Tissue oxygenation essentially relies on DO₂, VO₂, and the ability of the tissue to extract oxygen. DO₂ is the rate at which oxygen is transported from the lungs to the tissues and is the product of the CO and the arterial oxygen content (CaO₂): DO₂ = CO × CaO₂, where CaO₂ = hemoglobin concentration (Hb) × arterial oxygen saturation (SaO₂) × 1.34 (the oxygen carrying capacity of Hb). DO₂ can, therefore, be influenced by changes in CO, hemoglobin concentration, and oxygen saturation. VO₂ is the amount of oxygen removed from the blood by the tissues per minute and is the product of the CO and the difference between CaO₂ and mixed venous oxygen content (CvO₂): VO₂ = CO × (CaO₂ − CvO₂). VO₂ is determined by the metabolic rate of the tissues, which increases during physical activity, hyperthermia, shivering, etc. The ratio of the oxygen consumed to that delivered (VO₂/DO₂) represents the amount of oxygen extracted by the tissues, the oxygen extraction ratio (O₂ER).

As tissues are unable to store oxygen, it is important for them to have a system by which delivery of oxygen can be adjusted efficiently to oxygen demands. Under normal physiological conditions, as DO₂ decreases, oxygen extraction increases to compensate and maintain VO₂, ensuring adequate tissue oxygenation for aerobic metabolism and normal cellular function: VO₂ is independent of DO₂. Indeed, at rest, VO₂ is only about 25 % of DO₂, so that there is a large reserve of oxygen available for extraction if needed as DO₂ falls [12]. However, a point is reached at which oxygen extraction is unable to increase further and a so-called critical DO₂ is attained at which VO₂ becomes dependent on DO₂; any further decrease in DO₂ is associated with a decrease in VO₂ and anaerobic metabolism with a rise in blood lactate levels [13–15].

During septic shock, the ability of tissues to extract oxygen is reduced so that this VO₂/DO₂ relationship can be altered with the critical DO₂ set at higher values such that VO₂ is dependent on DO₂ over a larger range of values [13–15]. The reasons for the reduced oxygen extraction abilities in sepsis have not been fully elucidated but are likely to be related in part to the microcirculatory changes seen in sepsis, including increased heterogeneity, increased stop-flow capillaries, and increased shunting of DO₂ from arterioles to venules [12]. Impaired ability of mitochondria to use the available oxygen may also play a role in microcirculatory dysoxia [16].

Anemia, widely defined in ICU studies as a hemoglobin level <12 g/dl [7, 17], is common in critically ill patients [6, 7, 18]. In the ABC study [3], 29 % of patients had a hemoglobin concentration <10 g/dl on admission. In a Scottish cohort, 25 % of patients had a hemoglobin concentration <9 g/dl on ICU admission [19]. Hemoglobin concentrations decrease during the ICU stay, particularly in septic patients, in whom Nguyen et al. [20] reported a decrease of 0.68 ± 0.66 g/dl/day; this study also noted that hemoglobin concentrations continued to decrease after the third day in patients with sepsis but not in those without [20].

Multiple factors act together to cause anemia in the critically ill patient, including primary blood losses (trauma, surgery, gastrointestinal bleeding, etc.), phlebotomy losses, which can reach as much as 40 ml/day [20], hemodilution secondary to fluid resuscitation, blunted erythropoietin (EPO) production, abnormalities in iron metabolism, and altered red blood cell production and maturation [21–23]. In healthy subjects, compensatory mechanisms, including the increased oxygen extraction discussed above, but also reflex increases in CO because of decreased blood viscosity, increased adrenergic response, causing tachycardia and increased myocardial contractility, and blood flow redistribution (to heart and brain) enable severe anemia to be tolerated [24]. However, in critically ill patients, compensatory mechanisms are less efficient, and oxygen reserves are reduced so that lesser degrees of anemia may have greater consequences on organ function and outcome. Oddo et al., in a retrospective study of patients with traumatic brain injury who had had brain tissue oxygen tension (PbtO₂) measured, noted that anemia associated with reduced PbtO₂ was a risk factor for unfavorable outcome, but not anemia alone [25]. Patients with myocardial ischemic disease may be particularly sensitive to the effects of anemia because the associated tachycardia and increased contractility may increase myocardial oxygen demand, which will need to be met by increased coronary blood flow as myocardial oxygen extraction is almost maximal already at rest [23, 26].

Because DO₂ is the product of CO and CaO₂ is determined in part by the hemoglobin concentration, when the hemoglobin concentration decreases, DO₂ will decrease (if CO remains unchanged). Hence, one may anticipate that increasing the hemoglobin by giving a transfusion would help increase DO₂ as has indeed been shown in several studies [27–29]; although by increasing blood viscosity, some of the compensatory mechanisms of acute anemia on left ventricular pre- and afterload will be reduced, thus limiting the effects on DO₂ [5]. Moreover, even if DO₂ does increase, there is no guarantee that VO₂ and hence oxygen availability to the tissues will also increase, particularly in patients with an abnormal VO₂/DO₂ relationship and an altered microcirculation, such as those with sepsis [27, 29, 30]. There are several possible reasons for this including the fact that the ability of hemoglobin to download oxygen may be altered in sepsis because of microcirculatory changes, such as altered red blood cell deformability, altered oxygen extraction capabilities, reduced functional capillary density, and increased heterogeneity of flow. The ability of hemoglobin to deliver oxygen may also be influenced by changes that occur during storage of blood [31] and by increased blood viscosity following transfusion leading to reduced microcirculatory flow. Additionally, tissue oxygen demands are increased in patients with sepsis.

