Controversies in the physiological basis of the ‘anaerobic threshold’ and their implications for clinical cardiopulmonary exercise testing


Dr James Hopker


This article reviews the notion of the ‘anaerobic threshold’ in the context of cardiopulmonary exercise testing. Primarily, this is a review of the proposed mechanisms underlying the ventilatory and lactate response to incremental exercise, which is important to the clinical interpretation of an exercise test. Since such tests are often conducted for risk stratification before major surgery, a failure to locate or justify the existence of an anaerobic threshold will have some implications for clinical practice. We also consider alternative endpoints within the exercise response that might be better used to indicate a patient’s capacity to cope with the metabolic demands encountered both during and following major surgery.

The anaerobic threshold is a commonly used endpoint in cardiopulmonary exercise testing (CPX, also referred to as CPEX, CPET or CarPET) [1–3]. A recent review concluded that ‘peak oxygen consumption and possibly anaerobic threshold are valid predictors of peri-operative morbidity’ [4]. Cardiopulmonary exercise testing is purported to provide an objective evaluation of the body’s response to ‘stress’, in the form of exercise-induced increases in oxidative demand. Anaesthetists (and sometimes cardiologists or surgeons) who usually conduct or supervise CPX tests therefore need to be familiar with the normal ventilatory and metabolic responses to exercise that are essential to interpretation. Since (as we shall show below) some of the fundamental physiology remains unclear, it follows that those who conduct the tests should interpret their results quite critically before applying them widely. Although the anaesthetic literature is replete with descriptions of the basic technique of CPX [5], it is sparser in discussion of the underlying physiology. This review is therefore primarily a physiological critique, but with clear bearing on clinical practice.

Cardiopulmonary exercise testing is suggested to provide an indication of the patient’s ability to cope with the metabolic demands created by the planned insult of major surgery. Several endpoints in the exercise test have been proposed as important. For example, maximal aerobic capacity (Vo2max) has been suggested to be useful, with a low Vo2max predicting higher peri-operative complication rates [6, 7]. However, Vo2max requires the patient to produce a maximal effort to exhaustion, which is difficult or inadvisable in some patients.

An ‘anaerobic threshold’ occurs at an exercise level below Vo2max. The very broad idea is that the less ‘fit’ the patient, the earlier it is that anaerobic metabolism will occur during an exercise test [2, 7–10]. It has been suggested that a specific ventilatory anaerobic threshold (VAT, a concept we discuss further below) can be used to predict peri-operative complication rates (VAT ≤ 11−1.min−1) [3, 10] across a range of major surgeries [11].

However, using the anaerobic threshold can be problematic and it is debatable both whether there exists a ‘threshold’ and also if it is ‘anaerobic’ [12–19]. Several terminologies (e.g. ‘ventilatory threshold’, ‘ventilatory anaerobic threshold’, ‘individual anaerobic threshold’, ‘lactate threshold’, ‘onset of blood lactate accumulation’ and the ‘maximal lactate steady state’) all appear to be used interchangeably but may in fact have very different meanings. With an absence of methodological standardisation, questions also remain over how best to measure anaerobic threshold. If substantiated, all these concerns cast doubt upon the appropriateness of this measure to guide clinical practice. Nevertheless, despite these uncertainties, the anaerobic threshold continues to be widely used.

In this review, we critically discuss (i) what is conventionally understood by the term ‘anaerobic threshold’, (ii) whether there is, in fact, a ‘threshold’ (discernible using either ventilatory or lactate measurement criteria), and (iii) the nature of the ventilatory response to exercise.

What is commonly understood by the term ‘anaerobic threshold’?

