Association between preoperative haemoglobin concentration and cardiopulmonary exercise variables: a multicentre study

Background Preoperative anaemia and low exertional oxygen uptake are both associated with greater postoperative morbidity and mortality. This study reports the association among haemoglobin concentration ([Hb]), peak oxygen uptake (V˙O2 peak) and anaerobic threshold (AT) in elective surgical patients. Methods Between 1999 and 2011, preoperative [Hb] and cardiopulmonary exercise tests were recorded in 1,777 preoperative patients in four hospitals. The associations between [Hb], V˙O2 peak and AT were analysed by linear regression and covariance. Results In 436 (24.5%) patients, [Hb] was <12 g dl-1 and, in 83 of these, <10 g dl-1. Both AT and V˙O2 peak rose modestly with increasing [Hb] (r2 = 0.24, P <0.0001 and r2 = 0.30, P <0.0001, respectively). After covariate adjustment, an increase in [Hb] of one standard deviation was associated with a 6.7 to 9.7% increase in V˙O2 peak, and a rise of 4.4 to 6.0% in AT. Haemoglobin concentration accounted for 9% and 6% of the variation in V˙O2 peak and AT respectively. Conclusions To a modest extent, lower haemoglobin concentrations are independently associated with lower oxygen uptake during preoperative cardiopulmonary exercise testing. It is unknown whether this association is causative.


Background
Increased mitochondrial oxygen uptake requires increased cellular oxygen delivery. When oxygen delivery, or utilization, fails to meet metabolic demand, anaerobic cytoplasmic metabolism significantly augments aerobic mitochondrial ATP generation with a consequent increase in lactic acid production and accumulation. It has been suggested that an imbalance in oxygen demandsupply contributes to the point during cardiopulmonary exercise testing (CPET) known as the anaerobic threshold (AT) [1], although this physiological underpinning is not without controversy [2,3]. Major surgery places substantial metabolic demands upon the patient and may increase resting oxygen uptake from an average preoperative value of 110 ml min -1 m -2 to approximately 170 ml min -1 m -2 [4,5]. Lower preoperative exertional oxygen uptake ( _ VO 2 ), both the peak and that noted at AT, are associated with postoperative morbidity and mortality [6][7][8].
Oxygen delivery limits maximal oxygen uptake during exercise under normoxic conditions [17,18]. Increases in [Hb] increase _ VO 2 peak, whilst acute reductions in [Hb] lower _ VO 2 peak and endurance performance [19][20][21]. However, the extent to which postoperative outcomes are dependent upon interactions between [Hb] and _ VO 2 is unknown. We analysed a large multicentre dataset to explore the relationship between preoperative [Hb] and _ VO 2 .

Patient population
We Following discussion among the researchers and the local Research and Development and clinical governance departments, formal ethical approval was waived and confirmation of audit status granted due to the nature of the data collection.

Cardiopulmonary exercise testing
Across all testing sites, CPET was performed according to the American Thoracic Society/American College of Chest Physicians (ATS/ACCP) guidelines, under stable environmental conditions, with continuous 12 lead ECG monitoring and in the presence of a clinician [22].
A 3-minute rest period followed fitting of relevant equipment, after which unloaded cycling was performed at a cadence of 60 to 70 rpm for 3 minutes. Thereafter, patients performed a symptom-limited continuous incremental exercise ramp protocol, determined by the physiologist or clinician on the basis of predictive work rate algorithms and patient-reported activity levels [23]. The test continued (usually for 8 to 12 minutes) until volitional exhaustion occurred, or the patient was unable to maintain a cadence of 40 rpm for more than 30 seconds despite encouragement. The clinician stopped the test if the patient developed a sign or symptom listed in the ATS/ACCP guidelines, which included: new arrhythmia; more than 2 mm of ST elevation or depression on the ECG; an arterial blood pressure of more than 250 mm Hg systolic or 120 mm Hg diastolic (see 2003 ATS/ACCP statement on Cardiopulmonary Exercise Testing for an exhaustive list) [23]. Following termination of CPET, patients were encouraged to perform a 'warm-down' period of unloaded cycling.
The anaerobic threshold was estimated by an exercise physiologist or consultant physician, both experienced in CPET interpretation, using a combination of the modified V-slope, ventilatory equivalents and end-tidal pressure methods [24], which improves the rigor of AT detection. The _ VO 2 peak was recorded as the highest average _ VO 2 over the final 30 s period [25]. The ventilatory equivalents for carbon dioxide ( _ VE= _ VCO 2 ) and oxygen ( _ VE= _ VO 2 ) were recorded at the AT [26]. We recorded age, sex, height, weight, body mass index (BMI, kg m -2 ) and Lee's Revised Cardiac Risk Index (RCRI) from patients' medical histories [27]. In addition, serum creatinine, obtained from hospital electronic records systems, was used as an index of renal function. At UCLH and the Whittington Hospital, [Hb] was measured on the day of CPET (HemoCue AB, Angelholm, Sweden). Preoperative [Hb] was recorded from hospital electronic record systems within 3 days of CPET at Newcastle and at the time of pre-assessment (usually within 4 weeks of CPET) at Torbay. No patient received a blood transfusion between the [Hb] measurement and the CPET.

