Key words
lactate - cardiovascular risk - blood lipids - glykolysis
Introduction
Both exercise physiologists and clinicians have studied lactate metabolism for
decades. Once assumed to be simply a byproduct of anaerobic energy metabolism, which
might act as a substrate when conditions are aerobic, lactate actually is a
signaling molecule with relation to metabolic regulation [1].
Clinically, blood lactate levels have been interpreted to point towards insufficient
oxygen delivery (e. g., in sepsis or cardiac or bowel ischemia) [2]
[3]
[4] to tissues and other pathologies [5]. Moreover, blood lactate levels at rest
(Lacrest ) and during exercise are also used as a marker of primary
mitochondrial dysfunction and hence, lack of oxidative capacity of the cells, due to
hereditary enzyme defects [6] or even deconditioning
[7].
Recent work based on data gathered during the Atherosclerosis Risk in Communities
(ARIC) study has demonstrated an association between Lacrest and
cardiovascular outcomes [8], development of diabetes
mellitus [9]
[10],
hypertension [11] und carotid atherosclerosis [12], which points towards a clinical significance of
this parameter, possibly through its link to general energy metabolism.
As unfavorable metabolic changes manifesting at a very young age might already impact
on long-term cardiovascular risk [13], the aim of this
article, therefore, is to investigate the relationship between Lacrest
and cardiovascular risk with respect to blood lipids, physical fitness and in a
relatively young study group of healthy patient-athletes.
Knowledge about both, the distribution of Lacrest in a healthy population
and its intraindividual variation in repeated measurements is an indispensable
prerequisite to assess whether Lacrest is suitable to point towards an
increased cardiovascular risk on an individual level. To our knowledge, there is a
scarcity of data concerning theses parameters. Therefore, and as a preliminary step
for the key study aim, the distribution of the Lacrest and the relation
between inter- and intraindividual variation will be studied based on a large
dataset of healthy patient-athletes.
Materials and Methods
The local hospitals ethics committee has approved the study. In a retrospective
analysis of our database, 9051 datasets of 5575 individual patients
(“patient-athletes”, for subject characteristics, see [Table 1]) who visited our outpatient department in
between 2007 and 2016 for lactate-based performance testing were identified that
qualified to be further studied. The term “patient-athlete” is used
because the subjects were either competitive or leisure-time athletes (from
strength/speed or mainly endurance determined type of sports) or were
performing the lactate-based performance test to establish regular physical activity
(sedentary group). The subjects were allocated to their predominant type of physical
adaptation by their self-reported primary sport during at least one year before the
examination. This assignment was taken from the criteria in Olympic sports medicine
[14]. These datasets were used to describe the
distribution of Lacrest.
Table 1a Subject characteristics for samples to
create reference range (n=9051).
|
a. Sex
|
6087 male/2964 female
|
|
Age (yrs.)
|
24.6±14.7
|
|
BMI (kg/m²)
|
22.1±3.8
|
|
Exercise type
(“Endurance”/“Strength/Speed”/“Sedentary”)
|
4670/3624/757
|
|
VO2peak (ml/min/kg)
|
49.6±10
|
|
b. Sex
|
901 male/414 female
|
|
Age (yrs.)
|
37.2±17.2
|
|
BMI (kg/m²)
|
22.9±3.9
|
|
Predominant type of exercise
(“Endurance”/“Strength/Speed”/“Sedentary”)
|
704/329/282
|
|
VO2peak (ml/min/kg)
|
43.5±11
|
|
Estimated 10-year cardiovascular risk
|
0.035±0.05 (median 0.015)
|
|
HDL-cholesterol (mg/dl)
|
59.0±14.7
|
|
LDL-cholesterol (mg/dl)
|
127.4±36.4
|
|
Serum glucose (mg/dl)
|
87.1±16.4
|
|
Serum triglycerides (mg/dl)
|
108.6±71.5
|
|
Lacrest (mmol/ml)
|
1.17±0.31
|
BMI: body-mass-index, Lacrest: Lactate at resting conditions.