The answer to this question must be provided in physiological terms. If an individual performs an incremental exercise test where the work rate is progressively increased, O2 uptake (Vo2) increases in strict proportion to the rising work rate to fuel aerobic muscle contraction (Fig. 1). However, a point is reached where aerobic metabolism alone cannot supply the energy needed for the ever-increasing work rate and then the contribution from anaerobic metabolism becomes significant: lactic acid is released into the circulation as a product of glycolysis. Under these circumstances, Vo2 measurement alone will no longer account for all of the energy used by the muscle. Consequently, the anaerobic threshold has been conventionally defined as ‘the work rate or oxygen consumption just below that point at which metabolic acidosis and associated changes in gas exchange occur’ [20]. Production and excretion of CO2 (Vco2), however, do not rise in strict proportion to the rise in work rate (Fig. 1). Up to a certain point, they are linearly related, but beyond a certain work rate (which is said to be the same point at which lactate production becomes significant) there is an ‘extra’ output of CO2 (Fig. 1). The generally accepted explanation for this ‘excess’ CO2 production, offered by Wasserman, Beaver and colleagues [20–23] (but as we shall see below, disputed by others), is as follows. Bicarbonate (HCO3) is the main buffer for lactate in muscles and the neutralising reaction between the two yields CO2, which is subsequently excreted by the lungs and is detected as an increased Vco2 at the mouth. This extra CO2 is therefore (according to these authors) of ‘non-metabolic’ origin.

Figure 1.

 Volume of oxygen consumption (Vo2) and volume of carbon dioxide produced (Vco2) plotted as a function of work rate for an exercising subject. Note the linear increase in Vo2 (○), but a biphasic increase in Vco2 (•).

Changes in ventilatory control are also relevant. At low work rates, minute ventilation (VE) is strictly proportional to the work rate (and therefore also to Vo2 and to Vco2). Above the point ascribed as the anaerobic threshold, however, VE exceeds these proportionalities (Fig. 2): there is thus ‘extra’ ventilation in excess of Vo2 but broadly matching Vco2 (Fig. 1). This break-point in the ventilatory response is generally seen at work rates of 47–64% of Vo2max in healthy untrained individuals [23]. It is tempting to think that the break-points in Vco2 (Fig. 1) and in VE (Fig. 2) are causally related, but we will discuss this in more detail later in the article.

Figure 2.

 (a) Illustration of the ventilation (VE) vs work rate relationship during an incremental exercise test (cycle ergometer) in an example subject. The VE increases as work rate on the ergometer increases by 20 W every minute. Each data point represents 10 s averaged breath-by-breath data. It is possible to plot a bilinear relationship (but the bilinear plot does not fit some data points well). The break point could be said to represent an ‘anaerobic threshold’. (b) The same data as 2a, but a non-linear plot now fits all the data much better, where a discrete break-point is not evident.

In summary, the anaerobic threshold is commonly described by reference either to a blood lactate change (lactate anaerobic threshold, LAT) or to ventilatory gas exchange data, (ventilatory anaerobic threshold, VAT). Intuitively, it is felt that the points of the lactate and anaerobic thresholds should be the same.

Locating the ventilatory anaerobic threshold

The break-points referred above in the ventilatory and gas exchange responses form the basis for approaches to locate the anaerobic threshold point. The methods employed have undergone several refinements over the years. Initially, it was suggested that the point at which the slope of the VE-work rate relationship abruptly increased (Fig. 2a), could be used [20, 24]. However, it became apparent that the plot of VE against work rate (or time) in an incremental test was not strictly bilinear (which would make it easy to model and identify the ‘break-point’) but rather curvilinear (where VE rises smoothly with no single break-point; Fig. 2b).

To overcome this limitation and more directly to take account of the concept of ‘excess CO2 production’, Wasserman et al. plotted the ratio VE/Vo2 and the ratio VE/Vco2 against work rate [23] (Fig. 3). These ratios were termed ‘ventilatory equivalents’ for O2 and CO2, respectively, and they found that a break-point could be identified more clearly using VE/Vo2.

Figure 3.

 (a) The data in Fig. 2 re-plotted using the ratios of ventilation/volume of oxygen consumed (VE/Vo2) and ventilation/volume of carbon dioxide production (VE/Vco2) vs. power output. These ratios remain relatively constant during the exercise test, until the point at which the VE/Vo2 ratio increases (which can be assigned as the VAT). At this point, the ratio VE/Vco2 remains relatively constant and rises at a later point in the exercise test. The work rate at the VAT using VE/VO2 ratio is represented by the arrow; the grey box illustrates the period of ‘isocapnic buffering’. RCP, respiratory compensation point.

