Anesthetic Technique for Radical Prostatectomy Surgery Affects Cancer Recurrence: A Retrospective Analysis

After approval by the ethics committee of the Mater Misericordiae University Hospital (Dublin, Ireland), we reviewed the medical records of all patients who underwent radical prostatectomy in Mater Misericordiae University Hospital and the Mater Private Hospital between January 1994 and December 2003. Only the records from patients with invasive prostatic carcinoma where radical prostatectomy was indicated were included. We excluded patients who presented only for transurethral prostatic resection or patients diagnosed with prostate cancer who did not have an open radical prostatectomy.

All patients undergoing radical prostate surgery at the Mater Misericordiae University Hospital are given balanced general anesthesia. Most typically, this includes 1–2 μg/kg fentanyl and 1–2 mg/kg propofol for induction, and neuromuscular antagonism to facilitate tracheal intubation with 0.5 mg/kg atracurium. Anesthesia was maintained with nitrous oxide in oxygen, with a fraction of inspired oxygen of 0.3–0.5 at the anesthesiologist’s discretion. Nonsteroidal antiinflammatory drugs, e.g. , 75–100 mg diclofenac, and 0.1–0.15 mg/kg morphine were given for postoperative analgesia.

Epidural anesthesia and analgesia has been an integral part of the Mater Hospital’s acute pain service since 1990. Typically, it is offered in combination with general anesthesia to patients undergoing radical prostatectomy. When an epidural is used, it is usually inserted preoperatively at a low thoracic level (e.g. , T11–T12), and a preemptive dose of local anesthetic is given before the surgical incision. Local anesthetic administration is then continued as an infusion for 48–72 h after surgery. However, a sizeable minority of patients for this procedure did not receive epidural anesthesia and analgesia for a variety of reasons, including the presence of absolute or relative contraindications, or by preference of the anesthetist, surgeon, or patient. These patients received postoperative patient-controlled morphine analgesia. Patient-controlled analgesia was administered via  a CADD-Legacy ambulatory infusion pump (model number 6300; Deltec Inc., St. Paul, MN) programmed to deliver morphine boluses of 1 mg with a lockout time of 6 min.

Our primary outcome measure was the incidence of “biochemical recurrence,”i.e. , increase in prostate-specific antigen (PSA) after radical prostatectomy compared with its immediate postoperative nadir, which prompted the clinician to instigate adjuvant treatment (radiation therapy, endocrine therapy, or chemotherapy). Postprostatectomy increase in PSA is indicative of metastatic spread or local cancer recurrence. The time epoch evaluated ended October 2006 (i.e. , follow-up interval of 2.8–12.8 yr). Recurrence-free time was defined as the time between the date of surgery and the date of the last PSA, which was at or below the early postoperative nadir. Biochemical recurrence was in turn defined as an increase in PSA above the early postoperative nadir value, which prompted adjunctive therapy (chemotherapy or endocrine therapy or radiation therapy). Recurrence-free time for those patients without a recorded recurrence was defined as the time between the date of surgery and the date of last follow-up.

The following data were obtained from medical records: demographic characteristics; tumor size, grade, and type; PSA status; extent of inguinal nodal disease; whether postoperative or preoperative adjuvant chemotherapy, endocrine therapy, or radiation therapy was used; and each patient’s current status as determined by documentation of follow-up visits to the hospital’s outpatient clinic or general practitioner. In addition, we recorded the histologic tumor margins and whether the tumor was androgen receptive positive. If the most recent follow-up documentation in the hospital records exceeded 3 months from the date of our assessment, we contacted the patient’s general practitioner by telephone to ensure that no consultation had occurred in the interim. This occurred in nine cases: six patients who received general anesthesia alone and three patients who received combined general–epidural anesthesia. In all cases, this clarified whether the patient had contacted his or her general practitioner regarding prostate cancer recurrence.