For a subgroup of 836 of these subjects, replicate measurements of
Lacrest were available. This subgroup was analyzed to determine
intra- to interindividual variation of Lacrest (see below).
The key study sample consists of the 1315 subjects for whom Lacrest as
well as clinical chemistry data, anthropometric data and results from an exercise
stress test are available. For these subjects, 10-year cardiovascular risk was
calculated using the Framingham score [15]. Subjects
with a Lacrest outside the normal range for Lacrest as
described below were not excluded from the analysis.
Lactate analysis and clinical chemistry
For lactate analysis, samples were taken at rest before routine ergometer testing
using capillary sampling from the hyperemized earlobe. Resting conditions were
uncontrolled but performing the resting ECG and echocardiography before the
ergometer procedure allowed for a resting period of at least 30 minutes
before capillary sampling. During this period, the blood pressure was determined
in a supine position using the Riva-Rocci-method on the left arm. Lactate was
measured from these hemolyzed whole-blood samples on either Eppendorf EBIOS plus
(Wesseling-Berzdorf, Germany) or EKF Biosen analyzers (Barleben, Germany) within
a maximum period of up to one hour after withdrawal enzymatic-amperometrically.
We calibrated the analyzers before each measurement cycle; hence, the impact of
ambient temperature variation in the laboratory is negligible. All lactate
analyzers used in our laboratory undergo interlaboratory comparison as requested
by German authorities for hospital laboratory analysis (RiLi-BÄK). All
other laboratory parameters were obtained from venous blood samples taken before
exercise stress testing from an antecubital vein and were either analyzed in our
own (full blood count, HDL, LDL) or our hospital’s central laboratory
(total cholesterol, triglycerides, glucose) by conventional methods. Not all
measures were determined in all subjects (see [Table
1]).
Performance testing
All subjects underwent a graded exercise test on either a bicycle ergometer or a
treadmill until subjective exhaustion. was estimated from maximal
performance using the ACSM equations. [16].
Statistics
All statistical analyses were performed using R [17]. Reference values for Lacrest were determined using
parametric descriptive statistics. To determine whether individual reference
ranges for Lacrest are applicable and to obtain a reference to
determine effect size, intra- versus inter-subject variability of the respective
measures was expressed as ratio r, which was introduced by Eugene Harris [18]. Variance analysis for calculation of r was
performed using a linear mixed-effects model fitted by maximum likelihood
(“nlme” package, [19]) on a subset
of 836 subjects for whom multiple measurements were available for analysis; we
analyzed all available replicates).
The Framingham risk score as an estimator of cardiovascular risk [15]
[20] was
calculated from all available data using the
“framinghamriskequation” R package [21] using the “CVD” preset. The Framingham risk score
estimates the 10-year-risk for the occurrence of a cardiovascular event
(coronary artery disease, stroke, congestive heart failure and peripheral
vascular disease) from input variables such as sex and age as well as the
lipoprotein profile, the systolic blood pressure, and the smoker status and
presence of diabetes. The parameter “left ventricular hypertrophy in
ECG” was omitted in calculating cardiovascular risk because left
ventricular hypertrophy in our cohort was most probably a benign adaptation to
training
A correlation analysis was used to display the raw association between
Lacrest and the estimated cardiovascular risk. Cohen’s
definition of effect size for correlation coefficients [22] was used to describe the effect size of observed correlation
coefficient. To account factors confounding the association between
Lacrest and the cardiovascular risk, and to investigate the
association between Lacrest and other established predictors of
cardiovascular risk, linear regression modelling was used. Because of the large
sample size, a k-fold cross validation method (with k=10) could be used
for all linear modelling and correlation analyses [23]. We performed all analyses using the “caret”
package in R [24]. The descriptive correlation
coefficients for the correlation between Lacrest and the
cardiovascular risk in the subgroups displayed in [Fig.
1b–d] were calculated without k-fold-validation. An alpha
level of 0.05 was accepted for statistical significance.