Above the VAT identified using the break-point in VE/Vo2 ratio is a region where VE/Vco2 ratio remains relatively constant (Fig. 3) and where, by definition, arterial Pco2 also remains relatively constant. This area of the graph is termed the region of ‘isocapnic buffering’ [23]: the implication here is that the lactic acid produced is being buffered by HCO3 (to generate the ‘excess’ CO2). At work rates higher than this, lactic acid production overwhelms blood buffers, the pH of blood declines and this constitutes an extra drive to ventilation. The ratio VE/Vco2 thus rises. Some authors have argued that this second change (termed the ‘respiratory compensation point’) should define the anaerobic threshold [14, 25, 26], so it is a matter of debate as to which of these two break-points (i.e. in VE/Vo2 or in VE/Vco2) should properly constitute the true ‘anaerobic threshold’ [25].

Another common method used to determine the anaerobic threshold is the ‘V-Slope’ method [21]. One way of conceptualising this method is to consider that a straight line is fitted to the plot of Vco2 vs Vo2 using data obtained from both the early and later portions of an incremental exercise test. Anaerobic threshold is defined as the point where the slopes of these two lines intersect (Fig. 4). There exist several mathematical approaches to performing the calculations [27]. One potential limitation is of course the number of data points or range used to plot the two respective slopes (i.e. the gradient of each slope and intersect will depend on the number of data points used in the calculation). Nevertheless, Beaver et al. [21] have suggested that the results are relatively insensitive to this effect.

Figure 4.

 The V-Slope method for determining AT 4. The figure shows the same data set as that in Fig. 3. Here, the volume of carbon dioxide produced (Vco2) is plotted against the volume of oxygen consumed (Vo2) data from which a regression line for both the lower (line 1) and upper (line 2) parts of the curve is subsequently drawn. The inflection point of the relationship is established from the intersection of the two regression lines. Arrows represent the respective values for Vo2 and Vco2 at the putative anaerobic threshold.

Despite using all these methods, VAT is not always immediately discernible in practice in all subjects and there appears significant variability between independent reviewers when identifying the VAT in the same data sets across a range of different methods [19, 28]. For example, three different methods of calculating the V-slope resulted in VAT estimates using the same data of 886, 1178 and 1793−1.min−1 for one subject and 2027, 3196 and 2900−1.min−1 for another (i.e. a range of ∼58–102% of the lowest estimate) [27]. When the V-slope method is assessed for repeatability in the same subjects over a 6-week period without any training, it shows coefficients of variation of 12–17% and correlation coefficients of as low as 0.65 (compared with the highly repeatable Vo2max, μ which shows correlation coefficients of ∼0.96) [29].

Poor agreements of this type may occur because some of the assumptions on which these methods of identification of the anaerobic threshold are based may themselves be flawed. Recall that these methods assume the notion of ‘excess CO2’ produced as a result of HCO3 buffering of lactate. If this were not the case, then the basis of these tests would in turn be undermined. Indeed, it appears that: (i) HCO3 is not the main buffer in muscle; (ii) buffering of lactate by HCO3 in any case does not produce excess CO2; and (iii) the disproportionate rise in VE above the point commonly termed ‘anaerobic threshold’ is not due to the purported excess Vco2 also occurring above this point; rather the opposite: the excess minute ventilation causes the excess Vco2 (and as we shall see below, something else causes the excess VE). We will elaborate these points below:

Bicarbonate is not the main buffer in muscle

Beaver et al. [21] estimated that ∼92% of lactate is buffered by HCO3, and Wasserman et al. [20] further proposed that the concentrations of HCO3 and lactate mirror each other, such that as the magnitude of one rises, that of the other declines in a 1:1 ratio (perhaps aided by an antiport transport system in the muscle fibre membrane [23]). These observations were taken to support the hypothesis of HCO3 buffering.

However, detailed work by Hultman and Sahlin [30] estimates the HCO3 contribution to buffering to be just ∼16%. In muscle, buffers such as protein histidine residues (31%), phosphocreatine (29%), phosphate (8%) and carnosine (4%) are collectively more important. Furthermore, no antiport has been found and lactate movement occurs via monocarboxylate transporters which act as symports not antiports [31].

