Aggressive Warming Reduces Blood Loss During Hip Arthroplasty

Abstract

We evaluated the effects of aggressive warming and maintenance of normothermia on surgical blood loss and allogeneic transfusion requirement. We randomly assigned 150 patients undergoing total hip arthroplasty with spinal anesthesia to aggressive warming (to maintain a tympanic membrane temperature of 36.5°C) or conventional warming (36°C). Autologous and allogeneic blood were given to maintain a priori designated hematocrits. Blood loss was determined by a blinded investigator based on sponge weight and scavenged cells; postoperative loss was determined from drain output. Results were analyzed on an intention-to-treat basis. Average intraoperative core temperatures were warmer in the patients assigned to aggressive warming (36.5° ± 0.3° vs 36.1° ± 0.3°C, P < 0.001). Mean arterial pressure was similar in each group preoperatively, but was greater intraoperatively in the conventionally warmed patients: 86 ± 12 vs 80 ± 9 mm Hg, P < 0.001. Intraoperative blood loss was significantly greater in the conventional warming (618 mL; interquartile range, 480–864 mL) than the aggressive warming group (488 mL; interquartile range, 368–721 mL;P = 0.002), whereas postoperative blood loss did not differ in the two groups. Total blood loss during surgery and over the first two postoperative days was also significantly greater in the conventional warming group (1678 mL; interquartile range, 1366–1965 mL) than in the aggressively warmed group (1,531 mL; interquartile range, 1055–1746 mL, P = 0.031). A total of 40 conventionally warmed patients required 86 units of allogeneic red blood cells, whereas 29 aggressively warmed patients required 62 units (P = 0.051 and 0.061, respectively). We conclude that aggressive intraoperative warming reduces blood loss during hip arthroplasty.

Implications 

Aggressive warming better maintained core temperature (36.5° vs 36.1°C) and slightly decreased intraoperative blood pressure. Aggressive warming also decreased blood loss by approximately 200 mL. Aggressive warming may thus, be beneficial in patients undergoing hip arthroplasty.

Perioperative hypothermia is a common but preventable complication of anesthesia and surgery (¹). Mild perioperative hypothermia increases the incidence of morbid myocardial outcomes (²), reduces resistance to surgical wound infection (³), and prolongs both postanesthetic recovery and hospitalization (⁴).

An additional complication resulting from mild hypothermia is increased blood loss (⁵). Coagulopathy results, in part, from inhibition of platelet function (⁶), which is apparently related to defective thromboxane A₂ release, up-regulation of platelet surface protein GMP-140, and down-regulation of platelet glycoprotein Ib-IX complex (⁷).

Despite ample evidence that mild hypothermia (1.4°–1.9°C) provokes numerous adverse outcomes, the definition of intraoperative normothermia is controversial. Some authors indicate that patient temperatures exceeding 35°C are adequate, whereas others suggest temperature should be maintained >36°C (⁸). Neither value, though, appears to be based on thermoregulatory physiology, because normal body temperature averages 37°C and is rarely <36.5°C (⁹). We, therefore, tested the hypothesis that prevention of very mild hypothermia by aggressive warming and maintenance of normothermia during total hip arthroplasty with spinal anesthesia reduces surgical blood loss.

Methods

The study protocol was approved by the institutional review board at the University of Vienna and written, informed consent was obtained from each participating patient. In a preliminary study of 55 hip arthroplasty patients, 0.5°C of core hypothermia increased blood loss by 250 mL with a standard deviation of approximately 500 mL (¹⁰). We thus studied 150 patients because a sample-size estimate indicated that this number would provide a 90% chance of identifying a significant hypothermia-induced increase in blood loss at a one-tailed α level of 0.05. Eligible patients were scheduled to undergo primary, unilateral, cement-free total hip arthroplasty. None of the arthroplasties was performed for treatment of tumor.