Fig. 1 Association between Lacrest and cardiovascular
risk. Red line=linear fit with confidence interval (standard
error) for the slope. a: whole study sample (n=1315);
b: endurance-type athletes (n=706); c:
strength/speed -type athletes (n=329), d: sedentary
(n=284). *: p<0.05.
Results
Descriptive statistics and intra- to interindividual variation of
Lacrest
The descriptive statistics for the characteristics of the subjects
(n=9051 samples; 6087 taken from male and 2964 taken from female
subjects, mean age 24.6±14.7 years) for each sample taken for this
analysis is displayed in [Table 1a].
The mean lactate value at rest was
1.16±0.29 mmol·l-1. Based on these values
of an apparently healthy group of subjects, the normal range for resting lactate
concentration, expressed as mean±2 SD, spans from 0.58 to
1.74 mmol·l-1. 3.3% of all lactate values
observed at rest were higher than this observed upper limit of normal in our
cohort.
Analysis of intra- to interindividual variation of Lacrest yielded an
r value of 3. The intra-individual standard deviation was
0.75 mmol·l-1, the interindividual standard
deviation was 0.25 mmol·l-1.
Association between Lacrest and 10-year cardiovascular
risk
For this analysis data from 1315 samples (414 female) with a mean age of
37.2±17.2 was analyzed (for further details see [Table 1b]). Of these, 704 samples were from
“endurance” adapted athletes (thereby were 312 leisure time
athletes predominantly performing endurance exercise, 140 road and mountain bike
cyclists, 105 track and field athletes, 83 triathletes, 34 Nordic skiers, and 30
from other endurance disciplines), 329 were from
“strength/speed” athletes (142 from leisure time
athletes predominantly performing strength training, 80 from soccer players, 19
from alpine skiers, 16 from handball players, 15 from track and field athletes
from sprint and throwing disciplines, 14 from wrestlers, 11 from tennis players
and 32 from other non-endurance disciplines) and 282 form those subjects
classified as “sedentary” or sports beginners.
The association between the estimated cardiovascular risk and Lacrest
is illustrated in [Fig. 1a]. Correlation analysis
yielded a “low” correlational effect (Spearman’s
rho=0.20, p<0.001) between Lacrest and the estimated
cardiovascular risk. Analyses of the association between the estimated
cardiovascular risk and Lacrest within the subgroups with respect to
the different types of predominant training adaptation are displayed in [Fig. 1b–d].
The results of the regression analysis investigating the contribution of
Lacrest to predicting cardiovascular risk with established risk
markers are presented in [Table 2]. In this
linear regression model Lacrest is not a significant predictor of the
estimated 10-year cardiovascular risk in our study sample. The association
between Lacrest and metabolic parameters determining cardiovascular
risk and measures of physical activity, respectively, are illustrated in [Tables 3] and [4].
Lacrest is statistically associated with sex, serum glucose,
triglycerides, and HDL-cholesterol but not LDL-cholesterol. Moreover,
Lacrest is associated with the predominant type of exercise that
is conducted by the subject but not aerobic performance per se. Both models
feature a “small” effect size of the predictors for explaining
Lacrest.
Table 2 Results of linear regression analysis (k-fold
validation, k=10) for explanatory parameters for
cardiovascular risk.
