Upregulation of Prostaglandin E²and Interleukins in the Central Nervous System and Peripheral Tissue during and after Surgery in Humans

After approval of the Institutional Review Board of Rush University Medical Center, Chicago, Illinois, consecutive osteoarthritis patients between the ages of 55 and 75 yr scheduled to undergo elective primary total hip arthroplasty by a single orthopedic surgeon were contacted and assessed for study eligibility with a screening medical history. A study consent form was sent to patients who agreed to participate in the clinical study. After obtaining consent, the patient was allocated a study number and the study drug was dispensed to each participant. Using a random number table, patients were allocated to one of three groups, without stratification by demographic characteristics: group 1 (n = 10; control) received a placebo each morning for 4 days before surgery and on the day of surgery; group 2 (n = 10; 1-day COX-2 inhibitor) received placebo each morning for 4 days before surgery and on the day of surgery received oral rofecoxib 50 mg; group 3 (n = 10; 5-day COX-2 inhibitor) received oral rofecoxib 50 mg each morning for 4 days before surgery and on the day of surgery. Study drug administration on the day of surgery was timed to precede surgical incision by 90 ± 30 min. All prior nonsteroidal antiinflammatory drug therapy was discontinued 14 days before surgery.

The physicians and nurses managing the patient during surgery and in the recovery room, the personnel involved with postoperative pain assessment and management of the intrathecal infusion, and the study patients were blinded to group assignments. Treatment assignment codes were not available to the investigators until all patients completed the study. A global pain score using the visual analog scale (VAS) with 0 corresponding to no pain and 10 to the worst imaginable pain for the patient was obtained before surgery.5In the operating room, patients were sedated with midazolam (0.05 mg·kg^–1titrated to effect) and an intrathecal catheter was placed in the sitting position, at the L3–L4 or L4–L5 vertebral level to deliver spinal anesthesia with bupivacaine 0.5% (7.5 mg) and fentanyl 25 μg. Before administering spinal anesthesia, 0.5 ml CSF was removed with simultaneous sampling of venous blood (5 ml) for the analysis of PGE²and cytokines. Patients were maintained at normothermia in the operating room.19A sensory analgesic level of T10 was obtained before commencement of surgery. At completion of surgery an intrathecal infusion of fentanyl 0.5 μg·ml^–1and bupivacaine 0.1 mg ml^–1was initiated16using a continuous basal infusion with superimposed patient controlled intrathecal analgesia bolus doses. Initial infusion rates were 4 ml·h^–1basal intrathecal infusion plus patient controlled intrathecal analgesia of 1 ml every 12 min with a 4-h lockout of 40 ml. The patients were instructed before surgery to use patient-controlled intrathecal analgesia at their discretion to maintain the VAS pain score of less than 4, following a previously applied protocol.18A standardized surgical technique of noncemented hip arthroplasty performed through a mini-incision (<10 cm long) anterolateral approach was used in all patients.

The intrathecal catheter was left in place for 30 h after surgery to provide analgesia. CSF (0.5 ml) and blood (5 ml in plasma collection tubes) were collected at the time of intrathecal catheter placement and also at surgical incision for baseline measurements (to act as control for intrathecal anesthesia). In addition, CSF and blood were collected at: 1, 3, 9, 24, and 30 h after incision for measurements of PGE², interleukin 1β (IL-1β), IL-6, interleukin 8 (IL-8), and tumor necrosis factor α (TNF-α). In a preliminary study, CSF interleukin 10 was undetectable during the postoperative period, and that cytokine therefore was excluded from the assays. When the hip replacement was complete, a drain was placed in the hip wound and the exudates were collected in a Snyder 400-ml capacity reservoir. Samples from the hip drain exudate system were collected at 3, 9, 24, and 30 h after incision. One hour before scheduled collection time, the hip drain reservoir was emptied and the exudate was collected over the next 60 min. Drain fluids and blood were centrifuged and the supernatant removed. These fluids and CSF were frozen at –80°C for subsequent assay.