Buffering of lactate by bicarbonate does not produce ‘excess’ carbon dioxide

Peronnet and Aguilaniu [32] explained elegantly why, even if lactates were buffered in the main by HCO3, no excess CO2 would result. The simple reason for this is that HCO3 is formed in the first place by dissolution of CO2 in blood: any subsequent decomposition of HCO3 back into CO2 does not result in any excess CO2, but rather the same quantity of CO2 that dissolved in the first place (In other words, the laws of mass balance hold and there is no net creation of extra CO2 during neutralising reactions by HCO3).

Hyperventilation causes the rise in Vco2, rather than vice versa

The standard interpretation proposes that the purported excess Vco2 (Fig. 1) itself acts as a stimulus to VE above the so-called anaerobic threshold (Fig. 2). We discuss below, in more detail, the mechanisms and reflexes involved in the control of breathing during exercise, but one problem with this suggestion is that the body possesses no receptors to detect Vco2 so it is unclear how an increase in VE would be brought about. Secondly, it is well established that simple voluntary hyperventilation can increase Vco2 measured at the mouth transiently (e.g. for 2–3 min). Because incremental exercise is a series of short-lived, step-increases in work rate, such a transient effect could recur over the whole course of the exercise test. Thus, it seems that the ‘excess Vco2’ may simply be a consequence of the increased minute ventilation (itself brought about by other factors) rather than by any excess production due to buffering. And if that is the case, then it follows that any measurement methods (such as VE/Vco2 or V-slope) based upon the assumption that CO2 production increases due to buffering, may in turn be prone to error.

Locating the lactate anaerobic threshold

Given the problems inherent in the techniques of identifying a VAT, measuring blood lactate itself might enable more robust identification. However, identifying a ‘LAT’ leads to its own problems.

Numerous studies have supported the very plausible link between muscle hypoxia and lactate production [33]. Blood lactate increases more when cardiac output is impaired [34, 35] and also when hypoxic gas mixtures are inspired during exercise. Conversely, breathing hyperoxic mixtures attenuates the exercise-induced increase in blood lactate [36, 37]. However, these lines of evidence do not confirm the occurrence of tissue hypoxia as a key trigger for lactate production. There are contrary observations, to which we now turn.

First, it is the accumulation of blood lactate (suggesting the rate of lactate production has exceeded its removal from muscle and/or blood), not simply the production of any lactate, that should be considered to represent the body’s response above the anaerobic threshold. Lactate production has been shown to occur within a single muscle even at rest [38–40] and under fully aerobic conditions at low exercise levels [14, 41, 42], although lactate does not start to accumulate at this point. The term ‘anaerobic’ is therefore arguably a misnomer, as muscle hypoxia is not required for lactate production to occur [41, 42]. Rather, it might be that the capacity for lactate removal exceeds the rate of lactate production by the muscle [41]. These processes of production and removal are continuous and the balance of these determines whether lactate accumulates [41]. As lactate production increases with, for example, incremental exercise, so too does lactate sequestration. Since these two processes occur gradually and at different times in each muscle fibre or muscle unit, the aggregate effect on blood lactate levels is unlikely to be discrete but a smooth change over time as exercise intensity increases (Fig. 5; i.e. not all tissue units reach their maximum limit for lactate sequestration at exactly the same time) [16, 18, 19, 28, 42].

Figure 5.

 The smooth relationship between blood lactate production and exercise intensity in an experimental subject. Like the ventilatory response with incremental exercise (see Fig. 2b), there is no clear threshold point.

Another problem is that factors unrelated to exercise also control lactate production. For example, adrenaline (which can be released during normal exercise) has itself been shown to cause an increase in blood lactate concentration [26, 42, 43], whereas β-blockade reduces blood lactate accumulation [44]. Similarly, recruitment of type-II muscle fibres with increasing work rate causes an increase in blood lactate production regardless of ambient Po2 [45].

In summary, anaerobic metabolism does occur with heavy exercise and lactate does accumulate as a consequence, but a clear time-point at which this occurs is not easily discernible.