|
Estimate
|
Std. Error
|
t value
|
P
|
|
Intercept
|
−12.7900
|
9.84600
|
−1.299
|
0.194187
|
|
Age (years)
|
2.0090
|
0.05537
|
36.288
|
<0.0001
|
|
HDL-cholesterol (mg/dl)
|
−0.7782
|
0.06033
|
−12.900
|
<0.0001
|
|
LDL-cholesterol (mg/dl)
|
0.1928
|
0.02344
|
8.226
|
<0.0001
|
|
Sex=female
|
−12.7700
|
2.01300
|
−6.344
|
<0.0001
|
|
Serum triglycerides (mg/dl)
|
0.0505
|
0.01142
|
4.421
|
<0.0001
|
|
VO2max (ml/min/kg)
|
−0.3851
|
0.10130
|
−3.801
|
<0.001
|
|
Serum glucose (mg/dl)
|
0.1677
|
0.04655
|
3.603
|
<0.001
|
|
Predominant
exercise=”Strength/Speed”
|
5.7890
|
2.00300
|
2.891
|
0.0039
|
|
BMI (kg/m²)
|
−0.2212
|
0.14410
|
−1.535
|
0.1250
|
|
Predominant exercise=”Sedentary”
|
−3.3200
|
2.45900
|
−1.350
|
0.1771
|
|
Lacrest (mmol/l)
|
−0.4694
|
2.49400
|
−0.188
|
0.8507
|
|
R² for final model=0.7090968. RMSE of
cross-validation: 0.0268284. f²=2.44
|
|
|
|
|
Cardiovascular risk: 10-year-cardiovascular-risk estimated by the
Framingham score. BMI: body-mass-index, Lacrest: Lactate at
resting conditions.
Table 3 Results of linear regression analysis (k-fold
validation, k=10) of explanatory parameters for
Lacrest (metabolic parameters).
|
Estimate
|
Std. Error
|
t value
|
P
|
|
Intercept
|
0.9463365
|
0.0632382
|
14.965
|
<0.0001
|
|
Serum triglycerides (mg/dl)
|
0.0008238
|
0.0001262
|
6.526
|
<0.0001
|
|
Sex=female
|
−0.1124217
|
0.0189772
|
−5.924
|
<0.0001
|
|
HDL-cholesterol (mg/dl)
|
0.0018476
|
0.0006213
|
2.974
|
0.0030
|
|
Serum glucose (mg/dl)
|
0.0012227
|
0.0005129
|
2.384
|
0.0173
|
|
LDL-cholesterol (mg/dl)
|
−0.0003699
|
0.0002440
|
−1.516
|
0.1298
|
|
R² for final model=0.07333333. RMSE of
cross-validation: 0.2964259. f²=0 .08
|
|
|
|
|
Lacrest: Lactate at resting conditions.
Table 4 Results of linear regression analysis (k-fold
validation, k=10) for explanatory parameters for
Lacrest (type of sport and aerobic
capacity).
|
Estimate
|
Std. Error
|
t value
|
P
|
|
Intercept
|
1.3063134
|
0.0492455
|
26.527
|
<0.0001
|
|
sex=female
|
−0.1150247
|
0.0192080
|
−5.988
|
<0.0001
|
|
Predominant
exercise=”Strength/Speed”
|
−0.0839597
|
0.0210391
|
−3.991
|
<0.0001
|
|
Predominant exercise=”Sedentary”
|
−0.0763436
|
0.0257877
|
−2.960
|
0.00313
|
|
VO2peak (ml/min/kg)
|
−0.0014581
|
0.0009672
|
−1.508
|
0.1319
|
|
R² for final model=0.04858364. RMSE of
cross-validation: 0.2993456. f²=0.05
|
|
|
|
|
Discussion
Lacrest and cardiovascular risk
Recent analyses using a population-based approach are linking atherosclerotic
disease with Lacrest: Publications based on the ARIC study have
highlighted a link between Lacrest with mortality [8], arteriosclerosis [9]
[12] and metabolic disease [10]. It was speculated whether this was due to
Lacrest pointing towards an “insufficient oxidative
capacity” [8].
Our data, based on a young and very fit study group featuring a low
cardiovascular risk (median: 1.5%), also confirms an association between
Lacrest and 10-year cardiovascular risk, when Lacrest
is used as a single predictor, which is consistent within subgroups of the
respective predominant type of exercise (see [Fig.