Importantly, different tissues have different critical DO₂ values and VO₂/DO₂ relationships and develop hypoxia at different degrees of acute anemia [32]. Hence, global assessment of the VO₂/DO₂ relationship cannot be used to guide therapy. “Coupling of data,” which occurs when both variables have been calculated from the same values, is also a problem when using this relationship [15]. Cardiac output represents total body blood flow and can be monitored almost continuously but offers no information on regional organ perfusion. Cardiac output is also highly variable among individuals and varies according to oxygen requirements; for example, in sepsis the typically “normal” or high CO seen may be insufficient because of increased sepsis-related tissue oxygen requirements.

Mixed venous oxygen saturation (SvO₂) has been widely used as a marker of tissue oxygenation, and, indeed, as oxygen extraction increases to meet oxygen demands, SvO₂ will decrease. However, although low SvO₂ indicates poor tissue oxygenation, normal or high SvO₂ values do not necessarily mean that tissue oxygenation is adequate; for example, if a tissue is unable to extract oxygen, the venous return from that area may still have a high oxygen content although the tissues may be hypoxic. Central venous oxygen saturation (ScvO₂) is increasingly used as a less invasive surrogate for SvO₂, but again this is a global measure. Rivers et al. [33], in their landmark study, randomized patients admitted to an emergency department with severe sepsis and septic shock to receive standard therapy (targeted at a central venous pressure [CVP] of 8–12 mmHg, mean arterial pressure [MAP] ≥65 mmHg, and urine output ≥0.5 ml/kg/h) or to the so-called early goal-directed therapy (EGDT) in which an ScvO₂ of at least 70 % was also targeted by optimizing fluid administration, giving blood transfusions to maintain hematocrit ≥30 %, and/or giving dobutamine to a maximum of 20 μg/kg/min. The EGDT group received more fluids, and more were treated with dobutamine; the number of transfused patients was also greater than in the standard therapy group. Patients in the EGDT group had significantly lower mortality rates than other patients, and this study therefore seemed to support the use of ScvO₂ values to guide therapy, including transfusions [33]. However, in the recently published Protocolized Care for Early Septic Shock (ProCESS) study [34], there were no significant differences in 90-day mortality, 1-year mortality, or the need for organ support in patients managed with protocolized EGDT – using a similar protocol to that used by Rivers et al., protocolized standard therapy or usual care.

The O₂ER is easy to calculate, and plotting cardiac index (CI) against O₂ER and relating them to isopleths of VO₂ can help identify whether a patient has reached the point of VO₂/DO₂ dependency and evaluate the adequacy of CO in complex patients [35]. In patients with anemia and normal cardiac function, a CI/O₂ER ratio <10 suggests an inadequate CI that is likely due to hypovolemia [35].

As tissue oxygenation becomes inadequate, anaerobic metabolism begins to take over from aerobic metabolism and blood lactate levels rise. Although other factors can also result in increased blood lactate levels [36], a blood lactate level greater than 2 mEq/l suggests inadequate tissue perfusion and oxygenation. Hyperlactatemia is associated with a poor prognosis in critically ill patients in general and in those with sepsis [37]. As with many other measures, trends in lactate levels are of greater value than any individual value [38].

There is no ideal measure for determining optimal tissue oxygenation, and adequacy of DO₂ must be assessed using a combination of the above variables along with clinical examination.

With the advent of new techniques to monitor the microcirculation, several studies have now reported the effects of transfusion on the microcirculation in human subjects. In a small early study using orthogonal polarization spectral (OPS) imaging, Genzel-Boroviczény et al. reported an improvement in functional capillary density following transfusion in anemic preterm infants, indicating improved microvascular perfusion [39]. In patients undergoing on-pump cardiac surgery, Yuruk and colleagues [40] reported, using sidestream dark-field (SDF) imaging, that blood transfusion was associated with microcirculatory recruitment resulting in increased capillary density, thus reducing the oxygen diffusion distance to the cells. Using near-infrared spectroscopy (NIRS), the same authors reported that transfusion increased thenar and sublingual tissue oxygen saturation (StO₂) and thenar and sublingual tissue hemoglobin index (THI) in outpatients with chronic anemia [41]. In critically ill patients, Creteur et al., using the same NIRS technique, noted that blood transfusion was not associated with changes in muscle tissue oxygenation, VO₂, or microvascular reactivity in all patients, but that muscle VO₂ and microvascular reactivity did improve in patients in whom these variables were altered prior to the transfusion [42]. Similar findings have been made in patients with severe sepsis [43, 44] and trauma [45]. In a retrospective study of patients with severe sepsis who had a microdialysis catheter inserted for interstitial fluid measurements, blood transfusion was associated with a decrease in the interstitial lactate/pyruvate ratio, and these changes were again correlated with the pre-transfusion lactate/pyruvate ratio [46].