The tight link between ventilation and metabolism: key role for carotid bodies in the shape of the VE response

Central to understanding the exercise test – and therefore to understanding the anaerobic threshold concept – is the nature of the ventilatory response. Yet this is an aspect sparsely discussed in the anaesthetic literature. The physiological basis of this response is one of the oldest (and yet incompletely resolved) questions in physiology. The properties of the individual elements of the respiratory control system in isolation (e.g. the peripheral and central chemoreceptors, various chemo- and mechano-receptors in muscles and joints, the efferent control system of the intercostal muscles and diaphragm, the central neuronal structures in the brainstem) are well-described. However, what is less clear is how these elements work together to regulate ventilation so precisely to the work rate, such that arterial blood gas levels and pH are kept constant at low-to-moderate work rates.

Since arterial Po2, Pco2 and pH are unchanged from their resting levels, it was assumed that the peripheral chemoreceptors (i.e. carotid bodies in humans) were not active during such low-to-moderate levels of exercise, but in fact they are more active than at rest. This is evidenced by administering hypoxic gases to individuals whose VE can be observed to rise by a greater magnitude during exercise than at rest [46].

Carotid body stimuli during exercise

Physiologists realised that if the factor raising the activity of the carotid bodies was a chemical released into the bloodstream in proportion to the work performed, it would constitute an ideal mechanism to link ventilation to metabolism. Asmussen and Nielsen called this (yet unknown) substance ‘Factor X’ [47]. Substances such as catecholamines [48] or adenosine [49] have been proposed as candidates, and Paterson et al. [50] persuasively argued that arterial plasma K+ (whose levels reach ∼8 mm in heavy exercise) fulfilled many of the criteria required of a Factor X. The observation that arterial blood gas levels may oscillate rapidly around a (constant) mean value led to the notion that such oscillations (their amplitude or frequency) may relate to the level of work done and form a drive to breathe [51].

However, a key result against such a blood-borne drive to ventilation is the finding that when metabolites from the working muscles are prevented from entering the bloodstream (e.g. by the use of tourniquets) [52], VE still matches the increase in metabolism. Therefore, ‘neural drives’ such as the activity of proprioceptors in joints or muscle spindles may be more important in maintaining the tight link between ventilation and metabolism.

Doubts about the ‘neural’ drive to breathing in exercise; the concept of redundancy of the system

However, in paraplegic subjects (in whom the connection between muscles made to work by direct electrical stimulation and the central nervous system is disrupted), not only is the link between ventilation and metabolism maintained, but the activity of the carotid bodies is also increased, just as it is in exercise in ‘able-bodied’ subjects [53–55]. If neither blood-borne signals to breathe, nor those carried via afferent nerves are essential to the ventilatory response, then what is left?

One suggestion is that the respiratory control system demonstrates considerable ‘redundancy’ [56]. Because the homeostasis of blood gases and pH is so important, and because exercise is such a common and necessary state for the organism, the body has evolved several ways of ensuring the tight link between ventilation and metabolism, such that loss of any one mechanism through disease or injury means that other mechanisms ‘fill in’ or compensate instantaneously to yield a near-normal response.

The nature of the ventilatory response at high intensity exercise

If determining the physiology underlying the ventilatory response to exercise at low-to-moderate levels was problematic, the nature of the response at the higher levels of exercise appeared (for a time at least) clearer. Wasserman et al. [57] demonstrated that individuals after carotid body resection (an old treatment for the breathlessness of asthma) lost the biphasic or curvilinear ventilatory response to incremental exercise. The relationship between VE and work rate in these patients became essentially linear (Fig. 6). They argued that as these patients did not have peripheral arterial chemoreceptors, they were unable to detect changes in pH caused by accumulation of lactic acid and, therefore, they lost the extra drive to ventilation that normally accompanied heavy exercise. These subjects had ‘lost’ their VAT but had retained a LAT. Wasserman et al. concluded that intact carotid bodies were responsible for the break-point (or curvilinearity) in the ventilatory response. Since a break-point is central to the idea of the ‘anaerobic threshold’, it followed that the carotid bodies were key to this concept.

