., Enschede, The Netherlands), sampling at 1000 Hz. Through the recordings from the sensors, length of gait cycle (meter), time (milliseconds), and velocity (meter/second) were quantified, by extracting knee joint movement.

To quantify cognitive function, a mathematical dual-task task was administered [29]. Participants had to walk on the same 25-m path as in the single-task gait test, while performing continuously subtractions of seven from a number between 200 and 500. The subtractions should be recited aloud and was then noted.

Both the single-task and dual-task gait test were repeated three times, in a randomized order. An average of the three tests were used for further analysis.

To assess fitness of the older adults, three physical fitness tests were used, the 30-Second Chair Stand, 2-Minute Step-in-Place, and the 8-Foot Up-and-Go test, all from Rikli and Jones Fullerton Battery test for older adults [30]. The 30-Second Chair stand tests the leg strength and endurance and is measured as the number of times the participant can raise from the chair, without using the arms to push off. The 2-Minute Step-in-Place test is used to test aerobic endurance among older adults and is measured by the number of knees raises with the right knee. The 8-Foot Up-and-Go test measures speed, agility, and balance while moving, and is scored by the nearest 1/10th second.

Independent samples T-tests were performed between the three demographic set of data (age, height, and weight), regarding fall accidents, to secure that a possible difference between groups was not due to differences within demographic data. A 2 × 2 contingency table with a Pearson chi-square test was performed between the variables Osteoporosis and falls. A Pearson chi-square test was calculated for the variables Osteoporosis and fall-related fractures (“no fall,” “fall – no fracture,” and “fall – fracture”). For all of the variables, a MANOVA with a Bonferroni-corrected pairwise comparisons test used to analyze if there were differences within any of the 14 dependent variables (VAS, CoP sway area, CoP sway velocity, walking speed with and without dual-task, gait cycle length with and without dual-task, gait cycle time with and without dual-task, mini-BESTest, 30-Second Chair Stand, 2-Minute Step-in-Place, 8-Foot Up-and-Go, body fat percentage) and the fixed factors (previous falls and osteoporosis), as well as a MANOVA with a Tukey post hoc test for the 14 dependent variables and the fixed factors (fall-related fractures and osteoporosis). For each variable in the MANOVA tests, the observed power was calculated. For the Bonferroni-corrected pairwise comparisons, the partial ETA squared was calculated. A Walds forward stepwise logistic regression test was used to calculate if there were any statistical association between the groups of fallers and non-fallers, within each of the 14 parameters. If pain levels differed significantly in the MANOVA or were included in the logistic regression model, an independent T-test were calculated for the body chart drawings between fallers and non-fallers. Statistical significance is set to P < 0.05.