Statistical analysis
Statistical analysis was performed using Stata Version 11 (StataCorp, Texas, USA). Gaussian distributions of the data were verified by the Kolmogorov-Smirnov test, in conjunction with visual inspection of histogram charts. A difference between data was considered significant if P <0.05. We transformed seven continuous variables with skewed distributions by taking their log 10 : weight, BMI, AT, _ VO 2 peak, _ VE= _ VCO 2 , _ VE= _ VO 2 , and creatinine. These variables are presented as geometric means and approximate SD.
Historically, measurements of oxygen uptake have been indexed to body mass (ml kg -1 min -1 ), as it allows comparisons between individuals [28,29]. However, this value may still vary with body mass [30][31][32]. We therefore adjusted the measured oxygen uptake by raising the body mass to a power determined by allometric scaling using the power function ratio (Y/X β ) [29,33]. Specifically, the allometric relationship between body size and performance measure (AT or _ VO 2 peak) is determined by the allometric equation below (see equation 1), where Y is AT or _ VO 2 peak, X is body mass, β is a scaling exponent, a is the proportionality constant (intercept), and ε is the multiplicative error term, which overcomes the problem of heteroscedasticity [34].
The allometric relationship between body mass (X) and fitness parameter (Y) is expressed using the logarithmic transformation of equation 1 so that where β is the sample specific slope of the linear least squares regression line calculated by log-linear regression analysis (that is, scaling exponent β was 0.83 in the current study) and log a is the equivalent constant value (a) [34]. We further built models by adjustment for the determinant variable (AT and _ VO 2 peak) to potential confounders. Three levels of increasing adjustment were used: i) a basic adjustment for testing site; ii) an extended adjustment for testing site, age and sex; and iii) a fully adjusted model for all known confounders (testing site, age, sex, revised cardiac risk index, diabetes, creatinine and operation category). Results were standardised for testing centre by the inclusion of dummy variables in the regression model.
The effect size was expressed as the percentage increase in _ VO 2 for a 1 g dl -1 (or one standard deviation) increase in [Hb]. Partial correlations between [Hb] and CPET markers were performed controlling for confounding variables. Regression models assessed the associations between _ VO 2 and [Hb] and the proportion of variance in oxygen uptake explained by variation in [Hb]. Covariance models were analysed with [Hb] as a clinically relevant categorical variable ([Hb] <10 g dl -1 ; 10 to 12 g dl -1 ; >12 g dl -1 ), as similarly described [15,35] Relationships between haemoglobin concentration and oxygen uptake  Table 2). More patients awaiting colorectal surgery were anaemic: in 144/530 (27%) of these, the [Hb] was 10 to 12 g dl -1 and in 32/530 (6%) the [Hb] was <10 g dl -1 .

