Effects of Variations in Resistance Training Frequency on Strength Development in Well-Trained Populations and Implications for In-Season Athlete Training: A Systematic Review and Meta-analysis

This systematic review design was developed in adherence to the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA). The PRISMA guidelines include 27 items within a checklist that is designed to be used as the basis for reporting systematic reviews [24]. The research question being investigated within this review as defined by the PICO model (population, intervention, comparison and outcome) is whether in a well-trained population, during volume matched resistance-training interventions, does the frequency of training have an effect on both lower and upper body strength outcomes during experimental randomized and non-randomized studies? A protocol was not pre-registered for this review.

A Boolean/phrase search mode was applied using the following keywords: “resistance training” AND “frequency” AND “volume” AND “intensity”. The keywords were inputted using this format into four different databases, including PubMed, SPORTDiscus, Ovid and Scopus. Filters were applied to all databases to include studies that were written in the English language and presented in peer-reviewed academic journal articles. No restrictions were placed upon the sex of subjects; age, however, was restricted to no greater than 35 years, with no lower age cut-off.

The primary focus of this literature search was to identify studies that have assessed the effect of different RT frequencies in trained/athletic populations. The search timeframe was restricted up to and including 1^st April 2020, with no earliest date restriction, and following this search there were 2134 articles identified for further inspection. All duplicated studies were removed initially with the remaining studies then being screened utilizing the subsequent criteria. Research articles were included and eligible within this review provided that (1) a measure of maximal strength was assessed, (2) a minimum of two different training frequency groups were included, (3) the populations of the studies were stated as well trained, and finally (4) multi-joint exercises were included within the training programmes. Studies were excluded for using subjects that were not injury free for the 6 months prior, along with any systematic or narrative reviews. A summary of the above selection process is outlined in Fig. 1. Means and standard deviation (SD) were required from all papers to be analysed further; if these values were not present but the study met the rest of the criteria, the corresponding authors were contacted in order to obtain these values.

Following the identification of the studies included within this review, the quality and risk of bias were assessed. This included a Cochrane risk of bias assessment tool to assess the risk of bias within the randomized controlled trials. The Cochrane risk of bias assessment tool evaluates randomized controlled trials based on several categories that include sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective outcome reporting, and ‘other issues.’ Grades for these categories were provided as either ‘high risk of bias’, ‘low risk of bias’, or ‘unclear risk of bias’.

Means and SDs of upper and lower body maximal strength measures were independently extracted from the included studies for further analysis. Maximal strength tests included 1RM back squat, leg press, and bench press, and maximum voluntary isometric contraction of the knee and elbow flexors. Hedge’s g effect sizes (g) were calculated from the pre- to post-intervention results of each study to provide a standardized value whereby the magnitude of differences can be determined and compared across interventions whilst accounting for differences in sample size. The scale for interpretation of g was proposed by Hopkins [25] as follows: trivial (≤ 0.20), small (0.21–0.59), moderate (0.60–1.19), large (1.20–1.99), or very large (≥ 2.00). An estimation for between-study variance was calculated using a random-effects model, with associated Z value, p value and 95% confidence intervals (95% CI), absolute heterogeneity was assessed using Tau² and this was estimated using the restricted maximum likelihood method. Finally, a test for relative heterogeneity (I²), as outlined by Higgins et al. [26], was used to quantify the relative inconsistency of effects, using a scale of low (< 25%), moderate (25–75%) and high (≥ 75%) and the associated significance with an a priori alpha level of p < 0.05. All statistical analyses were carried out using Jamovi [27]. Training frequency of the lower and upper body was defined as the number of sessions that included exercises targeting those areas, respectively, per week. Comparisons were made between the lower frequency and higher frequency groups within each study. Training frequencies can be interpreted across the whole spectrum, with some sports considering a ‘high’ frequency to be the equivalent of a ‘low’ frequency in other sports; for example, within basketball, the majority of the season is spent playing three games per week, whereby one or two dedicated RT sessions would be considered high frequency, compared to a sport such as rugby or American Football whereby that same frequency would be considered low. From a research perspective, there is also no clear definition of what constitutes high and low frequencies, with some studies for example labeling three sessions a week as low [28, 29] frequency and some as high [30‐33]. The authors, therefore, have not definitively classified any frequency as either being low or high but made comparisons as lower and higher. Due to body mass not being reported for individual groups in all of the included studies both pre- and post-intervention, changes in relative strength of each group could not be assessed. Instead, baseline strength between groups within each study were compared separately to highlight magnitude of strength differences between the study groups. All studies included within the meta-analyses were independently evaluated by two of the authors (NR and PC) for methodological quality.