Patients were eligible for the study when they consented to regional anesthesia, were ASA physical status I–III, ages 40 to 80 yr, and weighed between 50 and 100 kg. Patients were excluded when preoperative coagulation tests were abnormal, which we defined as a partial thromboplastin time exceeding 40 s (normal 28–40 s), a prothrombin time <70% of normal (normal 70%–140%), platelet count <100,000/μL, or fibrinogen <200 mg/dL. Patients who had consumed aspirin products within a week of surgery, or who had a history of bleeding disorders, deep-venous thrombosis, or pulmonary embolism were also excluded. Roughly one-half (n = 72) of the enrolled patients predeposited 2–4 units of autologous blood the month before scheduled surgery and were given oral iron until the day before surgery. Autologous donation was per surgical routine and patient preference and was not controlled for study purposes.

This study was conducted over a 2-yr period. All patients were given perioperative thromboprophylaxis with either 20–40 mg Enoxaparin (Gerot Pharmazeutika, Vienna, Austria) or 5000 international units Dalteparin (Pharmacia AB, Stockholm, Sweden) subcutaneously. The first dose was given the evening before surgery and was then continued on a daily (self-injected) basis until 6 wk postoperatively. Hip surgery was performed in the supine position, by using a lateral transgluteal approach. All patients were given midazolam 7.5 mg orally the night before surgery, and again approximately 2 h before surgery.

Operating room temperature was maintained near 23°C. IV and radial arterial cannulae were inserted. Combined spinal/epidural anesthesia was performed via the L3-4 interspace (CSE MAXI-SET^®; Medimex Holfeld, Hamburg, Germany). Isobaric 0.5% bupivacaine 2.8–3.5 mL was intrathecally injected to establish a sensory block between the T-8 and T-10 vertebrae. Reductions in systolic blood pressure exceeding 30% of preinduction values were treated with IV volume expansion and, if necessary, 1–2 mg IV etilefrine. Otherwise, blood pressure was not actively managed. IV administered midazolam (1–2 mg) and propofol (2 mg · kg⁻¹ · h⁻¹) were used for intraoperative sedation. All patients breathed supplemental oxygen via a face mask.

After the induction of anesthesia, patients were randomly assigned to either aggressive warming to maintain normothermia (core temperature near 36.5°C) or conventional warming (core temperature near 36.0°C). We used computer-generated randomization numbers that were kept in sealed envelopes until informed consent was obtained and regional anesthesia was successfully performed. All patients were actively warmed with upper- and lower-body forced-air covers connected to individual forced-air heaters (Bair-Hugger^®; Augustine Medical, Eden Prairie, MN). Temperature of the warmers was adjusted, as necessary, to maintain the designated target core temperatures. Additionally, all IV fluids were warmed to 37°C.

As in previous studies (⁵), target minimum hematocrits were prospectively determined for each patient, based on age and cardiovascular status. The target hematocrit was 26% in patients aged <65 yr having no significant cardiovascular disease. The hematocrit was allowed to decrease to 28% in patients aged >65 yr or having cardiovascular disease. Significant cardiovascular disease was defined as previous myocardial infarction, angina, congestive heart failure, cardiomyopathy, hypertension (a diastolic blood pressure exceeding 90 mm Hg or requiring chronic drug treatment), or peripheral vascular disease. Hematocrit was maintained >30% in patients having both cardiovascular disease and an age >65 yr.

Intraoperative blood loss was scavenged by using a Shiley Stat^® (Dideco, Mirandola Modena, Italy) or a Cell Saver^® (Haemometics, Braintree, MA) autotransfusion system primed with 30,000 international units of heparin in 1000 mL of normal saline, of which patients received approximately 150 mL. Patients were given 10 mL/kg of crystalloid during the induction of spinal anesthesia. Crystalloid solutions were then infused at a rate of 5 mL · kg⁻¹ · h⁻¹ throughout surgery; additional crystalloid was given to replace the first 400 mL of estimated blood loss at a ratio of 3 mL/ml. Additional blood loss was replaced with a 3.5% gelatin volume-expanding solution (molecular weight, 30 000; pH, 7.3 ± 0.3; 301 mOsm/L; 293 mOsm/kg) (Haemaccel; Behring Werke, Marburg, Germany) at a ratio of 1:1 (mL). All scavenged blood was autotransfused near the end of surgery, irrespective of the hematocrit. Autologous-packed red blood cells were transfused as necessary to maintain the target hematocrits previously listed. Units of allogeneic cells were transfused as necessary to patients in whom autologous blood was unavailable or when autologous blood was exhausted.

