We performed a systematic search in the PubMed, Embase, and Web of Science databases from the date of their inception to 18 February 2024, with no language restrictions. When searching for eligible articles, we used the search terms (“strength training” OR “resistance training” OR “weight training” OR “power training” OR “plyometric training” OR “complex training” OR “weight-bearing exercise”) AND (“swimming athlete” OR “athlete” OR “swimmer” OR “competitive swimmers”) (Supplementary Appendix S2). The reference lists of the included studies and previous reviews were screened for additional studies. GLJ and JYW screened all search results to exclude duplicates. The titles and abstracts of the remaining studies were independently screened by GLJ and JYW based on inclusion and exclusion criteria. Full texts that met these criteria were independently screened by GLJ and JYW, and all disagreements were decided by YY.

In alignment with the PICOS framework (Hutton et al., 2015), our inclusion criteria were meticulously defined as follows: (a) Participants consisted of competitive swimmers; (b) The intervention included resistance training along with its advanced variations; (c) The comparison was made against habitual aquatic training routines; (d) Outcomes measured encompassed pre- and post-intervention maximum muscle strength of upper limbs, performance in the 25–200 m front crawl (seconds), maximum velocity (meters/second), stroke rate (cycles/second), and stroke length (meters/cycle); (e) The study design was restricted to published Randomized Controlled Trials (RCTs), whether individually designed, cluster-designed, or the initial phase of crossover studies. We excluded research focusing on the acute effects of a singular session on competitive swimmers and studies that lacked clear descriptions of the resistance training employed. Studies were also excluded if they did not provide means and standard deviations in their results or if the authors did not respond to our data inquiries.

Upon reviewing the titles and abstracts, we conducted a comprehensive assessment of the full texts for all pertinent articles. During this phase, two authors (GLJ and JYW) meticulously extracted data on (1) characteristics of the participants, such as sample size, age, and gender; (2) detailed descriptions of the interventions, including methodology of training; (3) specifics of the training variables; and (4) the main results of the study. In instances where raw data were incomplete, we reached out to the corresponding authors for clarification. Studies were excluded if the authors were unresponsive. GLJ and JYW independently evaluated all studies, utilizing the extracted data as their guide. Any discrepancies regarding the inclusion of a study were resolved through consultation with YY. The data from the studies that met our inclusion criteria are summarized in Table 1.

The risk of bias for each individual study was assessed independently by GLJ and JYW using the using the Cochrane Risk of Bias version 2 tool (RoB2) (Sterne et al., 2019), including five domains: randomisation process; deviations from intended interventions; missing outcome data; outcome measurement; and selection of reported results. For cluster randomized controlled trials, the RoB 2.0 tool uses an additional domain to assess risk of bias due to the timing of identifying and recruiting participants (Eldridge et al., 2016), in addition to the five domains above. Each area was assessed as (1) high risk, (2) low risk and (3) some concern. For each study, if all domains showed low risk, the overall risk of bias was low; if any of the above domains showed high risk, or the assessment results of multiple domains showed some concern, the overall risk of bias was high; otherwise, Overall risk of bias was low. Disagreements were resolved by consensus among reviewers or in consultation with a third reviewer.

Two independent reviewer (GLJ and JYW) also assessed the strength of the current evidence for each outcome and subgroup using the GRADE method (Guyatt et al., 2008). In our review, the evidence started with high quality and was downgraded for each of the flowing issues: (1) publication bias (at least 10 studies, Egger test p-value < 0.05) (Ioannidis and Trikalinos, 2007); (2) imprecision when fewer than 400 subjects were included in the meta-analysis (Mueller et al., 2007); (3) when more than 25% of the studies from high risk of bias were included (Mascarenhas et al., 2021); (4) I² > 50% in the heterogeneity (Higgins and Green, 2008). We choose to summarise judgments across domains using the 4 levels of confidence of the GRADE approach: very low, low, moderate, or high (Brozek et al., 2009).