Two thousand, one hundred and thirty-four studies were identified within the four databases highlighted in Sect. 2.2. Figure 1 illustrates that of the total studies identified, 142 articles were duplicates and, therefore, removed first. Following the application of the predetermined inclusion/exclusion criteria to both titles and abstracts of the identified studies, and with further inspection of the full text if required, a total of ten studies remained for further analysis [28‐37].

The results of two different meta-analyses were calculated to evaluate the effectiveness of the interventions on both lower and upper body strength pre- and post-intervention as well as the differences in effect between the lower and higher frequencies for each study (Table 1). Pre- to post-intervention g values can be seen in Supplementary Information Figures S1 and S2, with most of the interventions, regardless of frequency, demonstrating small-to-moderate (g = − 0.40–1.11) increases in strength. The estimated overall effect for strength in both the lower body (g = 0.562) and upper body (g = 0.323) pre- to post-intervention demonstrated the effectiveness of resistance training with significant increases for each of the meta-analyses, respectively (p < 0.001) (Table 1). When comparing lower and higher frequencies groups of each study, no overall significant effect was observed for lower body (p = 0.453 and g = 0.088) and upper body (p = 0.505 and g = 0.088).

Heterogeneity assessments of the completed meta-analyses were conducted and can be seen in Table 1, with inconsistency of effects being extremely low pre- to post-intervention (Tau² =  < 0.001, I² = 0%). A Cochrane risk of bias assessment was completed (Fig. 2) on all studies that described some level of randomization within their methods, with the results of this generally showing low risk of bias around the reporting of blinding of participants as this was unlikely to influence the results. Selection bias for the most part was unclear as the majority of the studies, although stating that groups were randomly allocated, did not report methods of allocation.

Due to a whole range of frequencies being investigated across the studies analysed in this review (Tables 2 and 3), the authors were unable to make enough comparisons to moderate the meta-analyses based on the frequency of each group. The lack of grouping has, therefore, led to some crossover between studies that have used one frequency as the ‘higher’ frequency that has also been included in a different study as the ‘lower’ frequency. An example of the crossover in training frequencies is demonstrated by McLester et al. [28] and Schoenfeld et al. [29] who both utilized three times per week as their higher RT frequency, whereas three times per week was used as the lower frequency in a number of the other studies included [30‐33]. Although the crossover between descriptors of ‘lower’ and ‘higher’ frequencies may appear to be a possible issue in the reporting of data, it is important to bear in mind that the differences in effect observed range from trivial to small (g = − 0.10–0.33) for both lower and upper body regardless of the descriptor used. It is possible, however, to make a number of direct comparisons based on studies that utilized the same frequencies within their interventions. As mentioned previously, McLester et al. [28] and Schoenfeld et al. [29] both investigated once per week compared to three times per week, with neither resulting in once a week being favorable compared to three times per week. When comparing three times a week as a lower frequency, in studies conducted by Colquhoun et al. [33] and Saric et al. [30] who investigated three times per week compared to six times per week, the results were mixed. Lower body strength improved to a greater extent for the group training three times per week in the study by Saric et al. [30], whereas the greater improvement in the study by Colquhoun et al. [33] was observed in the six times per week group. The reverse was true for the upper body. The only frequency analysed within this review that consistently demonstrated superior effect when compared with others for lower body strength was five times per week. Gomes et al. [35] and Zaroni et al. [37] both investigated once per week compared to five times per week, with both favoring the higher frequency for lower body strength. A similar trend was shown by Gomes et al. [35] in the upper body; however, Zaroni et al. [37] found that once a week demonstrated greater increases in upper body strength. Although the study by Hoffman et al. [32] observed in Fig. 3 very slightly favors ‘lower’ frequency, this is likely due to the study itself investigating four different frequencies and, therefore, the results were aggregated; when comparing all the groups individually the trend observed by Gomes et al. [35] and Zaroni et al. [37] is also demonstrated, with five times per week consistently demonstrating the greater effect for the lower body (Supplementary Information Fig. S3). When inspecting the upper body strength changes individually for the study by Hoffman et al. [32], the pattern followed the same trend as the study by Saric et al. [30] whereby six times per week was consistently the superior frequency (Supplementary Information Fig. S4). Again, however, the differences in observed effect between the studies was trivial to small, much like the overall effect of the meta-analyses which also demonstrated non-significant differences. Due to the low heterogeneity for both the upper and lower body observed in Table 1 (I² = 0%), there would continue to be minimal differences even if the sampling error of the interventions were to be removed.

