The distribution of rest periods affects performance and adaptations of energy metabolism induced by high-intensity training in human muscle
J. PARRA,1 J.A. CADEFAU,1 G. RODAS,2 N. AMIGÓ 1 and R. CUSSÓ 1
1 Departament of Physiological Sciences I, Institut d'Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Faculty of Medicine, University of Barcelona, Spain
2 High Performance Sports Research Centre (CEARE), Secretaria General de l'Esport, Barcelona, Spain.
The effect of the distribution of rest periods on the efficacy of interval sprint training is analysed. Ten male subjects, divided at random into two groups, performed distinct incremental sprint training protocols, in which the muscle load was the same (14 sessions), but the distribution of rest periods was varied. The `short programme' group (SP) trained every day for 2 weeks, while the `long programme' group (LP) trained over a 6-week period with a 2-day rest period following each training session. The volunteers performed a 30-s supramaximal cycling test on a cycle ergometer before and after training. Muscle biopsies were obtained from the vastus lateralis before and after each test to examine metabolites and enzyme activities. Both training programmes led to a marked increase (all signi®cant, P < 0.05) in enzymatic activities related to glycolysis (phosphofructokinase ± SP 107%, LP 68% and aldolase ± SP 46%, LP 28%) and aerobic metabolism (citrate synthase ± SP 38%, LP 28.4% and 3-hydroxyacyl-CoA dehydrogenase ± SP 60%, LP 38.7%). However, the activity of creatine kinase (44%), pyruvate kinase (35%) and lactate dehydrogenase (45%) rose signi®cantly (P < 0.05) only in SP. At the end of the training programme, SP had suffered a signi®cant decrease in anaerobic ATP consumption per gram muscle (P < 0.05) and glycogen degradation (P < 0.05) during the post-training test, and failed to improve performance. In contrast, LP showed a marked improvement in performance (P < 0.05) although without a signi®cant increase in anaerobic ATP consumption, glycolysis or glycogenolysis rate. These results indicate that high-intensity cycling training in 14 sessions improves enzyme activities of anaerobic and aerobic metabolism. These changes are affected by the distribution of rest periods, hence shorter rest periods produce larger increase in pyruvate kinase, creatine kinase and lactate dehydrogenase. However, performance did not improve in a short training programme that did not include days for recovery, which suggests that muscle ®bres suffer fatigue or injury.
Keywords anaerobic exercise, enzyme activities, glycogen, glycolysis, lactate, recovery, skeletal muscle metabolism, sprint training.
Received 12 July 1997, accepted 16 March 2000
Muscle adaptation owing to exercise or physical training seems to be correlated with the amount, intensity, distribution and duration of muscle loads (Dudley et al. 1982). The combination of these factors is especially important when interval high-intensity training is designed because biochemical responses depend on the protocol of contractile activity to which the muscle is subjected (summarized in MacDougall et al. 1998). Hence, a direct relationship between adaptations and the components of the sprint training has been hard to find. However, certain desirable biochemical adapta- tions and sprint performance improvement seem to be associated with high-intensity training. Increases in enzymatic activities related to glycolysis, including phosphofructokinase, lactate dehydrogenase or glycogen phosphorylase and changes in metabolite concentration, including phosphocreatine or glycogen, have been recorded, although the extent of these changes varies from one sprint training programme to another (Thorstensson et al. 1975, Costill et al. 1979, Roberts et al. 1982, Cadefau et al. 1990, Linossier et al. 1993, 1997, Dawson et al. 1998, MacDougall et al. 1998).
The duration of each bout affects the adaptations induced by interval high-intensity training, as described by Costill et al. (1979), comparing 6 s in the left leg with 30 s in the right leg of maximal isokinetic exercise of the same subject. Bouts lasting less than 10 s are considered more anaerobically dependent (Hellsten- Westing et al. 1993, Linossier et al. 1993) than longer bouts, which are more demanding and during which power output decreases before the end (Bogdanis et al. 1995, 1996).
The recovery periods between bouts are also deci- sive and the time ratio between recovery and exercise phases has to be taken into account (Linossier et al. 1997). Intense and brief muscle loads with long recovery periods are proposed to induce an adaptative response in phosphocreatine metabolism (Thor- stensson et al. 1975), while an increased training load has no effect on this process, but seems to produce a greater adaptative response in lactate metabolism (Roberts et al. 1982, Cadefau et al. 1990). The distri- bution of rest periods between days of training is usually less carefully planned than the ratio of recovery and work in each session and its effects on the adap- tations induced by sprint training have not been studied. The rest periods between sessions prevent fatigue, which could appear when rest periods are insuf®cient and/or muscle load is exhausting.