A total of six patients who received an epidural had documented administration of intramuscular morphine at some point in the first 36 h postoperatively, indicating either epidural failure or inadequate block needing rescue analgesia. These patients were included in the general anesthesia plus epidural group, according to intention-to-treat analysis. Reasons for censoring data included loss to follow-up or inadequate or lost documentation in the medical records.

The anesthetic groups were compared on potential baseline confounders (table 1) using chi-square tests for categorical variables and either t  tests or Wilcoxon rank sum test for continuous variables, as appropriate. Univariable association between recurrence-free survival and anesthetic technique was assessed with Kaplan-Meier survival estimates, and the groups were compared with the log-rank test.29In addition, univariable association between recurrence and all potential baseline confounders was assessed using Cox proportional hazards regression.

For the primary analysis, we compared the anesthetic technique groups on recurrence-free survival using multivariable Cox proportional hazards regression while adjusting for any baseline or intraoperative factors independently related with the outcome. Variables considered were American Society of Anesthesiologists physical status, tumor size, length of surgery, Gleason score, margin (yes/no), preoperative PSA, estimated blood loss, transfusion (yes/no), and date of surgery. We used stepwise regression and a liberal significance criterion of P < 0.30 to be confident of having adjusted for any potential confounding of the relation between anesthetic and recurrence among the available baseline factors, as well as to obtain a more precise estimate of hazard ratio for anesthetic treatment.

We assessed the proportion hazards assumption of the Cox regression model graphically by plotting the log [−log (survival)] against log (time). We assessed the predictive ability of the multivariable model with the c  index.30This summary measure gives the proportion of all usable pairs in which predicted and actual survival times are concordant, such that a patient surviving longer than another is also predicted by the model to survive longer. However, the c  index is a less informative measure of predictive ability when a significant proportion of the data are censored. The linearity of the relation between continuous and ordinal variables and recurrence was assessed graphically.

To assess the robustness of our primary analysis results, we also used propensity score matching to assess the association between type of anesthesia and cancer recurrence. A propensity score, defined as the probability of receiving regional anesthesia as predicted from all available baseline and intraoperative variables (same variables as previously listed in primary analysis), was calculated for each patient using logistic regression. No significance criterion was used to remove variables; all were retained. A greedy matching algorithm31was used to match pairs of regional and general anesthesia patients to within 0.05 on the propensity score scale. Then, to assess association with outcome, the matched groups were compared on cancer recurrence using Cox proportional hazards regression—both univariably, and multivariably using a significance criterion of 0.05 to adjust for any residual confounding.

The significance level for all hypotheses was 0.05. A Bonferroni correction to the significance criterion for multiple comparisons was applied where appropriate. SAS software version 9.1 (SAS Institute, Cary, NC) and R software version 2.4.1 (The R Foundation for Statistical Computing, Vienna, Austria) were used for all analyses.

The groups were not perfectly balanced: The epidural–general patients seemed to have had slightly worse American Society of Anesthesiologists physical status (P = 0.11), more complications (P = 0.05), and slightly shorter surgeries (P = 0.06) than the general–opioid patients (table 1). Complications were postoperative bleeding, pneumonia or other respiratory tract infection, and urinary tract infection. In addition, the epidural–general patients had a smaller fraction of patients with clear surgical margins (P = 0.06). These factors were thus prime candidates for inclusion in our multivariable model.

Table 2contains univariable Cox regression model results for each baseline and intraoperative factor. Without adjusting for potential confounders (i.e. , univariably), patients treated with a combination of epidural and general anesthesia had a significantly lower estimated risk of recurrence compared with the general anesthesia and postoperative opioid group, with hazard ratio of 0.34 (95% CI, 0.19–0.61). Margin, Gleason score, preoperative PSA, tumor size, and duration of operation were also significantly associated with recurrence in univariable analysis. Kaplan-Meier recurrence-free survival estimates and 95% CIs at landmark follow-up times for each anesthesia group are in table 3, with the survival curves plotted in figure 1.

