The systematic review followed the 2020 updated Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (28). The search strategy was performed (July 2024) in PubMed, Scopus, SPORTDiscus, and Web of Science. The selected keywords and Boolean operators for the search were: (“critical power” OR “critical speed” OR “critical velocity”) AND (“test” OR “tests” OR “testing” OR “method” OR “methods”) AND (“running” OR “runners” OR “runner” OR “athletes” OR “athlete”) AND (“field” OR “track” OR “field condition”).

All records retrieved from the database search were imported into the Rayyan systematic review software (29) for screening. Two reviewers (DP and JK) independently conducted the screening process, and any disagreements were resolved by consultation with a third reviewer (LL). Exclusion criteria included non-English language publications, reviews, meeting abstracts, letters, corrections, editorials, and non-human studies. Titles and abstracts were initially screened to exclude studies lacking the predefined keywords. The inclusion of articles was restricted to those published between 2010 and 2024. Studies were further assessed based on the PECO criteria (Table 1), with eligibility determined during the full-text review. Studies focused solely on novice or untrained individuals were excluded at this stage to ensure applicability to athletic populations. Additionally, studies with fewer than six participants were excluded to minimize deviations due to small sample sizes (30). Data extraction was conducted by one reviewer (LL) and cross-verified by a second reviewer (MK) where necessary.

The assessment of study quality utilised the Downs and Black scale (31). The original checklist includes 27 questions assessing the quality of reporting, internal and external validity, and statistical power. For this review, the checklist was adapted by excluding 12 questions that were not relevant to the selected studies, such as those specific to intervention designs. As a result, 16 questions were used to evaluate the studies, focusing on reporting clarity, study design validity, and appropriate statistical analysis (Supplementary Table S1). Two reviewers independently conducted the quality assessment, with disagreements resolved by a third reviewer. Each item received a binary score, with 1 representing “yes” and 0 indicating “no” or “unable to determine.” The total points were converted to percentages and adjusted according to the number of selected questions. In this review, the Downs and Black (31) total score was adjusted to a maximum of 15 points, as one question was considered self-evident and not included in the scoring. Studies scoring less than 45.4% (6 points or fewer) were categorized as having “poor” methodological quality. Scores between 45.4% and 63.6% (7–8 points) indicated “fair” quality, while scores above 63.6% (9 points or more) reflected “good” methodological quality.

Table 2 summarises the main characteristics of participants from the included studies. A total of 285 participants contributed to this research (207 males, 66 females, mean ± SD: age = 30.6 ± 6.3 years). Smyth and Muniz-Pumares (41) analyzed a dataset of 31,190 runners from Strava®; however, due to the large sample size and its focus on aggregated training data, this study was not included in these statistics. Ten studies included only male participants, while seven had a mixed sample, and only one study focused solely on female participants. The limited representation of female participants may influence the generalizability of the results, which should be considered in further analysis. The participants’ training experience varied widely, ranging from as little as 6 months to over 18 years. Training volumes also differed, from 4 h per week to 110 km per week, encompassing a mix of recreational, regional, and national-level runners.

The research included a variety of field-based running test procedures to evaluate CS/CP, reflecting the diversity in approaches across studies (Table 3). The TT were the most popular method, involving fixed distances (iso-distance) ranging from 400 to 3,600 m (14, 16, 20, 22, 23, 32–36, 39, 40, 42), with the most popular over three fixed distances: 1,200, 2,400 and 3,600 m (22, 33–36). Some studies employed fixed-duration trials (iso-duration) (23, 40, 42, 44). The 3MT (14, 15, 38, 43, 44) challenged participants to achieve maximum speed over a brief period, offering an alternative perspective on performance. Hunter et al. (44) explored the application of 3MT and TTs alongside habitual training (HAB) data, demonstrating the potential for estimating CS/CP from non-invasive and remote methods. Similarly, Smyth and Muniz-Pumares (41) analyzed large-scale activity datasets logged on Strava®, offering an innovative approach to derive CS and CP from everyday training records.

In addition to these methods, Vassallo et al. (43) applied the gold-standard TTE test in a field setting and introduced an enhanced 3MT model that incorporates energetic calculations, offering a practical approach to estimate CP and CS in outdoor conditions with power derived from GPS and speed variations accounted for. Notably, Olaya-Cuartero et al. (37), Ruiz-Alias et al. (23), Hunter et al. (44) and Van Rassel et al. (22) utilized the Stryd power meter to determine CP, either through the 9,3-minute CP test (23, 37) or via the iso-distance model (22). Several mathematical models were employed for CS determination, including linear distance-time relationships (e.g., Linear-TD), inverse-of-time models (e.g., INV), work-time calculations (e.g., linear work-time), and hyperbolic models (e.g., 2-HYP). Some studies also compared field test results with laboratory-based protocols (15, 16, 23, 33, 35, 38), as well as with performance outcomes, enhancing the assessment of the validity and reliability of these methods.