1a–d]). However, when parameters of the lipid profile and
other known risk modifiers are included in the prediction model,
Lacrest does not appear as an independent predictor of
cardiovascular risk in our population. This is most probably because in our
cohort, a significant association was observed between Lacrest and
metabolic factors such as serum glucose, serum triglycerides, and
HDL-cholesterol but not LDL-cholesterol (see [Table
3]). Whilst the present cohort is younger and healthier than the
subjects of the ARIC trials, the general finding of an association between
Lacrest and the lipoprotein profile was also observed in subjects
of the ARIC trial (e. g. [8]). However,
the main difference between our study and the reports from the ARIC trial is
that in our cohort, Lacrest loses its predictive value for
cardiovascular risk when adjusted for confounders, whereas the reports from the
ARIC trial yielded significant associations even when fully adjusting for the
lipoprotein profile.
Our data illustrates that when studying an association between Lacrest
and cardiovascular risk and/or metabolic disease, these associations
need to be accounted for not to attribute direct predictive properties to a
possibly “bystanding” molecule in the cardiovascular risk
profile. The observation of a remarkable intra-individual variability (see
below) further supports this notion pointing towards a rather second-tier role
of Lacrest in our cohort when aiming to gain insights concerning
cardiovascular risk. In this context, it is important that our healthy study
group did not feature relevant cardiovascular disease or diabetes as opposed to
the subjects of the ARIC trial.
Interestingly, and without further specification, higher Lacrest
values were interpreted as pointing towards impaired oxidative capacity by other
researchers speculating on the mechanism by which Lacrest might
impact on cardiometabolic disease [10]. Of note,
Matsushita and colleagues report a “sports score” which is
towards less physical activity in subjects with higher lactate values [8].
The data obtained from our healthy study population do not confirm this link
between a reduced oxidative capacity and Lacrest. First,
Lacrest is not significantly associated with the maximum oxygen
uptake estimated from stress testing, which per definition, is the maximal
oxidative capacity of the body during exercise (see [Table 4]). Second, our analyses show an association (of
“small” effect size) between Lacrest and the
predominant type of exercise adaptation in our study sample (see [Table 4]), that is somewhat contrastive to the
concept “increased Lacrest – reduced oxidative
capacity”: while subjects, who conduct predominantly endurance-type
exercise showed slightly higher Lacrest values, subjects conducting
mostly “strength/speed” -type of exercise appear to have
lower Lacrest values. Sedentary subjects feature Lacrest
that are somewhere in between the two physically active groups. This points
towards other mechanisms of elevated lactate values than a reduced oxidative
capacity, as especially endurance athletes would be expected to feature muscle
tissue with a higher oxidative capacity [25].
Physiologically, Lacrest values are not necessarily related to
oxidative capacity: A relevant proportion of glucose coming from the digestive
tract bypasses the liver and is consequently metabolized in peripheral tissues
leading to glycolysis and lactate production [26]
[27]. The lactate then enters the
liver through the bloodstream in the so-called “indirect
pathway” [26]
[27]. Lacrest is also physiologically linked to the diurnal
blood glucose and insulin levels [28]. However,
this relation is affected by metabolic disease which might lead to a higher
basal lactate production of adipocytes but also deteriorate the capacity of
adipocytes to provide lactate in response to glucose ingestion as part of this
“indirect pathway” [29]. This
regulation of Lacrest appears to be independent from a so called
“oxidative capacity” of muscle tissue and might explain the
associations between the relation between Lacrest and cardiometabolic
disease that were observed in the ARIC trial.
Possibly, the systematically higher Lacrest levels that we observe in
subjects who predominantly conducted endurance training (and most probably
therefore feature an excellent oxidative capacity) in our study sample will not
have the same physiological basis as the elevation of Lacrest
observed in the subjects from the ARIC study featuring increased cardiovascular
risk and cardiometabolic disease. In those subjects the observation of an
association between Lacrest and cardiometabolic disease might be
explained by regarding Lacrest rather as a symptom of an
already-manifest, subclinical cardiometabolic condition.