Prostaglandin E²and cytokine concentrations in samples of plasma, hip drain supernatant, and CSF were measured by sandwich enzyme-linked immunosorbent assay in 96-well microtitration plates following the manufacturer’s protocols. Enzyme-linked immunosorbent assay kits were obtained from Assay Designs (Ann Arbor, MI) for PGE²and R&D Systems (Minneapolis, MN) for the cytokines IL-6, IL-8, IL-1β, and TNF-α. All samples were assayed in duplicate. The detection limits for the assays in pg·ml^–1were: PGE², 13.4; IL-6, 9.4; IL-8, 25.0; IL-1β, 3.9; and TNF-α, 15.6. For all assays, the intraday and interday coefficients of variation were both less than 10%.

Visual analog scores beginning at 3, 9, 24, and 30 h after surgical incision and total intrathecal medication consumption were recorded. Patients rated sleep disturbance during the night after surgery (0 = no sleep disturbance to 10 = greatest sleep disturbance).18Functional outcome was assessed by the time to ambulate 10 and 25 m with standardized walking aids,20time to achieve milestones such as transfer unassisted in and out of bed, ability to ambulate with a walker unassisted on a level surface, and the ability to ascend and descend four steps using a railing.21Tympanic temperature was monitored before surgery and at each blood and CSF measurement time point after surgical incision.

A sample size of 16 patients per group was required to provide 90% power to detect differences over time between groups with respect to changes in CSF PGE²at an α= 0.05. The trial was terminated prematurely when the study drug was withdrawn from the market. However, 30 patients completed the study protocol (10 per group). One patient from each group subsequently was dropped from the study because of consistent inability to sample CSF. Analyses were performed using repeated measures analysis with preplanned contrasts for both trend and between-subject factors. For statistical analysis, all assay results that were below the level of detection were replaced by the level of detection. Maximum-likelihood estimates and corresponding P  values were generated using the mixed regression procedure in the Statistical Analysis System package version 8.2 (SAS, Cary, NC). The Dunnett-Hsu procedure was used to control for type I family-wise error rate for multiple comparisons. In particular, in analyzing changes in variables over time (e.g. , CSF PGE²), later time-point concentrations were compared with the 0-h baseline; and in comparing differences among groups in the COX-2 modulation study, each COX-2 dose group was compared with placebo. All tests were two-sided, and an α value of 0.05 was chosen as the threshold for significance. Continuous descriptive measures are reported as mean and standard deviation. Demographic measures were tested for between-group differences using one-way ANOVAs or chi-square tests, depending on the scale of the measure. Correlations between variables were performed by Pearson correlation coefficient of area under the curve (AUC) up to 30 h. An autoregressive time series model was used to evaluate relationships over time. Concentration-time slopes were calculated as changes in concentration divided by change in time. Regression coefficients (slopes) were estimated using the mixed regression procedure, and family-wise error rate was controlled using the stepdown Bonferroni procedure.

In control (placebo) patients, PGE²was increased in CSF at all measurement points from 9 through 30 h from the start of surgery (fig. 1). There was also an increase in CSF IL-6 that was measured at 3 h and persisted through 30 h. CSF IL-8 was increased only at the 9-h measurement. CSF IL-1β and TNF-α concentrations were below the detection limit of the assay at baseline and during the postoperative period. Differences in the time course of the CSF PGE²and cytokine levels were characterized by significance of the CSF concentration-time slopes (table 1). Over the 0- to 3-h period, only IL-6 had a significant positive concentration-time slope. At 3 to 9 h and 9 to 24 h, only PGE²showed significant positive CSF concentration-time slope estimates.

In control patients, local tissue (surgical site hip drain) PGE², IL-6, and IL-8 were increased from the 3-h initial sampling time at all measurement points from 9 through 30 h (fig. 2). Hip drain IL-1β increased at 24 and 30 h measurement points. However, the local tissue TNF-α was decreased at 24 and 30 h. In plasma, the only changes in concentration were an increase in PGE²at 1 h and a late increase in IL-6 at 24 and 30 h (fig. 3). Plasma IL-1β and TNF-α concentrations were below the detection limit at baseline and during the postoperative period. The hip drain concentrations of PGE², IL-6, IL-8, and IL-1β were greater than the plasma concentrations at 9 h and at all subsequent measurement points. Hip drain TNF-α was greater than plasma TNF-α at 3 and 9 h.