Figure 6.

 In normal subjects (solid line) incremental exercise yields a ‘bilinear’ relationship in which an anaerobic threshold (AT) may be identified. With carotid body resection (CBR; dotted lines), no such break-point is discernible and the relationship between ventilation (VE) and volume of oxygen consumption (Vo2) is more linear. Error bars are ± SEM (copied from Wasserman et al. [57] Journal of Applied Physiology, 1975, Am Physiol Soc, used with permission).

However, the details of this simple explanation were questioned by Hagberg et al. [58] and Paterson et al. [59], who observed that patients with myophosphorylase deficiency (McArdle’s disease) exhibit a biphasic (curvilinear) ventilatory response to incremental exercise even though they cannot produce lactate. Paterson et al. [59], noted that the rise in plasma K+ during the exercise test paralleled the ventilatory response and that both K+ and VE patterns were similarly curvilinear. Thus, although the carotid bodies were likely to be responsible for the unique shape of the ventilatory response, it was arguably K+, rather than lactate, that acted as the relevant stimulus. In contrast to patients with carotid body resection, McArdle’s patients had ‘lost’ their LAT, yet retained a VAT. Clearly, lactate is not necessary to generate a biphasic or curvilinear ventilatory response, but are even the carotid bodies (regardless of the stimulus acting upon them) needed?

Wasserman et al. [57] had ‘silenced’ the carotid bodies by surgical removal to convert a biphasic or curvilinear ventilatory response into a linear one. However, silencing them by other means (breathing high O2 mixtures or infusion of dopamine in healthy subjects) did not have the same effect [60–62]. This suggests that these organs are not essential for the shape of the ventilatory response. The system is again displaying its characteristic of ‘redundancy’ in generating a normal response even when individual components are removed or silenced.

Furthermore, if lactate accumulation underlies the anaerobic threshold, and in turn underlies the ventilatory response, then we might expect the time-point of the LAT to correspond to that of the VAT. Yet Hambrecht et al. [63] discovered a discrepancy in oxygen uptake at the LAT vs VAT. Dickhuth et al. [64] found the VAT to occur at a higher exercise intensity than the LAT. Chicharro et al. [65] showed significant differences between LAT and VAT when plotted against either heart rate, work rate or Vo2, during incremental cycling exercise. Overall, studies find that the correlation coefficient between VAT and LAT can be as low as 0.53 [18]. This dissociation of LAT and VAT suggests that the two events are independent and therefore governed by different underlying physiological processes [17].

Critique of the anaerobic threshold notion: a summary

Table 1 summarises the main tenets of the conventional views of the anaerobic threshold and also the main counter-arguments to these.

Table 1.   Summary of the central tenets of the ‘anaerobic threshold hypothesis’ (first column) and the main criticisms or counter-arguments to these tenets (second column).
Central tenets of ‘anaerobic threshold hypothesis’ during incremental exerciseCriticisms and counter-arguments
  1. VE, minute ventilation; Vo2, oxygen consumption; Vco2, carbon dioxide production.