The epidural catheter was used for postoperative pain relief. After a test dose (3 mL of 0.25% bupivacaine with epinephrine 1:200,000), 6–8 mL of 0.125% bupivacaine was injected. Subsequently, a continuous epidural infusion was maintained with 0.125% bupivacaine, 4 μg/mL fentanyl, and 3 μg/mL clonidine. This combination was infused at a rate of 3–5 mL/h until the second postoperative day.

Core temperatures were recorded from the tympanic membrane. The aural probes were inserted until patients felt the thermocouple touch the tympanic membrane; appropriate placement was confirmed when patients easily detected a gentle rubbing of the attached wire. The aural canal was occluded with cotton, the probe securely taped in place, and a gauze bandage positioned over the external ear. Core temperature was also recorded from the urinary bladder. Mean skin temperature was recorded from the weighted average of four sites (¹¹). These temperatures were recorded from Mon-a-Therm^® thermocouples (Mallinckrodt Anesthesiology Products, St. Louis, MO) at 20-min intervals throughout surgery and for the first three postoperative hours.

Prothrombin and partial thromboplastin times, fibrinogen, antithrombin III, platelet, and hemoglobin concentrations were determined by the clinical laboratory preoperatively, immediately after surgery, and on the first and second postoperative day. Coagulation tests were performed at 37°C. Hemodynamic variables, hemoglobin, hematocrit, fluid balance, and transfusion requirements were tabulated at 20-min intervals throughout surgery, and during the first three postoperative hours. Blood hemoglobin concentration was determined immediately after surgery, as well as on the first and second postoperative day. Cumulative blood loss and transfusion requirements were then determined after 6 h, and on the first and second postoperative days.

Intraoperative blood loss was estimated by combining changes in sponge weights (assuming a density of 1 gm/mL) with scavenged blood volume. Blood volume aspirated by the scavenging system was estimated by multiplying the hematocrit of the scavenged blood by its volume, and dividing by the average hematocrit of the patient over the relevant time period. Cell-scavenger was not used in an auto set-up mode. As soon as the suction-reservoir volume reached 700 mL (regardless of how concentrated the content appeared), the separation cycles were manually initiated. Postoperative blood loss was recorded from the wound drains 3 and 6 h after surgery, and on the first and second postoperative mornings. The surgeons were blinded to group assignment and perioperative core temperature, as were the observers who weighed gauze-sponges and calculated blood recovered by a red-blood-cell scavenging system.

Results were analyzed after completion of data collection and an audit confirming integrity of the randomization process. An intention-to-treat analysis was used. Thus, patients were considered to be in their assigned groups even when target temperatures were not reached (¹²).

Ambient temperature and hemodynamic responses were first averaged over the operative period within each patient, and then, averaged among the patients in each treatment group. Potential confounding factors in the two treatment groups were compared by using unpaired, two-tailed t-tests. Blood loss and allogeneic transfusion requirements were compared by using unpaired, one-tailed Wilcoxon or χ² tests. Our two prospective major outcomes were 48-h blood loss and allogeneic transfusion requirement. Our decision to use a one-tailed analysis for these two major outcomes was determined a priori and based on preliminary data and a sample-size estimate. Data were presented as mean ± sd or median and interquartile range;P < 0.05 identified statistically significant differences.

Results

Eight patients assigned to conventional warming had mean intraoperative core temperatures ≥ 36.5°C; similarly, four patients assigned to aggressive warming had mean intraoperative core temperatures ≤ 36.0°C. Data from these patients were included in the analysis on the basis of their intended treatments. One conventionally warmed patient returned emergently to the operating room after several hours of recovery because of a surgical complication. His data from the initial surgery were included in the analysis; however, his postoperative data were not. All other patients were treated per randomization and completed the study. Six different surgeons were involved in the study, and the distribution of their patients was similar in each treatment group.