After adjusting for tumor size, Gleason score, preoperative PSA, margin, and date of surgery, the epidural plus general anesthesia group had an estimated 57% (95% CI, 17–78%) lower risk of recurrence compared with the general anesthesia plus opioids group, with a corresponding hazard ratio of 0.43 (95% CI, 0.22–0.83; P = 0.012) in our multivariable Cox regression model (table 4). Gleason score and baseline PSA were strongly associated with recurrence independent of the other model factors. An additional model where all preoperative and intraoperative factors were forced into the regression model gave similar results, with hazard ratio of 0.42 (95% CI, 0.21–0.94; P = 0.014).

There was no evidence of nonproportionality in the hazards for the two groups, and the c  index for the final model was estimated at 0.91, suggesting excellent predictive ability. Although only 64% of the possible pairs were usable because of heavy censoring after 4 yr, the c  index is reliable here because a large fraction of prostate cancer recurrences are expected before 4 yr and little loss to follow-up occurred before then.2As a sensitivity analysis to explore the potential for bias due to censoring, we assessed the association between anesthetic group and recurrence using only the first 3 yr of data for each patient. Using data through 3 yr, before which only 8% of patients were censored, univariable and multivariable associations were similar to the full data results, with P = 0.012 and P = 0.033, respectively.

We also assessed the association between anesthesia technique and cancer recurrence using propensity score matching. Propensity scores were created (see Materials and Methods) for all 200 patients who had nonmissing data for all baseline and intraoperative potential confounding variables (table 5). Before matching, epidural and general anesthesia patients differed significantly on the propensity score and length of surgery, with smaller differences on other variables. Seventy-one matched pairs were obtained (n = 142 patients); matching was successful in improving the balance between groups (table 5). Length of hospital stay is shown in table 5for informational purposes but was not included in the formation of the propensity scores or in the assessment of association with recurrence.

Association between epidural and general anesthesia on cancer recurrence was then assessed on the propensity-matched pairs using Cox regression. Epidural patients were an estimated 52% less likely (0–77%) to recur at any given time compared with general anesthesia patients, with a univariable hazard ratio of 0.48 (95% CI, 0.23–1.00; P = 0.049). Although unnecessary because of excellent matching, a multivariable Cox regression analysis on the propensity-matched patients (table 6) resulted in a hazard ratio of 0.51 (95% CI, 0.25–1.06; P = 0.07). Our propensity-matched analysis results are thus quite similar to our unmatched multivariable model (table 4).

We evaluated cancer recurrence in men undergoing radical prostatectomy. After adjustment for confounding factors, patients who received general anesthesia combined with epidural analgesia had a 57% (95% CI, 17–78%) lower risk of cancer recurrence than patients who had general anesthesia and postoperative opioids. A propensity-matched analysis on a subset of the data gave a similar result: Epidural analgesia had a 52% (95% CI, 0–77%) lower risk of cancer recurrence. These results are strikingly similar to our previous report in women undergoing breast cancer surgery in which disease-free survival was 94% (95% CI, 87–100%) and 82% (74–91%) at 24 months, and 94% (87–100%) and 77% (68–87%) at 36 months in the paravertebral and general anesthesia patients, respectively (P = 0.012).

Both results are consistent with a “decisive period” during and after cancer surgery during which minimal residual disease is either controlled by host defense (presumably largely by NK cells) or is retained in the body—eventually becoming clinically apparent as local recurrence or metastasis. Considerable in vitro  and animal data support this theory and suggest that, unlikely as it might thus initially seem, use of regional analgesia may reduce the risk of recurrence after major cancer surgery.

Natural killer cells, the primary host defense against cancer, are a subpopulation of lymphoid cells that spontaneously recognize and kill a variety of tumor cells in vitro  and in vivo  4and are known to play a determinant role in controlling tumor development—and especially the metastatic process.5We did not evaluate NK cell function in our patients, but much previous work indicates that suppression of NK cell activity occurs within hours of surgery, lasts a few days, and is proportional to the invasiveness of the surgery.17,32Tissue damage, inflammation, pain, anesthetic and analgesic compounds, and psychological stress all contribute to NK cell suppression and the tumor-promoting effects of surgery33–36—and all are moderated by regional analgesia.