To determine the effectiveness of resistance training on performance in competitive swimmers. Missing standard deviations (SDs): When standard errors (SEs) instead of SDs were presented, the former was converted to SDs: (SD = SE*√n). If both were missing, we estimated the SD from credible intervals (CrIs), t values, or p values as described in Section 7.7.3 of the Cochrane Handbook for Systematic Reviews (Higgins and Green, 2011). The amount of baseline and post change between the experimental group and control group were calculated by the following formula: Mean change = Mean post   −   Mean baseline ;   SD change = SQUAT  SD baseline ^ 2 + SD post ^ 2   −   2 * R * SD baseline * SD post , where R is a constant (R = 0.5) (McKenzie and Brennan, 2019). The effect sizes were calculated as mean differences (MD) for continuous outcomes. Due to maximum muscle strength of upper limbs was assessed by different test, we chose the standard mean difference (SMD) to assess effect size. To evaluate the reliability of our estimates, we utilized 95% CrIs. We used the I² test and p-value to analyze the statistical heterogeneity between the studies, and when I² > 50%, it meant that there was substantial heterogeneity, we chose a random effects model, otherwise a fixed effects model. We conducted quantitative analysis through Egger’s test for all outcomes to investigate the presence of publication bias (p < 0.05, indicating significant publication bias). Based on the resistance training methodology, a subgroup analysis was performed to examine whether resistance training methodology influence the efficacy of resistance training on athletic performance. Data analysis was achieved based on the R statistical environment (V.4.3.2, www.r-project.org). The measured effects were considered significant at p < 0.05.

The literature search identified 243 potentially relevant studies (Figure 1). A total of 78 duplicates were removed, and a screening of the titles and abstracts excluded 132 studies. A total of 65 full-text articles were retrieved, GLJ and JYW confirmed the outcomes of interest by viewing the full text, and 13 studies (Toussaint and Vervoorn, 1990; Girold et al., 2006; Girold et al., 2012; Naczk et al., 2017; Gourgoulis et al., 2019; Sammoud et al., 2019; Barbosa et al., 2020; Amara et al., 2021a; Pinos et al., 2021; Sammoud et al., 2021; Amara et al., 2022; Khiyami et al., 2022; Norberto et al., 2023) with 267 participants (male: 168, female: 99), mean age of 16.1 years (age range 10–24 years), mean height of 151.1 cm (range 143–179 cm), mean body weight of 59.8 kg (range 36.2–73.6 kg), regional or national competitive swimmers with 2–8.7 years of training were included in the systematic review and meta-analysis (Table 1).

The resistance training lasted for 3–11 weeks, with total sessions of 12–54 times. Two studies in the included literature comparing concurrent resistance training (CRT) (using paddling or water parachute in the water and dry-land resistance training) with the habitual aquatic training (Amara et al., 2021a; Amara et al., 2022). Four studies were direct comparisons of aquatic resistance training with water parachute or hand paddles (ART) and the habitual aquatic training (Toussaint and Vervoorn, 1990; Girold et al., 2006; Gourgoulis et al., 2019; Barbosa et al., 2020). Four studies were direct comparisons of dry-land resistance training (DLRT) and the habitual aquatic training (Girold et al., 2012; Naczk et al., 2017; Pinos et al., 2021; Khiyami et al., 2022), 2 studies were direct comparisons of power training (PT) and the habitual aquatic training (Sammoud et al., 2019; Sammoud et al., 2021). One study involved direct comparisons of DLRT, PT, and habitual aquatic training (Norberto et al., 2023).

Five studies involving 129 swimmers assessed maximum muscle strength of upper limbs (Table 2). Overall, the results of low quality showed that resistance training significantly improved the maximum muscle strength of upper limbs of swimmers compared with the habitual aquatic training [SMD: 0.89, 95% CrI (0.54, 1.24), I² = 36%, Figure 2].

Our results showed that resistance training significantly improves the performance of competitive swimmers in the 25 m (MD: −0.78, 95% CrI (−1.15, −0.41), I² = 14.9%, 3 studies, low quality of the evidence), 50 m (MD: −0.62, 95% CrI (−1.01, −0.24), I² = 0.0%, 12 studies, moderate quality of the evidence), 100 m (MD: −2.36, 95% CrI (−3.33, −1.39), I² = 0.0%, 5 studies, low quality of the evidence) and 200 m front crawl (MD: −4.91, 95% CrI (−8.65, −1.18), I² = 13.0%, 3 studies, very low quality of the evidence) (Figure 3). It is worth noting that the improvement in swimming performance may be due to significant improvements in velocity (MD: 0.05, 95% CrI (0.01, 0.10), I² = 27.0%, 6 studies, low quality of the evidence, Figure 4) and stroke rate (MD: 0.04, 95% CrI (0.02, 0.06), I² = 59.0%, 6 studies, very low quality of the evidence, Figure 5), not due to improvements in stroke length (MD: −0.03, 95% CrI (−0.06, 0.01), I² = 0.0%, 6 studies, low quality of the evidence, Figure 5).