In order to analyse the effect of rest distribution on muscle adaptations, we designed two high-intensity training cycle protocols with identical daily muscle loads but different distribution of rest periods. Enzy- matic activities related to glycolysis, glycogen and creatine metabolism and aerobic metabolism were measured together with adenine nucleotides, glycolytic intermediates and creatine concentrations. We also evaluated the effect of the two training protocols on performance and muscle metabolic response by a 30-s all-out cycling test before and after training.
Ten healthy male student volunteers agreed to take part in this study. Their age, height and body mass were (mean +/- SD) 23.6 +/- 2.4 years, 171.1 +/- 3.4 cm and 70.2 +/- 4.8 kg, respectively. All were active, but none was currently participating in a regular training programme. During the experiment, all volunteers stopped their normal physical activity and only exer- cised within the experiment. Before the commence- ment of the experiment the volunteers underwent a medical check-up to verify that they were healthy and ®t. They were divided at random into two groups called `short programme' (SP) and `long programme' (LP).
The experiment was conducted in accordance with the code of ethics of the World Medical Association (Declaration of Helsinki) and approval was given by the Ethical Committee of Human Experimentation from the Pi i Sunyer Biomedicine Research Institute of Barcelona (Hospital Clinic i Provincial ± University of Barcelona). All subjects were informed before recruit- ment about the purpose of the study, known risks, and possible hazards associated with the experimental protocol and each gave his written consent.
In order to evaluate the anaerobic capacity of volun- teers and possible improvement owing to training, subjects were required to perform a supramaximal cycling test (30 s) the day before the initiation of training and 48 h after ®nishing 14 training sessions. The test was performed on a friction-loaded cycle ergometer (Monark 814 E, Varberg, Sweden). A microprocessor, interfaced with the cycle ergometer, counted ̄ywheel revolutions every second for 30 s of supramaximal cycling sprinting against a constant resistance of 0.075 kg ́ (kg body mass)±1. With the ̄ywheel progression per pedal revolution and the elapsed time, the following variables were calculated: peak power (the highest power output) and mean power (the average power output during the 30 s). The subjects were comfortably seated with feet secured to the pedals by toe clips. They were requested to pedal as fast as possible from the start and were encouraged to maintain maximum pedalling speed throughout the 30-s period.
Familiarization with the equipment, sprint cycling and testing procedures took place before the experiment started, until there was complete con®dence in reaching an all-out effort from a stationary start. Each group (SP and LP) participated in a different high-intensity programme designed to improve performance in high- intensity tests. These programmes involved the same 14 training sessions but with different duration of rest. The SP group trained every day for 2 weeks while the LP group trained for 6 weeks resting for 2 days between each session. The sessions comprised a number of warm-up repetitions of 15 s maximal cycling with 45 s rest-periods and a number of training repetitions of 30 s maximal cycling with 12 min rest-periods. The number of repetitions was modi®ed and the total muscle load increased during training. The first three sessions comprised of two bouts of 15 s sprints and two bouts of 30 s supramaximal cycling sprints. In the following sessions, the number of 15- and 30-s bouts were increased by one every two training sessions. The last three sessions consisted of seven bouts of 15 s and seven bouts of 30 s. As in the performance tests, subjects were instructed to remain seated during the cycle sprints in the training period. The ̄ywheel tension was set at 0.075 kg (kg body mass)±1 and remained constant for the duration of the training programme. The maximum number of pedal revolu- tions reached by each volunteer in every 30-s bout was recorded. All subjects were highly motivated and verbally encouraged during training and instructed to cycle with maximum effort in every session.