Distribution of Lacrest
To our knowledge, this study is the first to establish normal values for
Lacrest in capillary blood based on a larger population of young
healthy adults. When comparing the upper limit of normal based on our data,
which is 1.74 mmol·l-1, this appears to be slightly lower than
the arbitrary threshold of 2.0 mmol·l-1 that is commonly used in
clinical practice [30]. Noteworthy, our study
sample yielded an effect of sex on Lacrest with female sex being
associated with lower lactate values. When comparing the Lacrest
values of our cohort with data available from the ARIC publications, most study
groups featured lactate values that are compatible with our reference range
([10]: high risk group
1.21–1.63 mmol/l; [9]:
considered normal range 0.5–2.2 mmol/l), but some
authors observed Lacrest values well above our level of normal ([8], highest included Lacrest
55.5 mg/dl=6.16 mmol/l). From a
methodological point of view, it is relevant that lactate in the ARIC trial was
measured from plasma [8]
[11]
[31] and not from whole blood as in
our study. The difference between our observations and those made from the ARIC
study with respect to the association between cardiovascular risk or disease and
Lacrest do most probably not arise from systematic differences of
Lacrest levels between our study groups.
However, a reference range for Lacrest solely describes
between-subject variation. When developing and interpreting reference ranges for
a laboratory parameter, the inter- as well as the intra-individual variation
must be taken into account [18]. When the
intra-individual variability is much lower than the inter-individual variability
of a parameter, population-based reference values might not detect all relevant
alterations on the individual level.
Our analysis bases on r, the ratio between the intra- and inter-individual
variations, which was proposed by Eugene Harris (1974). When r is small
(<0.6, “marked individuality”), the intra-individual
distribution of measurements is smaller than the overall variance of samples
taken from a population. Laboratory parameters with such a small r are suitable
for clinical decision making based on individual threshold levels. We observed
an r value of 3 in repeated measurements of Lacrest in the same
subjects.
For r values≥1.4, the conventional limits include both, measurements of
subjects whose intra-individual variations are smaller than average, but also
those whose variation values are larger than average [18]. This further illustrates that a Lacrest value within
the normal range appears not to be suitable to be included into risk prediction
algorithms for the individual “chronic” cardiovascular risk as
it does not feature sufficient reliability for all subjects. In this context, it
is important to emphasize that the parameter estimates of the associations
between Lacrest and predictive variables (with a small but
significant effect size) such as sex, lipid parameters, and exercise type are
lower than the inter- and intraindividual variation of Lacrest
observed in our study sample. Although this may indicate a possible
physiological background, the clinical relevance of these observations appears
to be dubious. It can be speculated that future investigations aiming to confirm
a link between Lacrest and the statistically associated parameters
can be improved by vigorously standardizing the carbohydrate intake in the
period before the analysis because Lacrest is closely linked to blood
glucose and serum insulin levels [29].
Limitations
Per design, this study is unable to establish cause-and-effect relationships.
Bias might be introduced by the choice of cardiovascular risk calculation:
whilst the Framingham risk score was recommended by the European Society of
Cardiology’s Guideline at the time of data acquisition [32], it was recently shown that Framingham risk
models might overestimate cardiovascular risk in non-US cohorts and subjects at
higher risk [33]. Moreover, to our knowledge, the
Framingham risk score has not been validated in an athletic population, which on
the other hand side holds true for all algorithms for cardiovascular risk
prediction. In this context, it is noteworthy that “presence of left
ventricular hypertrophy in ECG” was omitted for the calculation of
cardiovascular risk, because in this cohort, left ventricular hypertrophy most
probably constitutes a benign adaptation to training. However, this might lead
to a systematic bias in calculation of the risk model. The population of the
present study is on average younger and most probably, much fitter and healthier
than the subjects participating in the ARIC studies. Therefore, a direct
comparison between the two study samples might be affected by fitness, presence
of disease and age, which might hinder direct transferability of our results. As
in all observational studies, although adjustments were performed using
multiparameter models, we cannot exclude residual confounding.
Conclusion
Our study describes the distribution of blood lactate concentration values at
physical rest based on capillary blood samplings in healthy subjects. The results of
regression analyses points towards a relation between lactate values at rest with
metabolic parameters and the predominant type of exercise, whereas no self-contained
predictive property is seen concerning the prediction of the estimated 10-year
cardiovascular risk in a young-healthy group of patient-athletes.
