Predictive Risk Factors for Persistent Postherniotomy Pain

• ❖ Significant persistent pain after herniorraphy affects nearly 10% of patients, yet no study has prospectively studied the relative contributions of preoperative, intraoperative, and postoperative factors, which predict the likelihood of this pain

• ❖ Both patient (preoperative pain with functional impairment) and surgical factors predicted the likelihood of persistent pain, providing a rationale for selecting surgical technique based on patient characteristics.

PERSISTENT postoperative pain is recognized as a substantial problem after several different surgical procedures,1,2including groin hernia repair in which approximately 5–10% of patients report persistent pain-related impairment of everyday activities.3,4Persistent postherniotomy pain (PPP) may be an ideal model to understand the pathogenic mechanisms for persistent postoperative pain overall because of the high patient volume, risk for nerve injury, a well-defined surgical procedure, no malignancy, chemotherapy or radiotherapy, and assumed low psychosocial predisposition. Although several factors have been suggested to be related to the development and intensity of PPP, such as psychologic status, preoperative pain, intraoperative nerve handling, surgical technique (open vs . laparoscopic), nerve injury, and acute postoperative pain,3–8the existing data are from studies on single modality mechanisms or interventions. Consequently, no study so far has investigated or clarified the relative role of the relevant pathogenic mechanisms for development of PPP, thereby precluding rational recommendations for prevention and therapy.

Therefore, we designed a prospective study with the primary endpoint of assessing for the first time the relative contribution of preoperative, intraoperative, and postoperative factors to the development of PPP substantially affecting everyday activities after groin hernia repair. Subsequently, we used these data to develop a model for preoperative prediction of patient risk for substantial PPP-related activity impairment.

The study was a detailed prospective consecutive cohort study carried out at two high volume hernia centers to ensure optimal surgical expertise. In each specialized center, an open sutured mesh repair with A.M. Lichtenstein (Hernia Section, Hørsholm Hospital, Hørsholm, Denmark)9or a laparoscopic transperitoneal preperitoneal (TAPP) hernioplasty with glue fixation and a lightweight mesh (Centre for Minimal Invasive Surgery, Bethesda Krankenhaus Stuttgart, Stuttgart, Germany) was performed. The study was approved by the relevant local ethics committees.††Patients were consecutively included after assessment of the inclusion criteria (primary unilateral hernia, age more than 18 yr, men, and ability to understand and write Danish or German) and exclusion criteria (bilateral or recurrent hernia, women, body mass index more than 35 kg/m², current malignant disease, abuse of alcohol, diseases impairing central or peripheral nerve function, other abdominal hernias, previous operations affecting sensitivity in the inguinal region, advanced dementia, or other inability to understand the study). All participating patients gave signed informed consent and received no payment except for travel expenses. Patients were screened at the primary visit to the clinic, and if eligible they were scheduled for clinical assessment 1–2 weeks preoperatively including an interview regarding medical history. A questionnaire regarding the variables of interest was filled on the day before surgery. A telephone interview was performed regarding pain and adverse events on days 7 and 30. Six months after surgery, a detailed clinical examination was performed also to exclude a recurrence.

The primary outcome was 6 months substantial pain-related impairment of everyday activities, assessed by the Activity Assessment Scale (AAS) questionnaire.10The original AAS consists of 13 questions investigating to what degree PPP affects each activity, ranging from “1 = no impairment” to “5 = unable to perform the activity.”10If a specific activity was not performed for reasons other than PPP, a score of “1 = no impairment” was assigned in the analysis. Because of our previous experience with activities affected by PPP, we replaced the activities “walking around inside” or “walking outside or at work” with “standing for more than 30 min,” thereby reducing the number of questions to 12. The summation of each answer to the 12 questions resulted in a raw score ranging from 12 to 60. According to the original AAS study,10the raw scores were transformed to percentages using the formula ([actual raw score − lowest possible raw score]/[possible raw score range]) × 100, resulting in scores ranging from “0%” if no pain-related activity impairment was reported, and “100%” for maximum impairment: a score of 8.3% was, thus, equivalent to being unable to perform an activity (“5 = unable to perform the activity”) but could also be achieved by having less than maximum impairment but in several activities. Thus, all patients with scores equal to or higher than 8.3% were considered as having a substantial PPP-related functional impairment.