The needle biopsy technique was used to sample muscle tissue. Muscle biopsy samples (30±50 mg) were taken under local anaesthesia (mepivacaine 2%) from the mid-region of the quadriceps femoris muscle (vastus lateralis) from both legs, 15 cm above the top edge of the patella on the ®rst day and 5 cm above it next day (48 h after ®nishing training). On the day of the performance test, volunteers reported to the laboratory at least 3 h after their last meal. After a light warm-up, they sat quietly on an examination couch while small incisions were made in both legs through the skin and fascia and the ®rst muscle biopsy was obtained from left leg (rest). Subjects then performed the test and the second biopsy from right leg (30 s) was taken imme- diately after, while they were still seated on the cycle ergometer. The same protocol was performed 1 day before (pre) and 2 days after (post) training. The samples were directly frozen, removed from the biopsy needles under liquid nitrogen and stored at ±80 °C until they were lyophilized and analysed.
Freeze-dried samples were dissected free of blood and connective tissue and powdered. A part (20 mg) of the dry tissue was treated with 0.5 M HClO4 and centrifuged at 13 000 ́ g at 4 °C for 15 min. The supernatant was neutralized with 2.1 M KHCO3. The neutralized extract was assayed for phosphocreatine (PCr), ATP, creatine (Cr), free glucose, glucose 1-phosphate (G-1-P), glucose 6-phosphate (G-6-P), fructose 6-phosphate (F-6-P), fructose 1,6-bisphosphate (F-1,6-P2), pyruvate (Pyr) and lactate (Lac). All muscle metabolites were assayed enzy- matically by ̄uorometric analysis (Lowry & Passonneau 1972). Glycogen concentration was measured both in the neutralized extract and in the pellet, by prior 1 M HCl hydrolysis extraction. Then the free glucose produced was determined by the method described above. From the neutralized extraction, IMP, ATP, ADP and AMP were measured using the HPLC method (Ingebretsen et al. 1982). Muscle metabolite concentrations were adjusted to the individual mean total creatine (PCr + Cr) because this mean should be kept constant during exer- cise (Harris et al. 1976). The adjustment to total creatine content enabled any variability in solid non-muscle constituents of the biopsies to be corrected.
For the enzymatic analyses, a portion of the muscle biopsies (10 mg) taken before the tests was homogen- ized in 30 volumes of ice-cooled extraction medium. The extraction medium contained 50 mM HCl±Tris (pH7), 4mM EDTA, 50mM KF and 30mM b-mercaptoethanol. The preparation was centrifuged at 15 000 ́ g at 4 °C for 15 min. The following activities were immediately measured in the supernatant: glycogen synthase (GS), glycogen phosphorylase (GPh), creatine kinase (CK), phosphofructokinase (PFK), aldolase (ALD), lactate dehydrogenase (LDH) and pyruvate kinase (PK), as described in Cadefau et al. (1990); hexokinase (HK), citrate synthase (CS), phos- phoglucoseisomerase (PGI) and 3-hydroxyacyl-CoA dehydrogenase (HAD) as described in Essen- Gustavsson & Henriksson (1984) and myokinase (MK) as described in Oliver (1955).
Values of pH in the muscle, before and after the test, were calculated from changes in lactate and pyruvate concentration [expressed as mmol (kg dry tissue)±1] as reported by Sahlin (1978):
pH = 7.06 – 0.00413 x (/Lac/ + /Pyr/)
ATP consumption [mmol (kg dry tissue)±1] during the tests, before and after training, was also calculated using a speci®c equation (Katz et al. 1986), where increments in ATP, ADP, PCr, Lac and Pyr were obtained from each value before and after the 30 s all-out cycling test.
ATP consumption = 2(-AhATP) - AADP - APCr + 1.5ª(Lac) + 1.5(APyr)
No corrections were made for lactate or pyruvate ef ̄ux during sprints or for anaerobic-produced ATP as a result of pyruvate oxidation.
The ̄ux in glycogenolytic and glycolytic pathways [mmol glucosyl units (kg dry tissue)±1] was calculated from a combination of the variation in concentration of several metabolites, as described by Spriet et al. (1987):
Glycogenolytic rate = AG-1-P + AG-6-P + AF-6-P + 0.5(ALac + APyr)
Glycolytic rate = 0.5(ALac + APyr)
Differences in the same groups before and after training were analysed by non-parametric Wilcoxon test for paired values. Differences between the two groups were evaluated by non-parametric Mann±Whitney test for unpaired values. Differences were considered signi®cant at P < 0.05 and values were expressed as means +/- SD.