Ketamine, thiopental, and halothane have each been shown to suppress NK cell activity and promote metastasis in an animal model.37Other volatile anesthetics also impair NK cell function17by as much as 90%.18Halothane and isoflurane comparably reduce neutrophil motility,38and sevoflurane impairs T lymphocytes.39Although the immune effects of other volatile anesthetics differ somwhat,40,41most—including isoflurane and sevoflurane, which were used in our patients—seem to substantially inhibit various immune functions. We were unable to determine the amount of volatile anesthetic given to our patients. However, it is highly plausible that patients in the epidural group—who were routinely given a preemptive dose of local anesthetic—required considerably less volatile anesthetic than those given general anesthesia alone.

Acute and chronic administration of opioids inhibits components of the cellular and humoral immune function, including antibody production, NK cell activity, cytokine secretion, lymphocyte proliferative responses to mitogens, and phagocytic activity.42,43The immunosuppressive effects of morphine are best studied21,44; however, other opioids, including fentanyl20,45and subtype-specific opioid receptor agonists,46produce comparable immune suppression in most studies.47Inhibition seems to be dose dependent.20Endogenous and exogenous opioids bind three major types of receptors: the μ-, δ-, and κ-opioid receptors that have been identified not only in peripheral sensory neurons and the central nervous system, but also in cells of the immune system such as polymorphonuclear leukocytes, macrophages, T lymphocytes, splenocytes, and macrophage-like and T cell–like cell lines.43Although we were unable to quantify opioid use in our two patient groups, those given epidural analgesia presumably required little opioid, whereas those given general anesthesia alone surely required considerable amounts of opioid for analgesia after open radical prostatectomy, which is a large and painful procedure.

Several important limitations are inherent in this study’s retrospective, observational design. Patients were not randomized and clinical care was not standardized, so that selection bias and the effects of unmeasured confounding variables cannot be excluded. Although our sample size was large enough to detect a strong association, our estimation was rather imprecise, evidenced by wide CIs. And as with most observational studies, even a propensity analysis was not able to completely balance the anesthetic groups on all potential confounders.

There was no significant alteration in clinical practice in the center in question in the relevant time epoch of the study. Laparoscopic or robotic prostate surgery was not introduced, nor were there substantial changes in the frequency of use of epidural analgesia for this type of surgery.

This study, like most retrospective analyses, should be viewed as generating a hypothesis and an estimated effect size for future large randomized controlled trials. (Two are already in progress [ClinicalTrials.gov identifiers NCT00418457 and NCT00531349], and others, including one to evaluate the effect of anesthesia on outcome after radical prostatectomy, will start soon.) Although these studies will require many years of enrollment and follow-up, even a smaller effect size would be clinically important, making the relation between anesthetic technique and cancer recurrence an important hypothesis to pursue, especially given its biologic plausibility.

After primary treatment for clinically localized prostate cancer, including radical prostatectomy, biochemical recurrence (defined as an increase in PSA above its posttreatment nadir) is usually the first evidence of either local recurrence or metastatic progression.48Radical prostatectomy has been shown to provide high 10-yr PSA recurrence-free survival regardless of whether the prostate tumor involves one or both lobes of the gland, once the tumor is histologically confined within the gland.49Others have used either PSA doubling time or any increase in PSA above the posttreatment nadir value as indicative of recurrence.50 

In summary, cancer surgery releases tumor cells into surrounding healthy tissue and into the systemic circulation. We speculate that whether this minimal residual disease becomes established as recurrent cancer or metastases depends on immune competence in the immediate perioperative period. Regional anesthesia and analgesia may help to preserve immune function by attenuating the surgical stress response, decreasing anesthetic requirement, and diminishing the need for opioids. Consistent with this theory, radical prostatectomy with epidural analgesia was associated with a substantial and statistically significant reduction in biochemical evidence of cancer recurrence. Although limited by its retrospective design, our study suggests that prospective trials evaluating the effects of regional analgesia and opioid sparing on cancer recurrence are warranted.