To summarize, the field-based methods for determining CS/CP were grouped into three main categories: TTs (including traditional iso-distance and iso-duration trials, as well as innovative approaches like habitual activity tracking and raw race data analysis), TTE, and 3MT.

The study quality was evaluated based on a maximum score of 15 points. Five studies achieved the score of 12 (80%), seven studies scored 11 (739%), four studies scored 10 (66%), and the remaining four studies scored 9 or 8 (60%–53%). No studies scored below 8, indicating generally fair to good methodological quality across all included studies (Supplementary Table S2).

In endurance research, two primary testing methods exist: the traditional TTE, where participants run at a constant speed over several trial (3–5) until exhaustion and the TTs, where athletes cover a fixed distance or run for a set duration at their maximal intensity. While the TTE is effective in controlled environments, they are less practical for field applications due to the difficulty of maintaining consistent pacing. For instance, Vassallo et al. (43) conducted a TTE test using audio tone for pace synchronization, ending the test when the athlete could no longer meet the required speed. However, even with auditory guidance, verifying that participants truly exert maximal effort remains challenging in the absence of physiological markers like VO_2max attainment. Moreover, environmental factors, such as wind, temperature, or uneven terrain, can further complicate pacing and reliability, making TTE a less optimal method for field conditions. These limitations underscore why TTE tests are less favored for field assessments compared to other methods.

The choice of mathematical model has a key influence on CS and D′ estimates. Linear models (INV, EXP) typically provide higher CS values, while non-linear models (2-HYP, 3-HYP) are more reliable for estimating D′ but exhibit greater variability. These differences should be carefully considered when selecting the number and duration of trials, as inaccuracies can be a source of variability in performance predictions. While mathematical models are not the primary focus of this review, they are extensively discussed in other studies (45–48), providing deeper insights into their application and limitations.

The 3MT is a valuable alternative to traditional TTE and TTs, offering a more time-efficient way to estimate CP and CS without requiring multiple trials. Its design is grounded in the principle of fully depleting the anaerobic reserve (D′) within the first 150 s of an all-out effort, after which the power output or speed stabilizes. This stabilization phase reflects the athlete's CS, representing the maximum sustainable speed without further depletion of D′ (14, 15). This stabilization represents the transition from anaerobic to aerobic energy pathways.

The simplicity and practicality of 3MT has been enhanced by tools such as GPS devices, accelerometers, power meters and stopwatches, enabling its use in field settings (14, 15, 43, 44, 54). Vanhatalo et al. (55) demonstrated that these technologies facilitate the application of 3MT in real-world conditions, making it a valuable tool for athletes and coaches. However, transitioning from controlled laboratory settings to outdoor environments introduces challenges, including surface variability, wind, temperature, and athlete motivation. These factors, combined with the potential for pacing behaviors, underscore the importance of robust protocol design. Hunter et al. (44) found that despite instructions, pacing behaviors were observed, highlighting the need for careful participant preparation and protocol design.

Studies have consistently validated the 3MT for estimating CS. Aguiar et al. (14) reported high reliability for CS determination, with test-retest coefficients above 0.90, though the test underestimated D′ by approximately 16%. Similarly, Broxterman et al (15). found a strong correlation (r = 0.92) between field-based 3MT and treadmill-based TTE tests, further confirming the validity of 3MT in estimating CS. Although W′ was again underestimated. Pettitt et al. (38) demonstrated that GPS-enabled 3MT provides reliable predictions for races ranging from 1,600 to 5,000 m, although errors increased for shorter events like the 800 m.

Innovations in 3MT methodology have aimed to address some of its limitations. Vassallo et al. (43) introduced a novel power-based model that tracked real-time variations in power output during over-ground running. By focusing on power rather than speed, this approach accounted for pacing fluctuations, acceleration, and deceleration, providing a more comprehensive assessment of mechanical demands. Although this model tended to overestimate CP and D′ compared to traditional protocols (by approximately 25 W for CP and 7 kJ for D′), it highlighted the potential for integrating power-based measurements into 3MT applications, particularly for sports requiring frequent speed changes. Another innovative approach was demonstrated by Hunter et al. (44), who showed that 3MT could be conducted unsupervised with reliable outcomes when participants followed clear instructions. However, this adaptation raises questions about whether unsupervised results are comparable to those obtained under supervised conditions. Future research should focus on establishing the reliability and validity of unsupervised 3MT across diverse field environments, particularly for recreational athletes.

