Test-retest reliability of the cardiopulmonary fitness test protocol
It was not possible to include a comparative matched control group in this longitudinal study of fitness measurements in professional soccer players. In consideration of this, the test-retest reliability of the primary cardiovascular exercise protocol was pre-defined using a convenience sample so to be able to subsequently infer meaningful change in performance over that of any variation in measurement error.

Test- retest data of maximal aerobic power (VO₂ max) and anaerobic threshold (AT) were collected from 20 male undergraduate students (20 ± 1.8 years; 77.6 ± 4.7 kg) (Table 1) who all performed the same laboratory test on two occasions separated by one week in similar laboratory conditions (~18^oC) and at similar times of the day. The information derived from these tests was used for the purpose of identifying pre-defined limits of test-retest agreement (Bland and Altman, 1986) for the test procedures. By using this process, acceptable reliability was set at the minimal positive and negative limits which encompassed all of the 20 observations. Subsequent seasonal variations between the repeated performances of the soccer players could thereafter be ascribed to either technical/trivial variability (if they fell within the pre-defined limits of agreement, or biological/meaningful variability (if they fell outside of the pre-defined limits of agreement).

Professional soccer players
Over the three year period, a pool of 42 male professional 1^st team soccer players (age: 25 ± 3.5 years, stature: 1.78 ± 0.3 m, body mass: 79.4 ± 1.6 kg) from a single English Championship Club performed the experimental protocols. Only players free of injury and in full training were tested as part of the study. During the three year period, some changes in personnel were observed on test-to-test occasions due to both temporary and permanent inter-club transfers, player availability, and also short- and long-term injuries.

Test series
The same sequence of tests were conducted on eight occasions over 3-years in the same order and, where possible, at the same time of day in a temperature controlled (~18^oC) exercise laboratory. Despite some changes in personnel, the cohort was a homogenous group of outfield soccer players and the physical characteristics of the players at each of the repeated measures are shown in Table 2.

Over the three seasons (1: 2001-2002, 2: 2002-2003, 3: 2003-2004) the same sequence of tests was performed on eight occasions: 1) immediately prior to pre- season training (PPS), 2) mid-season (MID), and 3) end of season (EOS). No tests were conducted at the MID interval of season one due to heavy match commitments at that time.

The testing procedure followed the same order for each player: 1) baseline blood lactate evaluation 2) basic anthropometric evaluation 3) countermovement jump (CMJ) tests (best of 3 jumps and 20-s sustained jumping) 4) combination test of AT and VO₂ max.

Test procedures
Body composition was assessed by bioelectrical impedance (BIA) (Tanita TBF-551 Body composition scales). Participants were allowed 2ml·kg^-1 body mass of water in the hour preceding testing to standardize hydration (Edwards et al., 2007) and were asked to refrain from the consumption of alcohol and products containing caffeine in the 24 hr period preceding testing. Finally, all participants were requested to void as fully as possible prior to the evaluation of body composition.

Body mass and stature were determined using standard laboratory measurement techniques (Tanita TBF-551 scales, Tanita, Middlesex, UK; SECA 240 Stadiometer, Hamburg, Germany).

A 10-min warm up was completed by each player prior to the CMJ tests which included light jogging, specific mobility exercises and a stretching routine.

The CMJ was performed according to the Bosco method (Bosco et al., 1983), whereby the highest vertical jump (jump height in cm) and the longest flying time (time between take-off and landing) were recorded. The 'best' CMJ of three efforts was recorded and the subsequent mean of 20-s sustained CMJ effort was also recorded to ascertain whether fatigue might be a factor in jump performance. The apparatus consisted of a digital timer connected by a cable to a jump platform (NewTest version 2.0). The timer was triggered at take-off and then by the participant's feet at the moment of touchdown (Hoffman and Kang, 2002). The subject started from an upright standing position on the platform, and following the eccentric phase (corresponding to a semi squatting position), the participants jumped vertically without using arms to aid further height (arms remained at both sides, hands on the hip throughout the tests).