The results of subgroup analysis showed that CRT (SMD: 1.60, 95% CrI (0.62, 2.59), 1 study), PT (SMD: 1.03, 95% CrI (0.25, 1.82), 1 study), and DLRT (SMD: 1.18, 95% CrI (0.53, 1.83), I² = 27.7%, 2 studies) can significantly improve the maximum muscle strength of upper limbs in competitive swimmers, excepted the ART (Table 3).

Only two forms of resistance training (CRT and PT) evaluated for their effects on the 25 m front crawl. The results of subgroup analysis showed that CRT (MD: −0.74, 95% CrI (−1.78, −0.38), 1 study), and PT (MD: −0.86, 95% CrI (−1.48, −0.24), 2 studies) can significantly improve the 25 m front crawl of competitive swimmers. Four forms of resistance training were evaluated for their effects on the 50 m and 100 m front crawl. The results of subgroup analysis showed that only CRT (50 m, MD: −1.08, 95% CrI (−1.78, −0.38), 1 study; 100 m, MD: −2.35, 95% CrI (−3.62, −1.08), 1 study) and PT (50 m, MD: −1.51, 95% CrI (−2.82, −0.19), I² = 0%, 3 studies; 100 m, MD: 4.00, 95% CrI (−6.74, −1.26), 1 study) significantly improved the 50 m and 100 m front crawl respectively. Three forms of resistance training (ART, PT, and DLRT) were evaluated for their effects on the 200 m front crawl. The results of subgroup analysis showed that only PT (MD: −7.00, 95% CrI (−13.57, −0.43), 1 study) significantly improved the 200 m front crawl. Three forms of resistance training (ART, CRT, and DLRT) were evaluated for their effects on the velocity, stroke rate and stroke length. The results of subgroup analysis showed that only CRT significantly improved the velocity (MD: 0.15, 95% CrI (0.06, 0.24), 1 study) and stroke rate (MD: 0.09, 95% CrI (0.05, 0.13), 1 study), and no form of resistance training found to significantly improve stroke length.

Of the 13 trials, for overall bias, 5 studies were assessed as low risk of bias, 7 as some concerns and 1 as high. In the randomization process, 7 trials were low risk, 4 trials were some concerns; for deviations from intended interventions, 11 trials were at low risk, 2 trials were some concerns; in the missing outcome data, 10 trials were low risk, 2 trials were some concerns, and 1 trial was high risk; in the measurement of the outcome, 13 trials were low risk; in the selection of the reported result, 11 trials were low risk, 1 trial was some concerns (Table 4).

Our results indicated large effects of resistance training on measures of maximum muscle strength of upper limbs in competitive swimmers, and significantly improved 25, 50, 100, and 200 m front crawl performance. The muscle strength in the upper limbs, main muscle groups involved in front crawl swimming propulsion (Catteau and Garoff, 1990), have great representativeness in a swimming training program’s final result (Trappe and Pearson, 1994; Maglischo and do Nascimento, 1999). Hawley et al. (1992) demonstrated a strong relationship (r = 0.82 and 0.93, respectively; p < 0.05) between the maximum muscle strength of upper limbs and front crawl sprint swimming performance over 25 m and 50 m. This finding was supported by Amara et al. (2022), who stated that resistance training effectively improves swimming initiation and turning performance, which was particularly important for short distance swimming. Resistance training not only improve upper body strength in competitive swimmers but may also optimize energy expenditure efficiency. According to Stian Thoresen Aspenes and Karlsen (2012), through specific resistance training, athletes could improve their propulsive efficiency in the water, which directly reduced energy expenditure during swimming. More efficient muscle use reduces the amount of energy required per stroke, allowing for higher speeds and better endurance performance during long-distance swimming. Additionally, systematic resistance training has been found to help reduce the risk of shoulder injuries, a common problem among swimmers. Previous research results showed that overuse injuries caused by repetitive shoulder movements can be effectively prevented by strengthening and stabilizing the muscles surrounding the shoulder. Strengthening these muscle groups can improve joint stability and reduce muscle strains and tendonitis that can occur during high-intensity training and competition (Tate et al., 2012; Hill et al., 2015; McKenzie et al., 2023). Together, these studies indicate that incorporating resistance training into competitive swimmers’ training programs is critical to improving their swimming performance and competitiveness.