Pre- to post-intervention showed trivial-to-moderate changes in maximum strength of both the lower and upper body with the majority of the interventions included within this review demonstrating the positive effects of resistance training on strength adaptations. It is, however, important to understand the potential mechanisms responsible when considering adaptations in strength. An increase in strength but no increase in muscle mass may suggest adaptations occurred predominantly due to increased fascicle length, reduction in pennation angle [38] and neural adaptations [39]. Alternatively, an increase in strength and increase in muscle mass will likely lean towards increased muscle thickness and pennation angle as well as possible increases in fascicle length [38]. This intuitively suggests that some strength adaptations will occur during hypertrophy training in response to an increase in muscle mass but may not be elevated to the level that would occur in response to a solely strength-focused training program. A summary of the studies analysed within this review can be seen in Table 2, whereby the set and repetition ranges of each of the interventions can be observed. It is clear that based upon our earlier definition of training methods seen in Sect. 1, the majority of these RT interventions are not focused on strength but heavily biased towards hypertrophy training with almost all interventions outlining sets above the recommended 3–6 repetitions [7], and most commonly employing 8–12 repetitions (see Table 2).

Despite all the interventions including exercises to RM, only three were explicitly reported to include the performance of sets to muscle failure [28, 30, 40]. The RM approach to load prescription is based upon performing the sets and repetitions with the maximum load possible to complete the full prescription, likely resulting in training to muscle failure. It is worth noting that within this review, ‘load’ is referred to when describing the amount of weight lifted, as Steele [41] and Arent et al. [42] have outlined how intensity (often used interchangeably with load) can be a better representative of effort. The reason for clearly defining the difference between load and intensity (effort) is to highlight that although the RM approach is performed with the maximum load possible for the sets and repetitions prescribed, this load may still be low–moderate, even when performed to failure and perceived to be high intensity by the athlete. Constantly training to muscle failure has been reported to have a potentially deleterious effect on performance [43]. Evidence of this effect has been observed when a group performing sets at a load relative to their maximum, compared to RM sets, demonstrated greater increases in jump performance, rapid isometric force production and muscular adaptations [43, 44]. The differences observed between the two groups is likely due to better fatigue management and potentially optimal performance adaptations, which is likely more appropriate for well-trained and professional athletes. The magnitude of load participants experienced throughout the majority of the interventions could explain why only trivial to moderate improvements in strength were observed across the 6–12 weeks. Schoenfeld et al. [40] have demonstrated similar hypertrophic responses are elicited when comparing a moderate load (three sets of 10RM) vs high load (seven sets of 3RM) when volumes are equated; however, the high load group demonstrated the greatest improvements in both back squat and bench press 1RM. In addition, well-trained populations will likely see less strength adaptations in response to hypertrophy training and specific strength training due to already having a greater base level of strength [20, 45]. Two out of the ten studies included within this review do not appear to be volume equated (Tables 2 and 3) [31, 32]. In a recent meta-analysis, Grgic et al. [21] used equated volume as a moderator, suggesting that increases in strength associated with higher frequencies of RT are largely attributed to the additional training volume. Due to the groups in studies by Hoffman et al. [32] and Kilen et al. [23] not being volume equated, it is not possible to determine whether training volumes observed in the higher frequency groups affected the resultant adaptations. The beneficial effect of increased RT volume on hypertrophic responses has previously been demonstrated [46]; however, the effect on strength is not as clear, or at which point increased volume may reduce the adaptive responses.