All participants completed the same combination test of AT and VO₂ max. All tests were completed on a computer controlled treadmill (Woodway PPS 55, Germany) and the test comprised a series of incremental steps which increased in speed every 3.5-min to a maximum of 4.03 m·s^-1. The test started with a warm up of 2-min at a walking pace of 0.97 m·s-1and was followed by 3-min at 2.78 m·s^-1 prior to the first stage of the test. The test was designed to enable the participants to reach aerobic steady state within 3-min in each stage (3.47, 3.61, 3.75, 3.89 & 4.03 m·s^-1) with a further 30-s included in each stage for the collection of blood samples. During each 30-s period, participants were stationary before recommencing the test for the next 3-min stage. After the final 3-min stage at 4.03 m·s^-1 was completed, the incline of the treadmill was increased by 2% every minute until volitional fatigue as reported previously (Edwards et al., 2003a). Criteria from British Association of Sport and Exercise Sciences (1997) were used to ascertain whether or not maximal aerobic power had been attained.

For the purposes of accurate assessment of the AT, the mean of VO₂ at both 1) the speed at which the production of lactate began to surpass its clearance (Noakes, 1998) and 2) the ventilatory threshold (VCO₂ /VO₂ measurements) were determined. AT was identified as the exercise induced VO₂ immediately preceding a 0.5 mmol·l^-1 increase in blood lactate concentration above the mean of two baseline values (Edwards et al., 2003a) and ventilatory threshold was identified by the V slope method, described by Beaver et al., 1986. These measures have been shown to be closely related using this protocol (Edwards et al., 2003a, Edwards and Cooke, 2004) and our experience in the laboratory support the combined approach for greater reliability in the measurement of AT (unpublished observations).

Blood sampling
Whole blood was sampled froma fingertip at rest prior to any testing. A small incision was made with a single use disposable lancet and capillary whole blood was collected into a 100 μl capillary tube (Becton Dickinson, New Jersey, USA). Thirty five l of whole blood were immediately analysed for lactate concentration (Analox GM7 Analyser, Analox Instruments, London, UK) at rest, and at the conclusion of each incremental stage in the anaerobic threshold element of the aerobic test.

Statistical analyses
Analysis of the repeated measurement on all primary dependent variables was considered in relation to the previous observations by members of this research team concerning appropriate statistical use for repeated measures experimental models (Morton, 2005). On this basis, a general linear model ANOVA (MINITAB Inc, State College, PA) was used to compare seasonal variations in physiological fitness measures. Post-hoc Tukey tests of honest significant difference were applied to examine where differences existed. Statistical significance was accepted when p < 0.05. All results are expressed as means ± unless otherwise stated.

The test-retest reliability of the cardiopulmonary test was determined by calculating limits of agreement between two tests (Bland and Altman, 1986). For both maximal aerobic power and anaerobic threshold, individual differences between tests 1 and 2 were plotted against the mean difference on each measurement. Upper and lower limits of agreement were calculated as the minimum standard deviation from the mean difference that incorporated all 20 test-retest data sets. These upper and lower limits were subsequently used to reflect the statistical distinction between 'trivial/technical' effects i.e. within the likely test-retest limits, and 'meaningful/biological' performance change i.e. change in excess of the limits of variability. The coefficient of variation (CV) was expressed as a percentage and calculated as: CV = 100 x SD_diff/X. The SD_diff indicated the standard deviation of the difference between the duplicate measurements, and X the mean reading of the measurements.

Preliminary data: test-retest reliability of cardiopulmonary fitness test measurements
The upper and lower limits of agreement represent the maximal variance that encompassed all 20 test-retest results of both VO₂ max and AT (Table 1).The results of the test-retest reliability evaluation demonstrated a coefficient of variation of 3.4% for VO₂ max and 4.6% for AT. The observed limits of agreement for VO₂ max and AT protocols were then applied to the professional soccer data to evaluate whether any changes between consecutive measurements exceeded the pre-defined test (non-biological) error.

Cardiopulmonary fitness
The mean VO₂ max for each test occasion was in excess of the 60 ml·kg^-1·min^-1 threshold (Figure 1) suggested by previous authors as a minimum threshold for elite soccer performers (Reilly et al., 2000; Stølen et al., 2005). There was variation between means across the consecutive test occasions; however, none of these changes were statistically significant. Although the PPS sample mean was consistently the lowest of the VO₂ max measurements this was not statistically significant (Figure 1). The three year consecutive measures of VO₂ max were subsequently plotted against the pre-defined limits of agreement but none of the changes between test occasions were in excess of the pre-identified maximal limits attributed to measurement error (Figure 2a).