Based on the distribution properties of the preoperative accessible variables in the final logistic model together with the type of surgical technique, a model for risk prediction of PPP was developed. Interaction between suggested risk factors was investigated by binary correlation analysis.

At the preoperative interview, information on the following preoperative variables was collected to assess their relative contribution to PPP (table 1).

Demographic characteristics are age, body mass index, and American Society of Anesthesiology score. Maximum daily physical activity level at work or during leisure activities was assigned a score ranging from 1–4 from a predefined list of activities with increasing physical intensity, for example, “1”= no physical activity, “2”= walking, “3”= running, and “4”= weight lifting/soccer.

General pain syndromes (headache, low-back pain, other musculoskeletal pain, abdominal pain other than hernia related, chest pain, or pain not covered by the above) were recorded, and pain occurring weekly or more often, with a Numeric Rank Scale (NRS) (0–10) pain score above 3, was considered as substantial pain. Similarly, persistent pain after previous surgery was recorded, and pain occurring weekly or more often, with an NRS pain score above 3 was considered substantial. Pain from the hernia was recorded, and a combined pain matrix score of intensity and frequency was assigned according to our previous report on preoperative hernia pain.11 

Occurrence of anxiety or depression was investigated by the Hospital and Depression Scale questionnaire,12and pain coping strategies were investigated by the Pain Catastrophizing Scale (PCS).13 

Sensory testing was performed after the interview at preoperative examination and 6 months postoperatively. Before sensory testing, groin and genital hair was trimmed down by the patient, using an electrical trimmer, taking care not to lesion the skin. Tests were performed with the patient in a semireclined position, in a quiet room with dampened light and a fixed room temperature (20°–24°C).14 

To get the patient well acquainted with the testing procedure, all tests began with a demonstration of the testing sequence on the lower forearm. The patient was instructed to keep his eyes closed and focus on the evoked sensations from the stimulation. The patient was unaware of the test results throughout the study. Threshold tests were preoperatively performed 2 cm lateral to the pubic bone and 2 cm above the inguinal ligament (hernia area) and postoperatively at the same area and always 2 cm distal to the hernia scar (open surgery), and anterior on the ipsilateral forearm approximately 10 cm distal to the cubital crease. Tests alternated between the groin and arm, always beginning in the groin. Site visits and follow-up visits between investigators were made to ascertain identical test procedures in the two centers, before and during the study.

A thermal stimulus (Modular Sensory Analyzer, Somedic AB, Hörby, Sweden) was used to asses warmth detection thresholds (WDT) and heat pain detection thresholds (HPT). A Peltier thermode with an area of 12.5 cm²was placed on the skin and five thermal stimuli for each parameter were applied with a randomized stimulus interval of 4–6 s, starting from baseline temperature of 32°C with a ramp rate of ±1°C/s. The patient pressed a button when experiencing warmth, thereby assessing WDT. Heat pain threshold was reached when warmth became painful. Failure to respond before the cutoff limit was reached resulted in assignment of the cutoff value. Cutoff limit was 52°, and WDT or heat detection threshold was calculated as the average of the five tests respectively.

A heat pain stimuli was chosen to assess pain responsiveness, as it is easily applicable, and heat pain is well recognized as opposed to cold or electrical pain. Four stimuli for each temperature of 45°, 46°, 47°, and 48°C (16 heat stimulations in total) were administered in a semirandomized order keeping the patient unaware of the ensuing temperature in relation to the previous, throughout the study. The ramp rate was 5°C/s with a return rate of 1°C/s. The temperature was maintained for 5 s with a 30-s interstimulus period. At the end of each stimulation, the patient reported NRS pain, and the average pain assessed for each temperature was calculated. The NRS pain responses to all four temperatures were closely correlated (ρ > 0.8, P < 0.001). The response 47°C stimuli was chosen for the predictive logistic risk factor analysis as it had the best normal distribution properties (skewness −0.287, kurtosis −0.631, Kolmogorov-Smirnov normality test P = 0.07).