Only one study reported negative effects in response to a lower frequency RT intervention, where the authors observed small decreases in upper body and lower body strength [31]. A potential cause for these findings could be the testing battery used. Rather than using 1RM testing, a maximum voluntary isometric contraction was used to assess the knee and elbow flexors, which was unlike the actions used within the studies training intervention. Another reason for the reduction in strength could have been due to this being the only study to use a concurrent training approach. Due to the population used (i.e., military personnel), there was a requirement to not only train muscular strength but also aerobic and muscular endurance. The requirement to train concurrently is also present in team sports, however, and the findings from Kilen et al. [31] support the complexity of this process. Due to the demands of team sports, ensuring appropriate development of all physical attributes (i.e. muscular strength and power, muscular endurance and aerobic endurance) is essential, not only to enable a greater ability to recover between efforts in training and competition but also to recover between fixtures within congested periods of a season, or during tournaments as highlighted in Sect. 1. Wilson et al. [47] have suggested that significant decrements in maximal strength may not occur as a result of endurance training but only through the incorrect training modality and/or dose. Kilen et al. [31] demonstrated the effect of traditional concurrent training whereby their “classical training” (lower frequency) group who performed training sessions of high-intensity cardiovascular, muscular endurance and strength training within their program, experienced a decrease in maximal strength. In contrast, increases in maximal strength were observed in the “micro-training” (higher frequency) group who performed the same exercises, intensity and volume, albeit divided over shorter, higher frequency bouts (Figs. 3 and 4).

One of the aims of this review was to identify if RT frequency influences strength in well-trained athletes. Quantifying training experience and categorizing an athlete as ‘well-trained’ is not simple. Rhea [48] has proposed possible thresholds for g values to use based upon training experience (categorized as ‘untrained’, ‘recreationally trained’ and ‘highly trained’) when inspecting treatment effects. The criteria for this review, however, were for studies to state their population as well trained, but as Tables 2 and 3 outline, the variation in criteria for this population was large, ranging from 6 months to 10 years. The duration an athlete has trained for does not necessarily dictate how well trained they are, as the training they could have experienced at times throughout their career may be suboptimal. It could, therefore, be more applicable to categorize athletes based on their relative strength levels as evident within the study by Colquhoun et al. [33] who accepted subjects based upon criteria that included both length of training history and a minimum strength level (150% of bodyweight for a deadlift), similar to the selection criteria for ‘previously weight trained’ individuals outlined by Willoughby [49], of a parallel back squat 1RM ≥ 1.5 times bodyweight as this is more likely to dictate the response to the RT interventions. Relative strength levels pre-intervention have been calculated and are outlined in Tables 2 and 3. A possible reason for there being small-to-moderate changes overall regardless of frequency could be due to the populations of these studies actually being well trained as the majority of groups exceed the 1.5 times bodyweight threshold previously described for lower body strength by Willoughby [49]. The greatest difference observed between two frequencies in the lower body was observed by Yue et al. [36] (Fig. 3), an explanation for this potentially being due to the lower frequency group being the weakest at baseline in comparison to the higher frequency group and in comparison to the other studies investigated within this review. The lower relative strength results in a greater potential for improvement over the same period when exposed to the same volumes. The length of the interventions within this review could have also had an effect on the small-to-moderate change observed overall in Supplementary Information Figures S1 and S2. Unfortunately, the authors were unable to calculate relative strength changes due to a lack of reporting bodyweights post-intervention, or bodyweights for the different frequency groups rather than the whole sample population. The duration of the RT interventions included within this review was 6–12 weeks. If the athletes were well trained, as described, it is unlikely that large changes to relative strength will be observed pre-post.