Morphometric characteristics, amount of predonated autologous blood, duration of surgery, and preoperative coagulation variables were comparable in the two groups. Preoperative anticoagulant use (Enoxaparin and Dalteparin) were also similar in the groups (P = 0.675). Prothrombin times, partial thromboplastin times, blood fibrinogen, and antithrombin III concentrations were normal, preoperatively, in both groups (Table 1).

Table 1: Morphometric Characteristics, Duration of Surgery, and Preoperative Coagulation Profile

Despite randomization, target minimum hematocrit values were significantly greater in the aggressive warming (normothermic) than conventional warming (hypothermic) group. Preoperative mean arterial pressure and heart rate did not differ significantly in the two groups. Intraoperative mean arterial pressure was significantly less in the aggressive warming than conventional warming group (80 ± 9 vs 86 ± 12 mm Hg, P < 0.001). In contrast, heart rate was significantly greater in the aggressive warming patients (75 ± 12 vs 71 ± 12 bpm, P = 0.043). Ambient temperatures were slightly, but significantly greater in the aggressive warming patients (22.6 ± 1.0 vs 23.1 ± 1.1°C, P = 0.004).

By design, average intraoperative core temperatures were approximately 0.5°C warmer in the patients assigned to aggressive warming (36.5 ± 0.3 vs 36.1 ± 0.3°C, P < 0.001). Three hours postoperatively, core temperatures remained significantly higher in the aggressive warming group (37.1 ± 0.7 vs 36.8 ± 0.6°C, P = 0.005). Mean skin temperature was nearly 1°C higher in the aggressive warming group (33.2 ± 1.2 vs 32.4 ± 1.1°C, P = 0.015, Table 2).

Table 2: Hemodynamic Responses and Temperatures

Intraoperative blood loss was significantly greater in the conventional warming (618 mL; interquartile range, 480–864 mL) than the aggressive warming group (488 mL; interquartile range, 368–721 mL;P = 0.002), whereas, postoperative blood loss did not differ in the two groups. Total blood loss during surgery and over the first two postoperative days was also significantly greater in the conventional warming group (1,678 mL; interquartile range, 1,366–1,965 mL) than in the aggressively warmed group (1,531 mL; interquartile range, 1,055–1,746 mL, P = 0.031). The 40 conventionally warmed patients required 86 units of allogeneic red blood cells whereas, 29 aggressively warmed patients required 62 units (P = 0.051 and 0.061, respectively, for number of patients and units, Table 3).

Table 3: Hemoglobin, Blood Loss, Fluids, and Allogeneic Transfusions

Discussion

The normal circadian body temperature variation is roughly 1°C (¹³). However, body temperature at any given time of day is regulated to within a few tenths of a degree (^14,15), with the peak value being approximately 37.5°C in the midafternoon and the minimum temperature being approximately 36.5°C near 3:00 am (¹³). Our target temperature for the aggressive warming group was thus, near the typical early morning value, although the conventional warming group target was only 0.5°C less. This small difference, only one-half the normal circadian variation, nonetheless increased surgical blood loss by approximately 200 mL in the conventionally warmed (hypothermic) patients (P = 0.031). We thus, confirmed our first hypothesis that very mild hypothermia increases surgical blood loss.

Consistent with increased blood loss, more allogeneic transfusions were required in the conventionally warmed (hypothermic) patients (86 units in 40 patients) than in those assigned to aggressive warming (normothermia) (62 units in 29 patients). Most of this blood was transfused between the first and second postoperative mornings. However, the difference in transfusion requirement was not statistically significant, with P = 0.051 and 0.061, respectively, for the number of patients and units. Failure to find a statistically significant difference in allogeneic blood loss might represent a Type II statistical error and indicate that our study was simply under-powered for this outcome. An alternative explanation is that the target minimum hematocrits were higher in the aggressively warmed patients. They were thus a priori more likely to have been given transfusions.