In addition, we used specific parameters during swimming, such as swimming speed, stroke rate and stroke length, to explore how resistance training can improve the front crawl performance of competitive swimmers (time taken to test the corresponding distance). Our results showed that resistance training might have increased swimming speed and ultimately front crawl performance (time taken to test the corresponding distance) in competitive swimmers simply by increasing swimmers’ stroke rate rather than stroke length. Strass (1988) conducted low-repetition high-intensity DLRT (3 sets of 3 repetitions, 90%–100% 1RM) on 10 male competitive swimmers for a total of 24 sessions over a 6-week period, and the results showed that DLRT increased swimming speed by increasing stroke length (reducing stroke rate) and ultimately improved 25 m and 50 m front crawl swimming performance. Girold et al. (2012) conducted a randomized controlled trial of 12 sessions of low-repetition high-intensity DLRT over a 4-week period (3 sets of 6 repetitions, 80%–90% 1RM) significantly improved swimmers’ stroke length and 50 m front crawl performance, but was not significantly different from habitual aquatic training. Recently, a randomized controlled trial by Amara et al. (2021a) conducted multiple sets of high-repetition moderate-intensity CRT (3–6 sets of 6–10 repetitions, 60%–80% 1RM), the results showed that compared with habitual aquatic training, CRT did not significantly improve the stroke length of competitive swimmers, but improved the front crawl performance by increasing the stroke rate and ultimately. In addition, many randomized controlled trials of ART (multiple sets and repeated with low-intensity, increased resistance during water training through resistance bands, hand paddles, or water parachute) had not found that resistance training significantly improved the stroke length of competitive swimmers (Girold et al., 2006; Gourgoulis et al., 2019; Barbosa et al., 2020). Our review of the literature seems to suggest that higher intensity, low repetition resistance training may be responsible for improving stroke length in competitive swimmers. However, since the included literature was only RCT studies, only Girold et al. (2012) evaluated stroke length for resistance training with higher intensity and low repetitions, and the data were merged. Other studies have used resistance training with multiple repetitions at a lower intensity. Therefore, overall, resistance training did not significantly increase the stroke length of competitive swimmers, but improved swimming performance by increasing stroke rate. In addition, more high-quality studies are needed in the future to verify the effectiveness of high-intensity, low-repetition resistance training models on the stroke length of competitive swimmers.

Our study had several limitations. First, the literature we included focused solely on the impact of various resistance training on competitive swimmers’ front-crawl swimming performance, and these results cannot be transferred to other swimming strokes. Therefore, our findings lack universality. Second, the small amount of literature included led to specific resistance training methodology having only one study in our subgroup analysis, which was the main reason for the low credibility of our results. In addition, the studies included in this meta-analysis involved a wide range of ages and swimming training years, which may have an impact on the interpretation and generalizability of the study results. Age and years of training are important factors that affect swimming performance and training adaptability. Different ages and main events of the swimmer may have significant differences in physiology, biomechanics, and psychological status, and these differences may affect training effects and competition performance (Gonjo et al., 2020; Gonjo et al., 2021; Raineteau et al., 2023a). Likewise, length of swimming experience may also have a decisive impact on an athlete’s performance. Therefore, differences in these factors may limit the generalizability of our analytical results, and future studies should consider the potential effects of these variables and attempt more detailed subgroup analyses. Due to the limited number of studies, we were unable to conduct an in-depth analysis of the dosage of specific resistance training. Previous literature has shown that specific dosages (training periods, sets, repetitions, and intensity) significantly affect the effect of resistance training on athletic performance (Lesinski et al., 2016). Therefore, future research should involve a large number of high-quality studies to thoroughly explore the impact of resistance training dosage on competitive swimmers’ performance, providing clear training guidelines for coaches and competitive swimmers. Lastly, in this study, we did not conduct a detailed distinction and analysis of the differences between men and women in the effectiveness of training administration. This also may be a limitation of this study, and future research can further explore the impact of gender factors on training effects.