Studies have also assessed the impact of transfusion of fresh versus stored blood cell units on the microcirculation. In healthy volunteers, there were no differences in sublingual OPS-derived microcirculatory variables or NIRS-derived StO₂ after transfusion with 7-day or 42-day stored blood [47]. Walsh et al. reported no difference between stored (>20 days) and fresh (<5 days) cells on gastric tonometry indices in anemic critically ill patients [48], but Weinberg et al. reported a decrease in NIRS StO₂ and sidestream dark-field (SDF) capillary vascular density with transfusion of older units in trauma patients [49]. Some of the negative effects of blood transfusion may be due to the presence of leukocytes, and it has been proposed that use of leukodepleted blood should be preferred in critically ill patients. A recent pilot study in which 20 patients with sepsis were randomized to receive either leukodepleted or non-leukodepleted blood showed no clear superiority of leukodepleted over non-leukodepleted blood although leukodepleted blood was associated with more favorable changes in MFI and blood flow velocity [50].

We have seen that blood transfusion can improve DO₂ but may not directly help improve tissue oxygenation. Patients with sepsis frequently develop anemia [20], which is known to be associated with worse outcomes in critically ill patients [11, 51, 52], but are blood transfusions actually of benefit? When should critically ill patients with sepsis be transfused? Several early observational studies suggested worse outcomes in critically ill patients who received a transfusion compared to those who did not [6, 7], casting doubt on the supposed benefits of transfusion, but more recent studies have suggested the opposite [8, 11]. Some of these differences may be related to the timing of transfusion as benefits are likely to be greatest in the early stages of disease than in later phases when patients are stable or already have established organ failure [53]. Indeed, the Surviving Sepsis Campaign guidelines give different recommendations based on the duration of the septic episode: during the first 6 h of resuscitation, they suggest that transfusion should be given to maintain the hematocrit above 30 % if ScvO₂ remains below 70 % despite initial fluid and vasopressor therapy; this recommendation was, however, largely based on the Rivers study [33], so may need to be reconsidered in light of the ProCESS results [34]. After this initial period, the SSC guidelines recommend transfusion when the hemoglobin concentration is less than 7.0 g/dl to maintain a concentration of 7.0–9.0 g/dl (grade 1B). In certain circumstances, such as severe hypoxemia, ischemic coronary artery disease, or acute hemorrhage, higher thresholds may be warranted [54]. Guidelines from the British Committee for Standards in Hematology make similar recommendations [55].

The Transfusion Requirements in Critical Care (TRICC) study published in 1999 [56] changed many intensivists’ conceptions of blood transfusion, and physicians worldwide began to reconsider their transfusion thresholds [57], although one recent study suggested that transfusion rates only decreased in high-volume ICUs (>200 admissions per year) but continued to increase in low-volume hospitals [58]. Importantly, much has changed in intensive care since 1994–1997 when the TRICC study was conducted. Blood transfusion medicine has evolved so that blood transfusions are now safer. The general process of care has improved, and patients are being diagnosed and treated more rapidly with appropriate and effective resuscitation. So what new evidence is available on transfusion thresholds? There have been no large-scale studies comparing one transfusion trigger with another in a general population of critically ill patients since the TRICC study, and there are few specific data in septic patients. But there have been several studies comparing different thresholds in other groups of patients. A randomized controlled study in more than 500 patients undergoing cardiac surgery with cardiopulmonary bypass reported that a perioperative restrictive transfusion strategy (to maintain a hematocrit at least 24 %) was associated with similar morbidity/mortality outcomes compared to a more liberal strategy (to maintain a hematocrit of at least 30 %) [59]. In addition, regardless of the transfusion strategy, the number of transfused red blood cell units was an independent risk factor for clinical complications or death at 30 days (hazard ratio 1.21 for each additional unit transfused; 95 % confidence interval 1.1–1.4, P = .002). In a recent pilot study that included 100 elderly (>55 years), mechanically ventilated ICU patients, there was a trend to reduced mortality in patients managed using a restrictive (hemoglobin threshold 7.0 g/dl) compared to a more liberal (9.0 g/dl) strategy [60]. In a small randomized study in 44 patients with subarachnoid hemorrhage and high risk of vasospasm, Naidech et al. [61] reported that targeting a higher hemoglobin concentration (11.5 g/dl) was as safe as targeting a lower hemoglobin level (10 g/dl) and may have reduced the incidence of cortical cerebral infarction. In a randomized study of 2,016 patients ≥50 years of age with a history of or risk factors for cardiovascular disease after hip fracture surgery, a liberal transfusion strategy (hemoglobin threshold 10 g/dl) was not associated with reduced mortality or function at 60 days compared with a restrictive strategy (symptoms of anemia or physician discretion for a hemoglobin level of <8 g/dl) [62].