Muscle metabolites in the pre-training test
One day before the beginning of the training period, subjects undertook a supramaximal test (pre-training test). Several metabolites were measured from biop- sies taken immediately before and after the test for both groups (Tables 1, 2). Before training, neither group showed any signi®cant differences in metabo- lite concentration at rest and the changes produced owing to the test were similar. ATP and PCr concentration decreased signi®cantly in both groups (in all cases P < 0.05). IMP increased signi®cantly during the test in both groups. Despite the extent of changes in total adenine nucleotide (TAN = ATP + ADP + AMP), the amount of TAN + IMP remained unchanged after 30 s of supramaximal cycling. Glycogen concentration decreased signi®- cantly in both groups (P < 0.05) and to a similar extent (SP 29%; LP 26%). As a result of glycogen degradation, G-1-P concentration increased signi®- cantly, about 3±4-fold (P < 0.01 both groups). G-6-P concentration increased more than 10-fold (both groups P < 0.01). F-6-P increased (both groups P < 0.01) in a similar manner to G-6-P. Although glucose concentration increased in both groups, no signi®cant differences were found. The increase in lactate concentration was higher than 10-fold in both groups and the increase in pyruvate reached 5-fold (P < 0.01 in all cases). As lactate and pyruvate concentrations increased, the muscle ®bres became acidic. The calculated pH fell sharply after the 30-s- test in both groups.
Enzymatic adaptations to sprint training
There were no signi®cant differences in enzymatic activities between the SP and LP groups before the start of the training programmes. However several enzy- matic activities were modi®ed in response to training (Table 3). Myokinase, glycogen synthase and glycogen phosphorylase did not vary significantly in either group, but the percentage of change was different between groups (P < 0.05).
Creatine kinase activity showed signi®cant increase in the SP group (44%, P < 0.05), but only a slight variation in the LP group (9%). Pyruvate kinase and lactate dehydrogenase increased signi®cantly (P < 0.05) and to a considerable extent (35 and 45%, respectively) in the SP group, but not in the LP group. All these enzyme activities (CK, PK and LDH) showed a different (P < 0.05) percentage of change between both SP and LP group.
Phosphofructokinase, aldolase, citrate synthase and 3-hydroxyacyl-CoA dehydrogenase increased in both groups (all cases P < 0.05). Although for the PFK and HAD activities, the increases were more extended in SP group (P < 0.05).
Hexokinase and phosphoglucose isomerase activities remained unchanged or showed only slight variations.
Muscle metabolites in the post-training test
After training and before the post-training test, both groups showed similar metabolite concentration at rest. However, the SP training caused a slight variation in total creatine (14%), together with increases in PCr (39%, P < 0.05) and glycogen (32%, P < 0.05) concentration, while only glycogen (30%, P < 0.05) increased in the LP group. In contrast, ATP and TAN concentration showed a slight change after 14 training sessions in both groups (not signi®cant).
After the post-training test, IMP and AMP varied as during the pre-training test in both groups. Whereas the change was smaller and not signi®cant in the SP group, the increase in the LP group was signi®cant (P < 0.05). In both cases, the variation was less pronounced than that previously produced in the pre-training test. ATP and PCr concentration changed in both groups, but in this second test only PCr decreased signi®cantly (P < 0.01, in both groups). Glycogen concentration decreased signi®cantly in both groups (P < 0.05). The increase in lactate concentration after the post-training test was more than 9-fold in both groups (P < 0.01). Long programme group attained values similar to those in the pre-training test, while the SP group showed a slightly lower lactate concentration than in the pre- training test.
The test gave a different result for the two groups after training (Fig. 1). The LP group signi®cantly improved their maximum peak power (20%) and mean power (14%), while the variation in the same parameters of the SP group was smaller and not signi®cant (3 and 3%, respectively). The difference in performance between SP and LP groups was signi®cant (P < 0.05).
The SP group improved the maximum peak power during the ®rst session of training, reaching 10% on the 10th day of training (data not shown). However, in the last three sessions, their performance deteriorated to the same extent as in the post-training test.
Figure 1 Mean and peak power values of SP and LP group during the 30 s tests, pre- (open bars) and post-training (striped bars). Values are means SD for ®ve subjects each group. * Signi®cant difference (P < 0.05) between values pre- and post-training.