Regression models
An increase in [Hb] by one SD was associated with a 9.7% (95% CI, 8.2 to 11.3) increase in _ VO 2 peak after adjusting for weight (P <0.0001), which was reduced to 6.7% (95% CI, 5.4 to 7.8) after adjusting for age, sex, weight and testing centre (P <0.0001). The percentage of the variance in _ VO 2 peak explained by [Hb] was 8.9% (P <0.0001) after adjusting for weight, and 5.5% (P <0.0001) after adjusting for age, sex, weight and testing site. An increase in [Hb] by one SD was associated with a 6.0% (95% CI, 4.8 to 7.3) increase in AT after adjusting for weight P <0.0001), which was reduced to 4.4% (95% CI, 3.3 to 5.5) after adjusting for age, sex, weight and testing centre (P <0.0001). The percentage of variance in AT explained by [Hb] was 5.9% (P 0.0001) after weight adjustment, reducing to 3.5% (P <0.0001) after adjusting for age, sex, weight and testing centre.  Table 3 and Figure 3). We analysed allometrically scaled _ VO 2 peak across individual surgical cohorts, with and without adjustment for confounders. _ VO 2 peak did not differ in hepatobiliary patients across [Hb] groups. _ VO 2 peak increased across each [Hb] group for colorectal, bariatric and other (mainly urological) patients, after adjusting for all confounders (testing centre, age, sex, weight, RCRI, diabetes, serum creatinine and operation category). _ VO 2 peak increased across each [Hb] classification for upper gastrointestinal and vascular patients after adjustment for testing centre, age, sex and weight, but not when additional confounders (RCRI, diabetes, serum creatinine and operation category) were added. _ VO 2 peak increased with each [Hb] group for maxillofacial patients after adjusting for testing and weight, but not when additional confounders were added to the model.