VE response exhibits a break-point (i.e. it is bilinear with respect to measures of work rate)VE response is curvilinear, such that no single break-point can be identified
Break-point in VE response is due to lactateCurvilinear VE response retained even in absence of lactate generation (eg, McArdle’s disease): arterial plasma K+ may be more important in generating the shape of the VE response
Break-point in VE response is due to intact, functioning carotid bodies (so is lost when they are surgically removed)Curvilinear VE response retained even when carotid bodies silenced (eg, using hyperoxia or dopamine infusion)
‘Excess’ Vco2 is due to buffering of lactate by HCO3Buffering of lactate by HCO3 is limited (other muscle buffers more important)
Even if it occurs, it does not release any ‘excess’ CO2; rather only the same CO2 as created HCO3 in the first place
‘Excess’ Vco2 causes the disproportionate rise in VE relative to work rate at high work ratesThere are no receptors for detecting Vco2
It is the disproportionate rise in VE relative to work rate at high work rates that causes the apparent ‘excess’ Vco2
Other mechanisms cause the disproportionate rise in VE relative to work rate at high work rates (e.g. rise on arterial plasma K+)
The anaerobic threshold is reliably identified by various techniques including break-points in profiles of VE, lactate, Vo2, Vco2, ventilatory equivalents, isocapnic buffering, etcThe various techniques yield very different values for purported anaerobic threshold, with poor repeatability and inter-observer or inter-algorithm agreement
There is often a choice of several purported break-points, such as the ‘ventilatory anaerobic threshold’ vs the ‘respiratory compensation point’ vs the lactate anaerobic threshold, and so on
Lactate response exhibits a break-point (i.e. is bilinear with respect to measures of work rate)Lactate is produced even during aerobic conditions at rest
The lactate response is curvilinear, such that no single break-point can be identified
Anaerobic threshold measured using ventilatory or gas exchange criteria matches that using lactate measurementsVentilatory or gas exchange profiles do not match the lactate profile
Adverse clinical outcomes correlate with a specific (low) value of anaerobic thresholdAssociations of adverse clinical outcomes with a specific (low) value of anaerobic threshold are spurious, and not all trials confirm the association
There are other, more appropriate measures within the exercise response that may have stronger associations, and that need further research

Implications of the physiology for clinical exercise testing

Cardiopulmonary exercise testing machines generate a value for the ‘anaerobic threshold’ (usually using one of the gas exchange methods described above) and the notion is integral to almost all training programmes and courses. Indeed, several clinical studies have found that using an anaerobic threshold serves as a useful predictor of patient outcome, with a cut-off of precisely 11−1.min−1 suggested [3, 10]. Carlisle et al. [66] measured the VAT of 130 patients before abdominal aortic aneurysm repair. Multivariate analysis demonstrated that their VAT correlated significantly with postoperative survival. Significant correlations between VAT and postoperative survival rate have also been shown by Older et al. [10], who investigated 187 elderly patients undergoing major abdominal surgery. In patients with a VAT of < 11−1.min−1 (n = 55), 10 died of cardiovascular causes (18%), whereas in patients with a VAT of > 11−1.min−1 (n = 132), only one died of cardiovascular causes (0.8%), p < 0.001.

How can this be so, given the real physiological uncertainties around the concept of an ‘anaerobic threshold’? One possibility is that these ‘positive’ findings arise by chance in relatively small studies. Not all trials confirm the positive correlations. For example, Nagamatsu et al. [67] investigated 91 patients before radical oesophagectomy and found no relationship between the VAT and postoperative complication rate. Similarly, Forshaw et al. [68] found no significant difference between the VAT of the patients (n = 78) who did and did not suffer complications following oesophagectomy (p = 0.07). A second possibility is that the interventions of anaesthetists during surgery may in some way differ between those with higher vs lower anaerobic thresholds, making more likely a significant relationship between pre-operative testing and postoperative outcomes. A third possibility is that there are two distinct patient populations. In one subgroup (e.g. those suffering cardiac dysfunction [69]) there is a pre-existing abnormality of the macrocirculation that can be worsened by major surgery, leading to tissue hypoxia. The onset of anaerobic metabolism during exercise might reflect these macrocirculatory limitations rather well. However, in the other subgroup, the macrocirculation is normal but major surgery elicits an inflammatory response predominantly affecting the microcirculation and cellular function, leading to tissue hypoxia by different mechanisms. Exercise testing might reflect this situation less well. If the proportions of patients in these categories vary between the different trials, this may explain why some trials show positive results and some do not. A final possibility relates to the nature of the measurement being performed. The method embedded in CPX identifies a break-point in the ventilatory or gas exchange response. Although somewhat arbitrary, the methods do very crudely identify a time-point somewhere near to when the non-linearity in respiratory responses becomes pronounced. Whether or not this is because of anaerobic metabolism is irrelevant, but what is important is that the non-linearity becomes pronounced at a somewhat earlier time-point in those who are ‘unfit’ vs those who are ‘fit’. Thus, it is feasible that applying the concept of the anaerobic threshold sometimes (or even often) ‘works’, albeit for the ‘wrong’ reasons. Approaches can be crude yet still work in practice (e.g. pH vs the more sophisticated Stewart analysis of acid–base disorders) [70].