ATP consumption and glycogenolysis and glycolysis rates during test
ATP consumption and glycogenolysis and glycolysis rates during the pre-training test was similar in both groups, however, differences in those rates were observed after training (Table 4). Short programme showed a decrease in ATP consumption (16%, P < 0.05) as a consequence of a reduction in the glycolysis (18%, P < 0.05) and glycogenolysis (30%, P < 0.05) rates. Long programme showed a slight but non-signi®cant variation in ATP consumption, prob- ably produced by an increase in the glycogenolytic rate (11%).
The SP group showed lower values of ATP consumption and glycogenolysis and glycolysis rates than the LP group. However only the glycogenolysis rate was signi®cantly different (P < 0.05) between groups after training.
Muscle metabolic response to the pre-training test
The muscle concentration of metabolites in the volunteers was within the same range reported in the literature (Bogdanis et al. 1995). Likewise, muscle metabolite response during the pre-training 30 s all-out test was in line with that described previously for normal active volunteers (Nevill et al. 1989, Bogdanis et al. 1995), with no differences being recorded between groups. However, it is worth noting the variability in glycolytic intermediate concentrations depending upon the experimental protocol (with or without warm-up) and the effect of previous training status of the volunteers.
The pre-training test led to a sharp reduction in ATP and PCr concentration after 30 s supramaximal cycling, which is consistent with other ®ndings (Stathis et al. 1994, Bogdanis et al. 1995, 1996). A large increase in IMP concentration was found in muscle after 30 s of supramaximal exercise in both groups. This increase in IMP muscle concentration is a consequence of a greater requirement of the adenine nucleotide metabolism, often described after high intensity exercise (Sahlin et al. 1978, Stathis et al. 1994).
Although changes in ATP, ADP and IMP were considerable, the addition of TAN + IMP remained constant in both groups after the test, which suggests that IMP was not dephosphorylated during the 30 s of sprint cycling and there was no loss of purine nucleo- tides.
During the 30-s test, approximately 30% of the glycogen was broken down. We found an average rate of 1.6 mmol glucosyl units (kg dry muscle)±1 s±1 after 30 s of sprint cycling, while Gaitanos et al. (1993) reported a glycolysis rate of 2.2 mmol glucosyl units (kg dry muscle)±1 s±1 after the ®rst 6 s and Jacobs et al. (1983) reported that after 10 s of maximal cycling the production of lactate was 60% of the total lactate produced during 30 s. All these results suggest that the expenditure of glucose via anaerobic glycolysis is not constant during a 30-s all-out cycling test.
Biochemical changes caused by sprint training
Sprint training produced changes in muscle metabolite concentration which seemed unaffected by rest period distribution, as variations were similar in both groups. Resting values of glycogen concentration increased after training (30%, P < 0.05) in both groups. Increases in glycogen storage after high-intensity training proto- cols has been described, although the amount depends on the programme design, as found elsewhere (Boobis et al. 1983, Cadefau et al. 1990).
TAN and ATP concentration at rest were lower (although not signi®cantly) after training than before in both groups. Such a reduction of TAN owing to high- intensity muscle contractile activity has been previously reported (Hellsten-Westing et al. 1993, Stathis et al. 1994). Moreover, this fall has been related to an insuf®cient resting-time between bouts for PCr resyn- thesis and to a parallel loss of muscle IMP by catabo- lism (Stathis et al. 1994). In our experiment, the recovery periods between bouts were 12 min, long enough for total PCr resynthesis (Bogdanis et al. 1995), but not long enough for the resynthesis of IMP to AMP. However, the later sessions probably generated sustained high IMP concentration because of the intensity of training (seven bouts of 30 s) and IMP could then be catabolized. Indeed, increased catabolism of IMP has been described after intense exercise repetition (Bangsbo et al. 1992). Thus, the recovery periods between bouts and the exercise intensity seem to be important in the adaptation of ATP and IMP metabolism.
Before training, enzyme activities were similar in both groups. However, after training, enzyme activities were clearly varied. Creatine kinase activity showed a signi®cant increase in the SP group. This enzyme usually shows only slight variations, probably because of its abundance (Cadefau et al. 1990). This study points out the possible importance of rest periods between sessions of training in order to produce an increase in the muscle CK activity.
In the glycolytic pathway, PFK and ALD activities increased in both groups. An increase in PFK activity is expected after sprint training, but its extent seems to depend on the training procedure (Cadefau et al. 1990, Linossier et al. 1993). PK and LDH activity increased signi®cantly only in the SP group. These data indicate that the more concentrated the protocol, the greater the changes in glycolytic enzymes (PFK, PK and LDH).