The differences between the preoperative and postoperative WDT and heat detection threshold measurements were calculated to investigate the degree of intraoperative nerve injury.

Open surgery was performed as a Lichtenstein sutured mesh repair (heavyweight mesh,108.5 g/m²monocryl-prolene-composite; Johnson and Johnson, Birkerød, Denmark, and monofilament polyprolene, surgipro suture; Tyco Healthcare, Copenhagen, Denmark). Laparoscopic surgery was performed as a (TAPP) hernioplasty as described previously15with mesh glue fixation (1 ml fibrin glue/,Tissucol; Baxter, Unterschleissheim, Germany) and implantation of a lightweight mesh (36 g/m²polypropylene; Optilene LP, B. Braun, Melsungen, Germany).

All procedures were performed under general anesthesia, (propofol and remifentanil, Copenhagen, and sevoflurane/desflurane in combination with a 70:30% mixture of nitrous oxide and oxygen, Stuttgart). Postoperative analgesia was acetaminophen 1 g 6 hourly and ibuprofen 400 mg 6 hourly for the first week. Patients were discharged the same day in Copenhagen and the next morning in Stuttgart. Intraoperatively, the hernia size was recorded, and the funicle, the ilioinguinal, iliohypogastric, and genitofemoral nerve, and any observed injury to these structures were identified. However, because of the surgical technique, nerve identification was not performed in TAPP-treated patients.

On days 7 and 30, a phone interview was conducted concerning wound healing, pain, and readmissions. At the 6-month follow-up, sensory thresholds were reassessed together with a questionnaire regarding AAS score, pain intensity, frequency, and a physical examination to asses recurrence or other complications.

The primary endpoint was identification of factors significantly related to substantial PPP-related activity impairment (AAS score ≥ 8.3) after 6 months. Because of a left-shifted distribution of postoperative AAS scores not normalized by log transformation, a logistic analysis was performed (see Results for details of 6 months AAS score distribution). All predefined factors were included in an overall logistic model. Categorical variables (AAS, the Hospital and Depression Scale, PCS, pain matrix score, and others) were subdivided into categories; for instance, preoperative AAS was divided into four categories (0 = no impairment, > 0–8.3 < = minimal impairment, ≥ 8.3–30.3 < = moderate impairment, and > 30.3 = severe impairment.) This was done to have fewer categories to improve sensitivity.

After construction of the primary logistic model, a backward elimination procedure was performed to reduce the model until all the remaining factors were significantly correlated to the primary outcome of an AAS score more than or equal to 8.3. Subsequently, we checked that no substantial interactions had been ignored by adding interaction terms to the final model and one by one including interactions with variables that had been eliminated. No significant interactions were found. Finally, we checked that the linear model was adequate by introducing polynomial effects of the continuous predictors in the model. None of the nonlinear terms were significant. To asses the overall predictive potential of the suggested prediction models, we computed two performance measures, namely the Brier score and the C statistic (also known as the area under the receiver operating characteristic curve). The Brier score is the mean squared difference between the risk predicted by the model and the binary outcome. The lower the Brier score, the better the individual risk predictions. A Brier score of 0 corresponds to making perfect predictions, and 33% corresponds to predicting the outcome by coin flipping. Although the Brier score measures the accuracy of the individual risk predictions, the C statistic measures the overall discriminatory power of the model. A model with no discriminatory ability has a C statistic of 50%, whereas a model with perfect discriminatory ability has a C statistic of 100%. It is well known that performance measures tend to be overly optimistic when evaluating the predictive ability on the same data that were used to fit the model. In particular, when many potential predictors are modeled, there is a risk that the data will be overfitted. The validity of the regression model can be assessed by a procedure known as bootstrapping in which a larger number of resamples are drawn from the original data with replacement, and the model is refitted on each bootstrap sample. Thus, we used regular bootstrapping to remove bias from our estimates as recommended by Ref. 16. A total of 1,000 bootstrap samples were drawn.