Discussion
To our knowledge, this is the largest study to explore the relationship between oxygen uptake and haemoglobin concentration in the clinical setting, and the first to control for body mass with allometric (log-linear) scaling. Oxygen uptake at peak exercise and at the anaerobic threshold (AT) increased with haemoglobin concentration [Hb], after adjusting for measured confounding variables.  The [Hb] was 10 to 12 g dl -1 in 353 patients (20%) and <10 g dl -1 in 83 (5%) patients. Other studies have reported rates of anaemia between 5% and 76%, a range partly dependent upon the indication for surgery and definition of anaemia [9]. The American College of Surgeons' National Surgical Quality Improvement Program (ACS NSQIP) recently reported a similar prevalence of preoperative anaemia (30.4%) [15].
The mean AT of 11 ml kg -1 min -1 is similar to that reported by other studies of preoperative populations [7,36,37]. However, this value is 2 to 3 ml kg -1 min -1 less than gender-specific reference values for age and height [38]. The mean _ VO 2 peak in our population, 15.5 ml kg -1 min -1 , is 11 ml kg -1 min -1 less than gender-specific reference values for age and height [38,39]. The aim of allometric scaling is to appropriately account for body size (that is, the scaled variable no longer varies with body size) [40]. Oxygen uptake is usually reported per unit body mass, ml kg -1 min -1 , a scale that requires further adjustment for body size [41]. In the obese, oxygen uptake expressed as ml kg -1 min -1 underestimates fitness and overestimates risk [42]. In the cachectic patient this scale overestimates fitness and underestimates risk [43].
Both AT and _ VO 2 peak increased with [Hb], for the whole population, across individual testing sites and across all groups, except hepatobiliary surgery. However, this relationship was weak and although being highly statistically significant does not necessarily reflect a magnitude of clinical association. Nonetheless, an increase in [Hb] by one standard deviation (that is, 1.8 g dl -1 rise in [Hb]) was associated with a 9.7% and 6.0% increase in weight-adjusted _ VO 2 peak and AT. The [Hb] explained 9% and 6% of the variance in _ VO 2 peak and AT respectively. The increase in AT and _ VO 2 peak with [Hb] may be due to increased oxygen-carrying capacity, or patients who are not anaemic exercising more than patients who are anaemic, or due to confounding. For instance, sick patients may be both anaemic and less fit. In addition, differences in AT, to some extent (although probably small), may be explained by inherent variations in measurement and/or interpretation or physiological context [44].
The cause of anaemia may be important. The most common cause is reported to be chronic disease, the severity of which being related to the degree of systemic inflammation [45,46]. Features of anaemia due to chronic disease include reduced red cell survival, impaired erythropoiesis, and impaired iron metabolism, all of which may reduce AT and _ VO 2 peak, directly or in combination [47]. Iron status was not routinely assessed BMI body mass index, _ VE= _ VO 2 and _ VE= _ VCO 2 , ventilatory equivalents for oxygen and carbon dioxide at the AT; AT, anaerobic threshold; _ VO 2 peak, peak oxygen uptake. Values are mean (SD). a ANOVA, P <0.0001; linear trend, P <0.0001. b ANOVA, P <0.0001; linear trend, P < 0.0001. c ANOVA, P <0.0001; linear trend, P <0.0001; d ANOVA, P <0.0001; e linear trend, P < 0.0001; f linear trend, P < 0.0001; g Adjusted for weight, testing site, age, sex, revised cardiac risk index, diabetes, creatinine and operation category. in our cohort but may independently influence fitness markers in the absence of anaemia. For example, iron deficiency with or without anaemia is associated with reduced fitness [48][49][50].
This study had some limitations. The observational, cross-sectional and retrospective nature of the data generates causative hypotheses but does not test them [51]. It would have been valuable to assess the association of overall survival and critical care use with both anaemia and oxygen uptake. The [Hb] may be an imprecise measure of blood oxygen carrying capacity, given that its value is influenced by disease-or therapy-related contractions or expansions in plasma volume. For instance, oxygen-carrying capacity may be normal if [Hb] is low due simply to an increase in intravascular volume. A better measure of oxygen-carrying capacity may thus be total mass of haemoglobin (tHb-mass). The tHb-mass displays a higher correlation with _ VO 2 peak (r 2 = 0.79) than either blood volume (r 2 = 0.76) or [Hb] [52,53]. The relatively small explained variance in AT and _ VO 2 peak by [Hb] (oxygen carrying capacity) in the current study suggests that other factors may play an important role in determining aerobic capacity. For example, other physiological factors that may limit _ VO 2 peak include pulmonary diffusing capacity, cardiac output and skeletal muscle limitations [54]. Although it is suggested that the AT reflects an imbalance in oxygen demand-supply, that is, that AT reflects onset of anaerobiosis, this is a much- Figure 3 Allometrically scaled oxygen uptake (ml kg -0.83 min -1 ) across [Hb] group (<10 g dl -1 , 10 to 12 g dl -1 , >12 g dl -1 ); _ VO 2 peak (A) and at the anaerobic threshold (AT) (B). Mean +/-SD values adjusted for weight, testing site, age, sex, revised cardiac risk index, diabetes, creatinine and operation category. P <0.0001 (ANOVA) and trend (P <0.0001) for both measures. Hb, haemoglobin concentration. debated and controversial concept [44,55]. In addition, variations in the AT, to some extent may be explained by inherent discrepancies in its measurement and/or interpretation [44].
Future studies of preoperative cardiopulmonary exercise testing should include [Hb] with long-term survival and quality of life as outcomes as well as considering alternative endpoints measured during exercise testing such as metabolic efficiency and the oxygen pulse [44].

Conclusions
In conclusion, anaemia is common in preoperative patients undergoing elective major surgery. There is an association between haemoglobin concentration and oxygen uptake during exercise, both _ VO 2 peak and AT, even after adjusting for measured confounding variables. Future studies may wish to address whether reversing anaemia before surgery improves these values, and thereby increases postoperative survival and function. Authors' contributions JMO, AFO, PJH, CS, JBC, and MS were responsible for data collection, drafting and revising the article and final approval of the version to be published. JAC was responsible for statistical analyses, revising the article and final approval of the version to be published. MPWG was responsible for substantial contribution to conception and design and facilitating acquisition of data, revising the article and final approval of the version to be published. TR and HEM were responsible for substantial contribution to conception and design, revising the article and final approval of the version to be published. All authors read and approved the final manuscript.