Aerobic metabolism improvement, represented by CS and HAD activities, is an unusual adaptation to sprint training. However, high production of lactate following repeated bouts might induce an aerobic adaptation by improving the metabolism of the excess of pyruvate through pyruvate dehydrogenase (MacDougall et al. 1998). If we consider this possibility, both programmes were intensive enough to induce aerobic adaptation, although shorter rest periods induced larger increases in HAD activity (P < 0.05).
It is of particular interest to note the lack of vari- ation in the HK activity of both groups, while increases of HK activity has been found after other sprint training protocols (Linossier et al. 1997, MacDougall et al. 1998). These authors described an increase in HK activity when recovery/work ratio was 8 (4 min of recovery between repetitions of 30-s bouts, MacDou- gall et al. 1998) or 11 (55 s of recovery between repe- titions of 5 s bouts, Linossier et al. 1997). Our sessions had a recovery/work ratio of 24 in both groups. Thus, this ratio of recovery periods between bouts could be involved in the adaptation of the external glucose use through HK.
Muscle metabolic response to the post-training test
Muscle metabolic response to the post-training test was different in the two groups. Glycogen consumption during the post-training test reached 25% of the rest concentration in the LP group but only 15% in the SP group. Consequently, glycolysis and glycogenolysis rates during the post-training tests were signi®cantly higher in LP group. However, part of this difference was a consequence of a signi®cant decrease in rates of the SP group. These suggestions are proposed when related to the weight of muscle and they could differ in the case when related to the whole body because the recruited muscle mass may be increased after training.
The increase in performance of the 30 s all-out test in LP group was associated with an increase in enzymatic activities related to muscle energy metabo- lism. The relationship between muscle enzymes and performance has been suggested by Linossier et al. (1997) and MacDougall et al. (1998). However, the lack of correlation between the improved performance (maximum power output appears at 5 s) in the 30-s test and the glycolytic rate averaged over the 30-s period is a question which invites further studies, specially focused on a better understanding of the muscle metabolism during the ®rst few seconds of high-intensity exercise.
With shorter rest periods, the SP group consumed less glycogen and anaerobically generated ATP and produced less lactate during the post-training test than before training. In contrast, performance was similar to the pre-training value. Hence, we suggest a decreased involvement of anaerobic metabolism and an enhanced involvement of aerobic metabolism through the increase in CS and HAD activities.
However, the reduction in anaerobic ATP consumption and the failure to improve performance in the post-training 30 s test were unexpected, because changes of enzyme activities in SP group were greater than in the LP group. A possible explanation for this might be that their muscles were suffering fatigue or injury (Allemeier et al. 1994). Repeated exercise at high intensity may cause a loss of K+ from the contracting muscle (McKenna et al. 1993) and a decrease in the gradient regulated by muscle Na+-K+-ATPase has been related to fatigue (Lindinger & Heigenhauser 1991). This K+ homeostasis in exercising humans has been connected with rest duration (Kowalchuk et al. 1988). Another critical point to explain fatigue is the intra- cellular Ca2+ exchange (Williams & Klug 1995). Possible training-induced alterations in Ca2+ uptake and alterations in the sarcoplasmic reticulum could be rest- dependent, given the different responses of the two groups. More than 48 h of recovery, after the last training session, had in all probability avoided part of the remaining negative effects of the last training session.
It can be concluded that sprint cycling training may produce major enzyme activity changes in human muscle such as PFK, ALD, CS and HAD activities, together with increases in glycogen concentration. However, part of these modi®cations depend on rest distribution. We suggest that some adaptations are better induced by shorter rest periods, such as increases in PFK, HAD, PK and CK activities or in PCr concentration, although LDH activity is the most sensitive to rest distribution.
Furthermore, in spite of the fact that shorter rest periods during high-intensity training induce greater biochemical adaptation in human muscle than a more restful training programme with the same muscle load, less rest hinders the improvement of short-time performance, probably because of fatigue.
This study was supported by grants from the Direcció General de l'Esport (Generalitat de Catalunya 1993), CICYT SAF95-1045 and FISS 95/0994 (Ministerio de Sanidad). The authors wish to give special thanks to the volunteers of the experiment and to Carmen Andrade for her skillful technical assistance.