A possible explanation for the observed increase in blood loss is the well established effect of hypothermia on platelet function (^6,7). The reduction in the current patients (approximately 200 mL with an 0.5°C difference) was roughly in proportion to the approximately 500 mL we observed previously with a core temperature difference of 1.6°C (⁵). However, we must consider another mechanism, specifically the effect of vigorous cutaneous warming on blood pressure.

The thermoregulatory system controls two distinct vascular responses. The first consists of arteriovenous shunts which are largely restricted to fingers and toes. These α-adrenergically controlled 100-μm vessels transmit 10,000 times as much blood as a comparable length of 10-μm capillary (¹⁶). The shunts constrict during cold stress, but are fully dilated at normothermia and do not dilate further even during excessive heating. The second thermoregulatory vascular response is cutaneous hyperemia. Cutaneous capillary flow is controlled by precapillary arterioles and mediated by a yet-to-be-determined factor released from sweat glands in response to cholinergic stimulation (¹⁷). Both shunt and capillary flow is largely under central control; however, both types of vessel are also mediated by local skin temperature (^17,18).

It is likely that cutaneous warming—which was necessarily more aggressive in patients assigned to normothermia—triggered considerable locally mediated capillary hyperemia. This theory is supported by our observation that the intraoperative mean arterial pressure was significantly less in the aggressively warmed patients, although their blood pressure values were virtually identical preoperatively and did not differ significantly postoperatively.

The importance of the 6-mm Hg intraoperative blood pressure difference in the two groups is that hypotension is a well established method of reducing blood loss. For example, reducing mean-arterial pressure 10 mm Hg decreases blood loss 30% in patients undergoing hip arthroplasty (¹⁹). It is thus, possible that the beneficial effect of maintaining intraoperative normothermia resulted, at least, in part, from locally mediated capillary hyperemia and the consequent relative hypotension. To the extent that this mechanism dominates, a comparable pharmacologic reduction in blood pressure may provide similar benefit.

Intraoperative hypothermia results from an internal core-to-peripheral redistribution of body heat and exposure to a cool environment (²⁰). Passive insulation decreases heat loss only 30%–50% (^21,22) and thus, is unable to compensate for large intraoperative cutaneous and evaporative losses (²³). Consequently, nearly all unwarmed surgical patients become hypothermic (²⁴). Minimum core temperatures are typically near 35°C after total hip arthroplasty (⁵), and initial postoperative temperatures across a wide variety of surgical procedures average 35.6°C (²⁵). Active warming was thus, required in both our treatment groups.

Ambient temperatures in the two treatment groups differed significantly because the extra heating in the normothermic patients also heated the room. However, the difference was only 0.5°C, an amount that did not unblind the study. The total blood loss in our hip arthroplasty patients is hardly unusual, especially because the type of arthroplasty performed on our patients is associated with more bleeding than other approaches (²⁶). Well over one-half of all blood loss was postoperative and we have previously reported similar values from another institution (⁵). Others (²⁷) report decreased losses, presumably the result of differing surgical technique and various strategies to reduce loss.

Our randomized, prospective data suggest that aggressive warming and maintenance of normothermia markedly influence blood loss. However, our current and previous (^5,28) results contrast with a recent prospective, randomized study in which a 1°C difference in intraoperative temperature was not associated with increased blood loss during hip arthroplasty (²⁹). Why these results differ is unclear, but may be a consequence of the populations studied or surgical or anesthetic techniques.

In summary, 150 patients undergoing hip arthroplasty were randomly assigned to aggressive warming to maintain normothermia (approximately 36.5°C) or conventional warming (approximately 36.0°C). The conventionally warmed (hypothermic) patients lost significantly more blood (1,678 mL; interquartile range, 1,366–1,965 mL) than the aggressively warmed group (1,531 mL; interquartile range, 1,055–1,746 mL, P = 0.031). This approximately 200 mL difference in blood loss may have resulted from the combined effects of reduced platelet function and impaired coagulation enzymes in the conventionally warmed patients and locally mediated hyperemia of cutaneous capillaries with consequent relative hypotension in the aggressively warmed patients. Our data suggest that aggressive warming to maintain normothermia (approximately 36.5°C) produces a clinically important reduction in blood loss during hip arthroplasty.