When three groups of patients are considered, there is a reduction of CSF PGE²by the COX-2 inhibitor with either dose regimen (fig. 4A). The 5-day dosing decreased CSF PGE²levels at 9, 24, and 30 h compared with the placebo patients, whereas the single dose reduced the PGE²level at the 9-h measurement. Surgical site PGE²also was reduced by either dosing regimen of COX-2 inhibitor (fig. 4B), but there was no COX-2 inhibitor modulation of plasma PGE².

CSF IL-6 concentration was reduced at the 3, 9, and 24 h measurement points by 5-day dosing of COX-2 inhibitor, and at 3 and 24 h with the 1-day dose (fig. 5). Surgical site IL-6 was not reduced by either dose of COX-2 inhibitor, nor was plasma IL-6. IL-8 concentrations were not modulated by COX-2 inhibitor in the CSF, hip drain, or in the plasma, and IL-1β and TNF-α hip drain concentrations were not COX-2 inhibitor modulated.

Demographic characteristics did not differ among the three groups of patients (table 2) and none of the patients received blood transfusion during the time of the study. COX-2 inhibitor administration, with either dosing regimen, reduced postoperative pain at 9-, 24-, and 30-h assessment points (fig. 6). In addition, 24 h from the start of surgery, intrathecal drug consumption for postoperative pain was reduced in the 5-day COX-2 inhibitor group (79.1 ± 18.5 ml) versus  the placebo groups (106.9 ± 16.5 ml), although the 1-day COX-2 inhibitor group (98.9 ± 12.3 ml) was not different from placebo. The VAS for pain was not significantly correlated with any of the PGE²or cytokine levels in the placebo patients. However, when all the three groups were combined, over the 30-h postoperative time period, VAS was positively correlated with CSF PGE²level (P = 0.0043). Similarly, CSF IL-6 concentration was correlated positively with VAS (P = 0.0057). CSF IL-6 also was correlated positively with postoperative day 1 sleep disturbance, both for the placebo group (P = 0.0482) and all three groups combined (P = 0.0270). CSF IL-8 also was correlated positively with sleep disturbance, but only for the three combined groups (P = 0.0330).

There was a positive correlation between poorer functional recovery and increased surgical site PGE²concentration measured at the 30-h time point. The decreased performance during physical therapy included increased time to ambulate 10 and 25 m (P < 0.01), to get out of bed (P = 0.03), and to walk up steps (P = 0.0003). Finally, there was no correlation of body temperature with any measured variable, including CSF PGE². All patients’ body temperatures were less than 37.1°C before surgery and 37.2°C after surgery.

The upregulation of PGE²in CSF, documented at 9 through 30 h after the start of hip replacement surgery, is similar to findings in animal models of peripheral inflammation. In rats with unilateral hind paw inflammation, CSF PGE²increased beginning at 6 h and peaked at 12 h after injection of complete Freund’s adjuvant.13After intraarticular injections of kaolin and carrageenan in the rat knee, there was a pronounced intraspinal release of PGE²at 7 to 9 h after injection.7In both animal models, the increase in PGE²was accompanied by an upregulation of spinal COX-2 protein. CSF IL-6 was increased at 3 h from the start of hip surgery and remained elevated at 24 h. This is similar to another hip replacement study in which a single CSF IL-6 level obtained on postoperative day 1 was increased.22The increase in CSF IL-6 at 24 h in our study was 283 pg·ml^–1, whereas in the latter study, the CSF IL-6 increase was approximately 150 pg·ml^–1, both under spinal anesthesia. In that study, CSF IL-10 did not change at the single measured time point, which matches our preliminary findings (see Materials and Methods). CSF IL-8 increased in our study, although it was only at the 9-h measurement. CSF IL-8 has also been shown to be increased after major surgery, such as thoracoabdominal aortic reconstruction.23Surprisingly, CSF levels of IL-1β and TNF-α did not increase above the detection limits in our patients throughout the postoperative period. These IL-1β results are different from a study in rats with peripheral inflammation, in which there was a large upregulation of CSF IL-1β, starting at 2 h after insult.13In patients undergoing lower-extremity revascularization under spinal anesthesia, the immediate postoperative CSF TNF-α level increased only slightly.24However, after thoracoabdominal aortic surgery under general anesthesia, there was no increase in CSF levels of IL-1β, TNF-α, or IL-6.23The latter results may be the result of suppression of central inflammatory responses by general anesthesia. However, different types of surgery may generate different central nervous system inflammatory responses. In this hip surgery study, an analysis of concentration-time slopes suggests that in the CSF there is an initial increase in IL-6 that is followed by an increase in PGE². Topical IL-6 application has been shown to increase PGE²production in the central nervous system both in vivo  25and in tissue culture.26,27 