An analysis including only preoperative factors and surgical treatment was performed to allow development of a model for risk prediction of PPP. The bivariate correlations between factors are analyzed by estimating Spearman's Rho. Significance was set at a 5% level for all estimates. Continuous variables were compared by Student t  test and reported as medians with 95% CI. Categorical variables were compared by Mann–Whitney U test for group comparison or Wilcoxon test for comparison of changes in paired values before and after surgery. All categories were ordered. Data are reported as medians with range and percentiles where appropriate.

A prospective power analysis for this hypothesis-generating study was not undertaken. The large consequent patient material was based on the research time available within a 12-month period. Thus, the statistical uncertainty of each parameter is reflected in the corresponding 95% confidence intervals. To assess improvement/aggravation from surgery, we calculated changes in AAS score as ΔAAS = AAS^post− AAS^pre.

Logistic regression was performed using the SAS version 9.1 software (SAS Institute Inc., Cary, NC). All other analyses were analyzed using SPSS version 13 software (SPSS Inc., Chicago, IL).

Four hundred sixty-four patients were included, and 442 (95.3%) were followed up per protocol. Details of exclusion and loss to follow-up are presented in figure 1. Two recurrences after Lichtenstein's procedure were seen. Demographic data and distribution of variables are shown in table 1for all patients and stratified for the two centers. Several factors were significantly different between the two centers, with patients scheduled for Lichtenstein's herniotomy having a significantly higher activity level and higher preoperative pain-related activity impairment (AAS score). TAPP patients had a significantly higher American Society of Anesthesiology score, higher pain matrix scores, larger hernia defects, and more affected anxiety, depression, and pain-coping strategies scores. The two centers were comparable with regard to age, body mass index, pain from other bodily regions or after previous other surgery, and hernia types. Assessment of preoperative nociceptive function was comparable between the two centers, that is, no systematic variation because of differences in testing method was seen, except for WDT in the hernia region being higher in the Lichtenstein group. However, as seen in table 2, WDT differences were clinically irrelevant, that is, less than 0.5°C. Preoperative AAS score was highly correlated to preoperative pain matrix score of pain intensity and frequency (ρ= 0.70, P < 0.001).

Three hundred thirty-nine patients (76.7%) did not report any PPP-related activity impairment at 6 months. Pain of any degree (NRS 1–10) was reported by 117 (26.5%), and 55 (12.4%) had substantial pain-related impairment at the 6-month follow-up (AAS score ≥ 8.3). Of these, 10 (17.5%) had “a little” difficulty in five or more activities, 33 (57.9%) had “some difficulty” in three or more activities, 10 (17.5%) had “a lot” of difficulty in two or more activities, and 4 (7.1%) were “unable” to perform the activity due to pain (table 3). Overall, the operation reduced pain-related impairment of any degree from 66.3% before operation to 23.3% after (P < 0.001), with a significantly reduced maximum impairment (P < 0.001) (table 3). Thirty-five patients (7.9%) reported a higher degree of pain-related impairment after 6 months than before surgery with median increase in AAS score of 5 (range 2.1–27.1).

Preoperatively, 52 (11.7%) patients reported pain from the hernia during sexual activity, with 6 (1.3%) finding pain to impair sexual function moderately (1.3%) and severely in 6 (1.3%). Postoperatively, 37 (8.4%) patients reported pain from the groin during sexual activity, whereof 25 (5.6%) were new cases. Three (0.7%) patients found the pain to impair sexual function moderately and severely in six (0.5%). Ejaculatory pain was experienced by six (1.3%) preoperatively and by six (1.3%) postoperatively; however, in five of these latter cases, the ejaculatory pain was new (four open vs . one laparoscopic).