The measurements of AT followed a similar pattern to VO₂ max, however, statistical evaluation demonstrated that MID2 (p = 0.04) and MID 3 (p = 0.03) were significantly higher than their respective PPS values (Figure 1). When plotted graphically utilising the pre-defined limits of agreement, the two largest gains in AT occurred between the pre-season and mid-season intervals (PPS2 - MID2 4.8% gain; PPS3 - MID3 6.8% gain) (Figure 2b). The largest negative change in AT occurred between MID3 and EOS3 (6.0% loss; p=0.04). As percentages of maximum, the highest at which AT was observed corresponded to MID2 (84.2%) and MID3 (85.2%) (Figure 1). The three PPS test occasions were the three lowest percentages of maximal aerobic power (PPS1: 80.5%, PPS2: 81.05%, and PPS3: 81.3%).

Countermovement jump (CMJ) performance tests
Mean 'best' CMJ performances improved significantly between PPS2 and MID2 (p = 0.03) and thereafter remained consistently higher than season one measurements (PPS1 and EOS1) until the final test occasion at EOS3 where the mean reverted to the magnitude of earlier observations (Figure 3a). The 20-s CMJ test showed a highly significant change in performance between PPS2 and MID2 (p = 0.007) and remained consistently elevated from this time irrespective of test occasion (Figure 3b).

Body mass and composition
Body mass did not change across the eight measurement occasions (Table 2). Mean body mass ranged from 77.8 kg ± 7.2 (EOS2) and 81.3 kg ± 9.9 (PPS1). Estimated percentage body fat also did not differ between test intervals and ranged from a low of 11.1% 2.7 (EOS3) and a high of 12.8% ± 3.0 (PPS2) (Table 2).

The main observation of this study was that mean VO₂ max remained relatively stable among a homogenous population of elite soccer players and, in contrast to AT, did not change with different phases of the off- and on-seasons. Despite occasional test-to-test changes in personnel, the significant differences observed in AT measurements compared with VO₂ max (in the same players) clearly identifies AT as the more sensitive measure of training state. This observation confirms our previous observations from single test off- and on-season (Edwards et al., 2003a). Although VO₂ max and AT responded similarly over the duration of the study, the magnitude of change according to our criteria was only of biological meaning in AT and these observations are supported by several generic laboratory investigations reporting increases in measures of the anaerobic threshold without parallel changes in VO₂ max (Denis et al., 1982; Edwards et al., 2003a; Henritze et al., 1985). Figure 2 illustrates the observation that AT demonstrated both significant changes and these were also in excess of the pre-set limits we identified as estimates of test-to-test measurement error. Our finding also supports the use of this technique to infer an element of control in elite athlete studies where it is not always possible to include a comparative control group. This may encourage others to report data from elite populations which would be of assistance to the further understanding of high quality performances across different sports.

Interestingly, the majority of mean performance changes in physiological and anthropometric variables occurred mid-season while there was a noticeable decline in some mean performance indicators towards the end of the three competitive seasons.

The observation of improved mid-season physiological performances is probably due to a number of inter-related factors. The typical English full-season is approximately 9-months and the mid-season measurements were taken after 5-months. At this stage, players involved in all games had usually completed ~25 competitive matches with a full season comprising ~55 matches (including all competitions). On this basis, the congestion of matches in the latter stages of the season may have negatively contributed to the test performances. In addition, the accumulative effect of minor injuries, reduced opportunities for aerobic conditioning training, and general fatigue could all be factors reducing test performances at that test occasion.

The mean VO₂ max test performance across all eight test intervals (61.6 ± 0.6 ml·kg^-1·min^-1) confirms earlier comments suggesting 60 ml·kg^-1·min^-1 as a minimal 'elite' threshold (Helgerud et al., 2001; Reilly et al., 2000). However, the stability of this measure in comparison to AT across different stages of off- and on-seasons questions the relevance of repeated measurement of VO₂ max beyond the broad identification of aerobic capability in the tested cohort. In support of this statement, studies of soccer players have not tended to differentiate between adult players of differing standard (e.g. Brewer and Davis, 1992; Edwards et al., 2003b), nor been correlated with high-intensity exercise during match-play (Krustrup et al., 2003), nor previously identified change between the on- and off-season in elite adult soccer players (Edwards et al., 2003a). This is consistent with the observation of a high genetic contribution to VO₂ max performance, thus probably restricting the potential for improvement (Bouchard et al., 1992), especially in elite athletes with already well-developed exercise capacities. Nevertheless, a previous study demonstrated significant improvements in the VO₂ max test performances of elite youth players as a consequence of specific aerobic training (Helgerud et al., 2001) but it is likely that those gains are probably, in part, attributable to maturation.