After hip replacement surgery, hip drain fluid evidenced a postoperative increase in PGE²after 3 h from the time of surgical incision. In dental patients, PGE²at the surgical site increased starting 1 to 2 h after molar extraction.5,6In our study, hip drain IL-6, IL-8, and IL-1β levels also increased after 3 h, but TNF-α decreased. This is similar to a previous hip replacement study under spinal anesthesia in which hip drain IL-6 and IL-1 increased during the postoperative period, but TNF-α did not.10After reduction mammoplasty in patients, wound fluid IL-6 increased after surgery and there was a trend toward increased IL-8 levels.11Because the first hip drain samples were obtained 3 h from the surgical incision in this study, we are unable to conclude definitively whether upregulation of any of the cytokines preceded the increase in local PGE²production.

The only changes we observed in the plasma levels were an initial increase in PGE²and a late increase in IL-6. Several studies have shown an increase in blood IL-6 levels after hip surgery.9,10,22,28–30Some of these studies also measured blood levels of IL-1β and TNF-α10,29and showed no increase, which is consistent with our study. Our circulating concentrations of PGE², IL-6, IL-8, and IL-1β were all less than the hip drain levels at 9, 24, and 30 h after surgery, and the plasma level of TNF-α was less than the hip drain at 3 and 9 h. At the 24-h measurement, hip drain levels were approximately two orders of magnitude more than the corresponding plasma levels of PGE², IL-6, or IL-8. In a similar hip surgery study, it was also noted that concentrations of cytokines were higher in wound drainage fluid than in the systemic circulation.10This suggests that hip drain fluid levels are not plasma derived but instead represent a local inflammatory response. Thus, it is not clear what role circulating cytokines, except possibly IL-6, play in the systemic response to hip replacement surgery. This is unlike the human response to endotoxin injection, where circulating levels of TNF-α, and then IL-6 and IL-8, rapidly increase.31The plasma levels of PGE², IL-6, and IL-8 were not correlated with the corresponding CSF levels that were measured in our patients, and so there is no indication that the CSF levels represent diffusion from the blood. This is consistent with animal studies showing that PGE²and IL-6 cross the blood–brain barrier poorly.32,33Mechanisms have been proposed by which circulating cytokines can trigger brain endothelial cells to produce PGE², which then can diffuse into the central nervous system.34Thus, the rise in blood IL-6 levels observed at 24 and 30 h may have contributed to the persistent increase in CSF PGE²at those times.

Animal models of peripheral inflammation have demonstrated an upregulation of spinal cord COX-27,13–15and subsequent increase in central PGE²levels.7,13Surgical incision also can increase COX-2 protein levels in the rat spinal cord.35An upregulation of both CSF and surgical site PGE²was observed in the current study, and the question arises whether those levels can be reduced by COX-inhibitors. Because of the risk of increased bleeding during surgery, nonsteroidal antiinflammatory drugs, which are primarily COX-1 inhibitors, were not tested. Rofecoxib, the COX-2 inhibitor used in the present study, has good penetration into the central nervous system after oral dosing,36and that may allow direct central COX-2 inhibition, in addition to the expected peripheral COX-2 inhibition.

