Participants
All volunteers had to be healthy with no orthopaedic or cardiovascular conditions and should be able to perform a squat with the tops of their femur parallel to the floor. Therefore, seven individuals were excluded before the start of the study. After being informed about the study procedures and inclusion criteria, 41 healthy women were enrolled on the study (15/03/2021) after signing informed consent. The classification of the participants in pre-and postmenopausal was based on hormone concentrations and the date of the last menstrual period [37]. In the case of low estradiol (E2) and high follicle-stimulating hormone (FSH) concentrations, and at least 12 months without menstruation, participants were declared postmenopausal [37]. Seventeen participants (n = 17) were classified as pre- (PreMeno) and 24 as postmenopausal (PostMeno). Subsequently, postmenopausal women were assigned after stratified randomization (MM, age, weight, and height) into two subgroups: moderate intensity (MI-RT; n = 12) and low intensity (LI-RT; n = 12). Premenopausal women were not divided into subgroups due to the small sample size and performed MI-RT.
Experimental design
The study was approved by the local ethics committee of the IST University of Applied Science, Dusseldorf (02/2021)
, according to the Declaration of Helsinki, and registered in the German Registry of Clinical Studies (05/03/2021; DRKS00023826). In addition, the hygiene concept for preventing the spread of COVID-19 was approved by the local regulatory authority. A total of 41 women aged 40–60 years were recruited into a local gym. The study design encompassed two phases, and three measurement points (T0, T1, T2), and lasted for 20 weeks (see Fig. 1).
#####Figure1#####
Initially, ten weeks (T0-T1) served as a control period without any training or systematic physical activity (lockdown due to the COVID-19 pandemic). This was followed by a 10-week RT intervention period (T1-T2) using a two-group matched pair parallel design. At T0, participants completed questionnaires about their health, menstrual status, and recent RT history. Subsequently, hormone status (estradiol [E2], progesterone [P], follicle-stimulating hormone [FSH], testosterone [T], dehydroepiandrosterone [DHEA]), body composition (total body water [TBW], fat-free mass [FFM], muscle mass [MM], fat mass [FM]), muscle thickness (vastus lateralis [VL], rectus femoris [RF], triceps brachii [TB]), grip strength (GS), and dynamic strength (1-RM squat [SQ], 1-RM bench press [BP]) were assessed. The same variables were also collected after the control and RT period except for hormone status. T2 testing was done 48–72h after the last RT session.
Procedures
Forty-eight hours before testing, the subjects were not allowed to engage in resistance training or any other strenuous physical activity. The volunteers were also not allowed to drink alcohol before testing and had to appear fast. All measurements (T0–T2) were conducted single-blind in the morning at the same time of the day by the same researcher.
Hormone parameters
For saliva samples, specific ELISA kits for E2, P, T, and DHEA concentrations were used (RE52281; RE62141; RE52651; RE30121046; Re62039). Dry blood concentration of follicle-stimulating hormone (FSH) was analyzed by an external laboratory according to the CLIA method (Ayumetrix, 17387 63rd Ave, Lake Oswego, OR 97035, USA).
Body mass and body composition
Body mass [BM] was assessed by a digital scale (Etekcity EB4074C, Anaheim, CA, United States of America), with participants only wearing underwear and no shoes or socks. Total body water (TBW), fat-free mass [FFM], MM, and fat mass (FM) were analyzed by bioelectrical impedance analysis (BIA 101, Akern, Firenze, Italy). Before measurements, participants had to lie supine for ten minutes on a treatment table to allow for fluid shift [38]. Subsequently, the measurement was performed, and the data were processed using BodyGramPro software (Version 3.0, Akern, Firenze, Italy).