A Cochrane risk of bias assessment (Fig. 2) was carried out to understand the bias across the studies that used a randomized approach. The overall conclusion would be that the risk of bias is low or even slightly unclear due to lack of detail around the way the randomization was carried out six out of the eight randomized studies. The concealment of allocation was also unclear in seven out of the eight studies, with the eighth explicitly outlining that concealment of allocation did not occur. Depending upon the setting of these studies, however, that is not always ecologically possible, particularly when working in a team sport setting. Ecological validity could also provide a rationale for the lack of control groups within all but one of the studies. Given links between strength training and reduction of injuries, it could be viewed as unethical to have a control group that only takes part in the sport if they already have a background in RT as this could put them at a greater risk of injury and possible reduction in competitive advantage over those without such a background.

The inclusion/exclusion criteria were initially designed to allow for a range of different potential moderators to be applied within the current meta-analysis. There was, however, a lack of consistent moderators available, which not only highlights a limitation of this review but also highlights gaps in the current literature and provides a strong rationale for future research areas in exercise prescription. The low consistency of effect (high heterogeneity) between the studies assessed in this review may have been attributable to certain commonalities. This low inconsistency is not necessarily a limitation but does highlight areas researchers need to expand on in the future. For example, all the interventions included in this review were completed on male subjects, with the vast majority being ‘recreationally trained’ and completing the same test for maximal strength. Some areas for future research would, therefore, be to investigate both sexes, but particularly females, to provide comparison with the current literature. Taking samples from athletes within different team sport settings would also be appropriate, as one set of sporting demands or type of sporting schedule may benefit from one particular approach compared to another. Research conducted within competitive team sports would, however, require the acceptance of ecological validity, whereby a number of factors that are likely outside of any investigator’s control would need to be considered. It is also important to understand that the description of team sport athletes as being ‘well-trained’ may only apply to their sport and not when to resistance training. The majority of interventions in this review included the exercises that were used to test their participants’ maximum strength (1RM). Utilizing the exercises tested within the intervention may have resulted in improvements purely based on improvement of technique or familiarization; however, due to the ‘well-trained’ nature of the population, this is unlikely. Another potential issue with the maximal testing used to assess strength was that only bench press was used as a measure of compound upper body strength, whilst most interventions included a full body approach, meaning there was a lack of evidence to demonstrate upper body strength increase as a whole. A possible limitation of only measuring maximal strength means that the rate at which the participants could produce force was not measured. The force production capabilities of athletes are important for performance and associated with injury risk reduction; therefore, not only is it important to apply force maximally, but the rate at which it is applied is also important. Measures of multi-joint rapid force production (e.g. using the isometric mid-thigh pull) should also be assessed when considering the implications for athletic populations.

Finally, as mentioned when considering concurrent training, Kilen et al. [31] described their higher frequency RT group as a “micro-training” group, and considering the overall lack of difference between training frequencies further investigation should investigate a variation of the term used by Kilen et al. [31] which has become more commonly used by practitioners which is “micro-dosing”. Micro-dosing was initially coined from a performance perspective by Hansen [50] but has not been widely used within the peer-reviewed literature, and therefore has no clear definition. We, therefore, define micro-dosing training as “the division of total volume within a micro-cycle, across frequent, short duration, repeated bouts” and suggest that such an approach should be thoroughly investigated in the future.