The percentages of VO₂ max at which AT occurred, which although reasonably high, are in agreement with previous observations (Edwards et al., 2003a, Helgerud et al., 2001; McMillan et al., 2005). As such, it is likely that there remains greater scope for AT to change in accordance with seasonal variations in the fitness of elite adult male players. A recent study of lactate thresholds in professional youth players (McMillan et al., 2005) also showed this was the case in that group. However, the test-retest reliability element of this study conducted among the student volunteers demonstrated a greater CV for AT (4.6%) compared with VO₂ max (3.4%) indicating that although AT is useful, it remains appreciably more complex to accurately quantify. Although the variability of the measurements in our study were reasonably small, the subjective intra-observer and inter-observer reliability and reproducibility of ventilatory and lactate thresholds has previously been questioned (Gladden et al., 1985; Yeh et al., 1983). This is due to the complexities of identifying a sudden advance in lactate production when more commonly, lactate increases proportionally with the increased exercise intensity or power output (Allen et al., 1985; Myers and Ashley, 1997). Therefore, a combined assessment incorporating ventilatory variables to produce a mean VO₂ at AT is, in our opinion, preferable so as to obtain a more accurate measurement which can be observed via the relatively low CV (4.6%) reported in this study.

It seems likely that, although both AT and VO₂ max are related as measures of cardiopulmonary fitness, the two measurements are probably limited by different physiological mechanisms (Myers and Ashley, 1997). Considerable debate exists on the mechanisms limiting VO₂ max test performance (Bassett and Howley, 2000; Noakes, 1998); however, it is beyond the scope of this study to examine the causal factors limiting both VO₂ max and AT.

The ability to achieve a high standard of jumping performance is clearly related to activities during soccer match-play (Reilly and Thomas, 1976) and a previous study reported a close relationship between vertical jump height and performance in a soccer league (Arnason et al., 2004). The CMJ performances in this study (Figures 3a and 3b) were less than those previously reported in elite male volleyball players (CMJ: 54.3 ± 4.6cm) (Forthomme et al., 2005) but compared reasonably well with results from soccer players reported using similar protocols (43.5 - 61.0cm) (Stølen et al., 2005). The performances were well maintained over the 20-s trials (mean of eight trials: 46.0 ± 0.3cm) and qualitative observations from our laboratory support the use of CMJ tests in examining inter-individual jumping characteristics which showed clear differences between junior performers and the elite 1^st team squad (unpublished observations). The 1^st team squad improved CMJ significantly from the 1^st season (p < 0.05) and maintained this stable performance thereafter. This is probably due to a training effect which was most evident in the 20-s test whereby contraction of the antagonist muscle groups was probably reduced, allowing for a greater coordination of these muscle groups so as to produce a more powerful and effective series of jumps (Schmidtbleicher et al., 1988).

Anthropometrics are also important in preparing for optimal performance in soccer (Reilly et al., 2000) as excessive adipose tissue acts as dead weight in activities where the body mass must be repeatedly lifted against gravity during locomotion and jumping (Reilly, 1996). Although the examination of anthropometrics was clearly basic in this study, previous estimates of fat values have ranged between 7.9% and 19.3% (White et al., 1988), while ~11% seems broadly normal for elite players (Ostojic, 2003, Rienzi et al., 2000, Rico-Sanz, 1998). The percentage body fat of players in our investigations are similar to those reported elsewhere (mean of eight tests: 11.8% in this study); however, the method of collection in this study was BIA and the greater technical error common to this method may explain why significant associations were not evident between % fat and the main outcome measures.

Reilly (1990) also reported an overall mean body mass of 74.0 ± 1.6 kg for nine English professional squads and the eight mean body masses in this study are higher than those; however, this may in part be attributable to the greater mean height of our players (1.78 ± 0.26m) while probably also reflecting gradual changes in anthropometrics of professional players between ~1990 - 2004.

• Maximal aerobic power remains fairly stable across inter- and intra-season measurements.
• Anaerobic threshold appears more sensitive of training state confirming our earlier observations.
• The professional players tended to attain optimal performances at the mid-season interval over the 3 seasons, presumably prior to the development of accumulative fatigue.