While the 3MT offers significant advantages in terms of practicality and efficiency, it appears better suited for experienced runners accustomed to performing at maximal intensity. Novice runners may struggle to deliver consistent all-out efforts, potentially affecting test accuracy. Additionally, field tests lack the ability to verify maximal efforts through physiological markers like VO_2max attainment, which adds complexity to standardizing results. The consistent underestimation of D′ presents challenges for prescribing high-intensity interval training, emphasizing the need for careful protocol design. Environmental factors, such as surface variability and weather conditions, further highlight the importance of controlled settings to ensure reliable outcomes.

In conclusion, the 3MT integrates anaerobic and aerobic parameters in a single trial, making it a valuable tool for estimating CS in both laboratory and field settings. Continued advancements, such as the integration of power-based monitoring and innovative technologies, hold promise for enhancing the test's accuracy and applicability. However, its dependence on maximal effort and sensitivity to pacing behaviors highlight the importance of robust protocols, participant preparation, and environmental consistency to optimize test reliability.

The CS is a valuable tool for assessing endurance performance and planning effective training strategy for runners. However, applying CS in real-world settings presents unique challenges, particularly due to variability in physiological responses around the CS threshold. This variability creates a “grey zone,” as described by Jones et al. (11), where uncertainty exists near the estimated CS. For example, a CS value of 5 m/s with a 5% error margin could place an athlete either in the heavy-intensity domain (below CS at 4.75 m/s) or the severe-intensity domain (above CS at 5.25 m/s), complicating training prescriptions and pacing strategies. Caen et al. (51) also emphasized that such variability is influenced not only by biological factors but also by methodological aspects, including differences in trial durations and testing protocols. Addressing these factors is essential to refine CS testing protocols, ensuring greater accuracy and applicability in both research and practical training settings.

Studies have validated the robustness of CS for training purposes. Figueiredo et al. (56) demonstrated that CS and peak running velocity are equally effective for prescribing endurance training in recreational runners when determined under controlled track settings. At the same time, gender-specific differences in pacing strategies highlight the practical relevance of CS. Female marathon runners, for example, tend to perform closer to their CS than males, particularly in time brackets of 170–360 min. This even pacing approach contrasts with the variability observed in male runners, who often experience significant slowdowns in the latter stages of a race, especially those with slower performances. For less experienced runners, individualized pacing strategies based on CS are essential to mitigate fatigue and sustain performance (41).

In competitive scenarios, the utility of CS extends to performance prediction and race strategy planning. Elite marathoners typically compete at approximately 95% of their estimated CS (57), while half-marathon runners have been shown to race at around 97.3% of their CP, with no significant differences between CP determined by the 9/3-minute Stryd CP test and the race's target CP (37). Supporting this, Smyth and Muniz-Pumares (41) demonstrated that runners initiating a marathon at approximately 87.6% of their CS generally achieve better outcomes, whereas those starting above this threshold often experience pronounced slowdowns as the race progresses. Similarly, marathon speeds average around 84.8% of CS, with faster runners competing closer to their CS (∼93%) compared to slower runners (∼78.9%). These findings highlight CS's versatility not only as a robust predictor of performance but also as a valuable tool for refining pacing strategies across different performance levels. By tailoring race strategies based on CS, runners can better balance energy expenditure and mitigate fatigue, ultimately enhancing race outcomes.

The application of CS is not limited to race-day strategies but extends to training program design. Field-based CS tests, such as TTs and 3MT, provide reliable metrics that coaches can use to design individualized training programs. Intervals prescribed just below CS help optimize aerobic endurance, while those above CS engage anaerobic capacity. Clark et al. (58) demonstrated that high-intensity interval training (HIIT) prescribed at intensities between 110% and 130% of CS has been shown to yield significant improvements in aerobic performance. Moreover, enhancing CS can translate directly into improved race results. For example, increasing CS from 4.90 m/s to 4.99 m/s has been shown to result in a 36 s reduction in a 10,000 m race time, underscoring the substantial impact of targeted CS training program (35).

By integrating CS into training and race planning, runners and coaches can better manage pacing, fatigue, and energy reserves, leading to optimized performance. These applications highlight CS's critical role in bridging theoretical models and practical strategies, making it an indispensable tool for endurance athletes aiming to achieve peak performance.

Field-based testing methods offer significant advantages for a wide range of athletes, including their simplicity, accessibility, and suitability for various performance levels. For recreational runners, tests can often be performed using minimal equipment, such as a stopwatch and a marked distance, and completed within a single day with a 30-min recovery period. However, several practical factors must be considered to ensure test reliability and validity.

To aid practitioners in implementing these methods effectively, Table 4 provides a summary of key recommendations and limitations for different protocols, including TTs, TTE, 3MT, and approaches based on raw or habitual training data. This table highlights critical parameters such as test duration, recovery times, environmental considerations, and equipment requirements, offering a practical guide for optimizing test design and execution.