Muscle thickness
For muscle thickness measurements, a B-mode ultrasound (Mindray DP-50, Mindray Medical International Ltd, Shenzhen, China) with an 8.5-MHz linear probe (Mindray 75L53EA, Mindray Medical International Ltd, Shenzhen, China) was used. Muscle thickness was measured at three anatomical sites on the right side in accordance with previous studies [38, 39]. M. vastus lateralis [VL] thickness was measured with the participants lying supine on an examination table at half distance between the most prominent point of the greater trochanter and the lateral condyle of the tibia (gain 50 dB; image depth 3.7 cm). The thickness of the rectus femoris [RF] was measured at 50 % betwen the anterior inferior supra iliac crest and the proximal border of the patella with participants lying on their left side (gain 50 dB; image depth 3.7 cm). For measurement of the triceps brachii [TB], the participants lay on their abdomen while mages were taken at 40 % distl between the acromial process of the scapula and the lateral epicondyle of the humerus (gain 50 dB; image depth 5.5 cm). To ensure the identical positioning of the ultrasound probe, the measuring points were marked with a waterproof pen. Ultrasound transmission gel was applied to the scanning head, and the probe was positioned perpendicular to the long axis of the extremity without depression of the underlying tissue. Three images were recorded at each site and saved on a USB flash drive. Subsequently, muscle thickness was analyzed in the images using the calliper measurement of the ultrasound machine. The mean values of the three images of each site were used for further analyses. The test–retest intraclass correlations coefficient for this analysis was reported from our laboratory as 0.998 (RF), 0.996 (VL), and 0.997 (TB) [38].
Strength
Following a standardized warm-up procedure (5min running), grip, upper and lower body strength tests were conducted. Grip strength [GS] of both hands was assessed using a digital hand-held dynamometer (digital Jamar+, Fabrication Enterprises, New York, United States). For testing, volunteers were seated upright on a chair with elbows bent 90° and in contact with the body. Then they were instructed to press the grip of the device as forcefully as possible for at least five seconds without changing their position. Three trials were carried out for each side. The best trial was documented for further analysis. For testing lower body strength, a “touch and go” barbell box squat (SQ, femur parallel to the floor) was used. The height of the box was adjusted for each subject individually and maintained throughout the RT period and during retesting. After a minimum of five minutes of rest, upper body strength was assessed using the free weight bench press exercise (BP). For this, grip widths were documented and stipulated throughout the study. For both tests, the participants first completed ten repetitions with an empty bar, followed by a two-minute rest period. Then, a second warm-up set was completed with ten repetitions and approximately 50% of the predicted ten-repetition maximum load. Following four minutes of rest, a final set was performed until momentary concentric muscle failure or failure of proper exercise technique. To this end, a load was set in advance, which the research team estimated to allow one to ten repetitions. From the load used and the completed repetitions, the 1RM squat and bench press were calculated according to the formula proposed by Brzycki [40], which was evaluated as sufficiently accurate for estimating 1-RM using fatiguing sets of less than ten repetitions [41].
Resistance Training Protocol
A detailed description of the RT protocol (sets, repetitions, intensity, tempo, and rest) throughout the study can be found in Table 1. The training program consisted encompassed two cycles of five weeks each. Weeks one to four of each cycle were “loading” weeks followed by one “unloading” week. The training was conducted twice weekly, 48-72h apart. All RT sessions were supervised by a qualified member of the research team (researcher-to-participant ratio 1:1–4). Exercise selection was identical for all intervention groups. Session 1 consisted of “touch and go” barbell box squats (femur parallel to the floor, barbell bench press, seated neutral grip cable row, dumbbell side bend, and prone plank. In the second training session, the same exercises were completed, except that the cable row was replaced by a lat pull-down with a wide pronated grip. Except for box squats and the plank, all exercises were completed with the maximum possible range of motion and at identical tempi. Volume loads (repetitions x sets x % 1-RM) were approximately equal between intervention groups, with the MI-RT group performing more sets per exercise to achieve a similar volume load compared to the LI-RT. The adjusted weight of the box squat and bench press in the first cycle was based on the initial 1-RM test (T1), whereas the resistance of the remaining exercises was determined by trial and error. During the “loading” weeks, the last set of each exercise was performed to momentary concentric failure or failure of proper exercise technique. During the following weeks, the weight was increased by 2.5-5% for each exercise, depending on the repetition to failure. From the last set of each exercise in week 4, a new 1-RM was estimated using the Brzycki formula [40] and used from week six.