Alternative endpoints within the exercise test

Although it is possible to continue to use the anaerobic threshold point as an empirical measure, without taking account of any of the physiological controversies outlined above, it would be more progressive to identify a more appropriate end-point, perhaps more clearly reflecting the two possible patient subgroups alluded to above. We suggest two: metabolic efficiency and the oxygen pulse.

An efficient use of oxygen during the process of aerobic energy production might be an important factor for survival [71, 72]. If a patient was, in a general sense, ‘metabolically efficient’, the reduced availability of oxygen both during and after surgery would have a lower impact than it would in a patient who was less efficient. Thus, rather than trying to identify a discrete point that may signify the limit of a patient’s aerobic metabolism (i.e. Vo2max or VAT or LAT) it might be more pertinent to assess how well a patient is able to cope with prolonged tissue hypoxia. A simple exercise efficiency test [73] performed as part of pre-operative screening could be used, in which ‘gross efficiency’ is defined as the ratio of work accomplished (or power output on the cycle ergometer) divided by energy expended. This last entity can be estimated from calculating the energy expenditure for a given Vo2, related to the known respiratory quotient from gas exchange data. For a reliable measurement of gross efficiency, gas measurement should occur under steady state conditions for at least 4 min [74], which is perhaps clinically preferable to the exhaustive exercise of an incremental test. A high pedal cadence can erroneously suggest low efficiency [75–78], so a standardised cadence must be used when conducting repeated measurements of efficiency in the same individual. Future studies should seek to test the hypothesis that such metabolic efficiency might provide a more powerful predictor of post-surgical outcome than those endpoints currently used for this purpose.

Evaluation of the ‘oxygen pulse’ might also provide valuable information during CPX testing. Oxygen pulse is the volume of o2 taken up by the pulmonary blood during the period of a heartbeat [79], often using data from an incremental exercise test. It is derived from a re-arrangement of the Fick equation:


Thus a reduced O2 pulse with exercise implies a reduced stroke volume response (e.g. cardiac dysfunction) or a reduced ability to extract O2 from the blood (as in microcirculatory dysfunction) [79]. Oxygen pulse displays a characteristic bilinear pattern; initially a linear rise with exercise intensity, which later flattens. In disease, both the slope of the initial rise and the point and degree of flattening are more pronounced (Fig. 7).

Figure 7.

 The relationship between O2 pulse and increasing exercise intensity. Pattern A inline image illustrates a normal trace; Pattern B inline image shows the response in a patient with cardiac disease. Reproduced from Lim et al. [79] with permission.

The submaximal O2 pulse has been shown to effectively risk stratify heart disease patients [80] and so may negate the need for maximal testing associated with the determination of Vo2max.

Finally, relatively little work has been undertaken on the recovery period after exercise, which may better reflect processes that occur with recovery after major surgery than does an incremental exercise test (which creates a transition into exercise). In cardiology, it has been recognised that recovery of heart rate after exercise is an independent predictor of survival [81], and respiratory measurements made in the post-exercise period are insightful of control mechanisms in physiology [82]. However, it remains to be explored whether gas exchange kinetics in this phase might predict postoperative outcomes [83].


The concept of the anaerobic threshold is more indeterminate and the methods of locating it are more complex than they may superficially at first appear. Rather than try to refine further the methods of locating exactly a putative ‘anaerobic threshold’, we suggest that the important research question is to establish which of the alternative endpoints (that have a more robust physiological basis) best predicts clinical outcomes. This question may be answerable, in part, by re-analysis of some existing data. In future, the best approach to assessing how well a patient might cope with the insult of major surgery is likely to be a (short) battery of fitness assessments as part of a pre-operative screening, perhaps including exposure to modest hypoxia. This last approach will require more sophisticated clinical exercise laboratories than are generally available, to deliver and monitor hypoxic gas mixtures. Clinical decisions may thereby become less reliant on the concept of an elusive entity.

Competing interests

JJP is an Editor of Anaesthesia and this manuscript has undergone an additional external review as a result. No external funding and no competing interests declared.