The present study demonstrates that COX-2 inhibition reduces CSF PGE²over the 9- to 30-h period from start of surgery, and the effect was more pronounced with multiple (5-day) dosing versus  a single dose. Our previous clinical, pharmacokinetic analysis showed that such multiple daily dosing doubles both the plasma and CSF levels of rofecoxib compared with a single dose.36Hip drain PGE²was suppressed by COX-2 inhibition and the single dose was as effective as multiple dosing. In an oral surgery model, pretreatment with another COX-2 inhibitor, celecoxib, suppressed local tissue PGE²at 80 to 240 min from the end of surgery.37Interestingly, the COX-2 inhibitor also decreased postoperative CSF IL-6 levels in our hip surgery patients. PGE²has been demonstrated to stimulate IL-6 production in various in vitro  studies: mouse macrophages,38rat leukocytes,39human astrocytoma,40osteoblast-like cells,41and rat astrocytes.42In human astrocytoma cells, PGE²stimulation increased IL-6 mRNA levels at 30 min, and maximal mRNA levels were achieved 60 min after PGE²stimulation.40In mouse macrophages, a cause-and-effect relationship (60-min delay) was shown between increased PGE²and IL-6 production, and indomethacin inhibited IL-6 mRNA expression.38Therefore, a plausible explanation for the reduced CSF IL-6 levels seen with administration of a COX-2 inhibitor is by reduction of CSF PGE²and less subsequent stimulation of central IL-6 production.

In the present study, COX-2 inhibition with single or multiple doses reduced postoperative pain, an observation noted in an earlier knee replacement surgery study.18In the current study, we demonstrate a correlation between CSF PGE²and VAS over the 30-h monitoring period. Peripheral PGE²sensitizes nerve endings and contributes to pain,43intrathecal PGE²is directly nociceptive in rodents,44and PGE²applied to spinal cord slices depolarizes spinal neurons.45When our three study groups are combined, the VAS was correlated positively with CSF IL-6 levels. CSF IL-6 concentration correlated with sleep disturbance, both for the control patients and all groups combined. Experiments in rats with central IL-6 administration have shown an initial enhancement of nonrapid eye movement sleep, after which sleep was suppressed.46 

Increase in surgical site PGE²was correlated with poorer functional recovery, including increased time to begin ambulating, get out of bed, walk up steps, and walk 10 and 25 m. This is consistent with our previously published study in patients undergoing total knee replacement: when COX-2 inhibitor was administered before surgery, there was an improvement in the functional outcome.18There was suppression of the hip drain PGE²with COX-2 inhibition in the present study, with improved outcome. We did not find a correlation between plasma IL-6 and functional recovery unlike another hip surgery study.20Fever has been related directly to brain levels of PGE².47In our study, there was no correlation between CSF PGE²and body temperature. However, no patient had a temperature of more than 37.2°C. After total knee arthroplasty, patients who were febrile had higher knee drain and serum IL-6 levels than afebrile patients.48Because we did not have any febrile patients, it is not surprising that there was no correlation of body temperature with any of the biochemical measures in our study.

In addition to the consideration that other types of major surgery may evoke a different spectrum of inflammatory responses than hip replacement surgery, there is another factor that may limit generalization of our results. All of our patients received spinal anesthesia with an indwelling intrathecal catheter. In a similar hip surgery study, with patients receiving general or spinal anesthesia, CSF IL-6 levels on postoperative day 1 were more variable in the latter group, raising the possibility that spinal anesthesia led to neuroaxial irritation.22However, we did not observe an acute change in any of our CSF measures, in the interval between placement of the spinal catheter and the initial surgical incision. Still, some of our measures may be secondary to the local anesthetic dosing. In rodents, general anesthesia can suppress the initial CSF PGE²inflammatory response to surgery,49and if this effect is also demonstrated in humans, that may explain anesthesia-related differences in CSF IL-6.22,23 

The primary conclusion of our study is that hip replacement surgery is associated with postoperative upregulation of PGE², IL-6, and IL-8, both centrally and at the surgical site. CSF PGE²is associated with postoperative pain, and tissue PGE²is associated with functional recovery. Preoperative administration of a COX-2 inhibitor reduces both CSF and surgical site PGE²levels as well as CSF IL-6 levels.