Table 1
Overview of the resistance training protocol
| Cycle 1 | Cycle 2 |
Week1–4 Loading | Week5 Unloading | Week6–9 Loading | Week10 Unloading |
Session 1 and 2 (sets x reps) |
MI-RT Intensity | 4sets 3x10 1x to failure @75% 1-RM1) | 3sets 3x10 @53.3% 1-RM1) | 4sets 3x10 1x to failure 75% 1-RM²) | 3sets 3x10 53.3% 1-RM²) |
LI-RT Intensity | 3sets 2x20 1x to failure 50% 1-RM1) | 2sets 2x20 40% 1-RM1) | 3sets 2x20 1x to failure 50% 1-RM²) | 2sets 2x20 40% 1-RM²) |
Tempo Rest | 2:0:1 (eccentric : isometric : concentric) 120seconds |
MI-RT = moderate-intensity resistance training group; LI-RT = low-intensity resistance training; 1-RM = one-repetition maximum; 1) based on pre intervention 1-RM; ² ) based on estimated 1-RM using the Brzycki formula (35) (weight and repetitions from the last set of each exercise following week 4 session 2).
Nutrition
Dietary habits were maintained throughout the 20 weeks. Only immediately after the RT sessions participants consumed a carbohydrate-protein source to maximize muscle protein synthesis. Participants could choose between three different meals with a similar amount of calories, protein and carbohydrates. Each meal was used in previous investigations [42, 43]. For detailed meal information, see the supplemental materials.
Statistical analyses
Data were analyzed by the R statistical language version 4.0.4 (R Core Team, 2021). Only data from participants with an adherence of > 85% were included in the analyses. The raw data of VL, RF, and TB were box-cox transformed before analysis. The transformation was applied to obtain approximately Gaussian distributions of the raw data which exhibited heavily skewed distributions otherwise. Transformations were applied to the data at T0 and the box-cox estimates for λ were then employed to transform the remaining data.
The values of FFM, MM, FM, GS, 1-RM SQ and BP were analyzed unchanged. Individual time intervals (∆t [h]) since the start of the study were introduced as an additional covariate. Data analysis was performed using linear mixed effects (LME) models, where FFMλ, MMλ, FMλ, VLλ, RFλ, TBλ, GS, 1RM SQ and BP, served as dependent variables. Model building was performed independently for each of these.
We were specifically interested in the effects of menopause on the trends of strength capacity and muscle growth. Therefore, all models included the interaction term of menopause with PreMeno MI-RT (T0 to T2) as a fixed effect. Likewise, PreMeno MI-RT itself was axiomatically included as a fixed effect. As it represents an ordered factor with three levels, second-order orthogonal polynomials were chosen as contrasts for PreMeno MI-RT.
Initially, random effects were merely assumed between individual intercepts of each measure. Subsequently, ∆t was included as a random effect, where linear individual trends were assumed. The presence of potentially non-linear individual trends then was investigated by upgrading to 2nd or 3rd-order natural splines of ∆t. After the development of appropriate random effect structures, it was tested whether ∆t also contributed to general trends in the population, i.e., whether it represented a significant fixed effect. In either case, model comparisons were based on likelihood statistics and changes in Akaike’s information criterion. Significant differences were set with p ≤ .05.
Finally, effect sizes were determined between discrete time levels according to the approximation of Cohen's d for mixed effects models (d = 2t/DF(1/2) where t = t-value; DF = degrees of freedom. Classifications were stipulated as follows: trivial < 0.2; small < 0.5; moderate < 0.8; strong > 0.8 [44]. All graphs were created using the latest version of GraphPad Prism.