Factors suggested to be related to the binary outcome of substantial PPP-related impairment of everyday activities was investigated in the logistic analysis (table 4). The whole model was significant (P < 0.001). By backward elimination, a reduced model was fitted with only independently significant variables for PPP: division of preoperative AAS score into four categories (P = 0.0036), pain response to 47°C heat stimulation (P = 0.0018), pain intensity on the postoperative day 30 (P = 0.0002), and changes in WDT in the groin (P = 0.029). The corresponding odds ratios (ORs) for substantial PPP-related activity impairment for each variable are shown in table 5, where the preoperative OR for the three groups with pain-related activity impairment are compared with the group without any pain-related impairment (intercept), and the OR for pain scores during the 47°C heat stimulation and the day 30 postoperative pain scores are per NRS point (11 point scale), and OR for sensory deficit are per degree of sensory loss (maximum loss 16°C). The bias-corrected Brier score for this model is 9.8%, and the C statistic is 77.0%, indicating a fair predictive and discriminatory ability.

Age (mean 57 yr, range 18–86, 25th and 75th percentile = 46 and 64, respectively) was significantly correlated to the postoperative AAS score (ρ=−0.13, P = 0.009) with lower pain-related impairment with increasing age when analyzed as a simple bivariate analysis but not when subjected to logistic regression analysis. Age was significantly correlated to preoperative WDT and HPT in the groin (ρ= 0.26, P < 0.001 and ρ= 0.16, P < 0.01, respectively) with increasing detection threshold with higher age. Preoperative pain response to the 47°C heat stimuli was significantly lower with increasing age (ρ=−0.15, P = 0.002).

Daily work or leisure activity level was reported as “heavy” by 100 (22.6%), “moderate” by 162 (36.7%), “light” by 155 (35.1%), and not being physically active by 25 (5.7%). Activity level was significantly correlated to age with less activity with increasing age (ρ=−0.17, P < 0.001). However, there was no significant correlation between activity intensity and substantial PPP-related activity impairment (ρ= 0.09, P = 0.06). American Society of Anesthesiology score was not correlated to the level of postoperative AAS score (ρ= 0.05, P = 0.3).

The prevalence of other frequent pain syndromes is shown in table 1. Twenty-one patients also reported pain after previous surgery, mainly joint surgery (knee, shoulder, spine; n = 16), fractures (n = 4), coronary bypass (n = 3), and sterilization (n = 2). There were no significant correlation between other pain syndromes and the postoperative AAS score (ρ < 0.05, P = 0.2). Patients with other pain syndromes did not have a higher pain response to the 47°C heat stimuli, than patients without (median NRS = 5 vs . 5, P = 0.16).

NRS pain score on day 7 was low (median = 1, range 0–9), with 24 patients scoring 4–6 and 8 patients scoring 7–10. At day 30, median NRS pain was 0 (range 0–10), with 48 patients scoring 4–6 and 16 patients scoring 7–10 on the NRS pain scale. Patients' pain scores on days 7 and 30 were closely correlated showing that substantial pain was reported by the same patient population but with higher pain scores. Acute postoperative pain (days 7 and 30) was significantly correlated to the preoperative 47°C pain response (ρ= 0.26, P < 0.001 and ρ= 0.13, P < 0.01, respectively). Nerve damage, assessed as changes in WDT in the groin, was significantly correlated to acute pain on days 7 and 30 (ρ > 0.2, P < 0.001 for both). Anxiety was the only psychometric parameter to be significantly correlated to the intensity of day 7 or 30 pain (ρ > 0.16, P < 0.001 for both days). Depression and PCS scores were not correlated to the acute pain response (ρ < 0.09, P > 0.06).

The incidence of substantial PPP-related activity impairment (AAS score ≥ 8.3) was significantly lower in TAPP- versus  Lichtenstein-operated patients (8.1 vs . 16.0%, respectively, P < 0.02). Although the overall reduction in pain-related impairment was not significantly different in Lichtenstein- versus  TAPP-operated patients (P = 0.1), patients operated with A.M. Lichtenstein had a significant higher maximum intensity of pain-related impairment at 6 months (P = 0.001) (table 3). The number of patients with increased pain-related impairment or level of increase in AAS score was not significantly different between Lichtenstein- or TAPP-operated patients (P > 0.1).

According to the Hospital and Depression Scale, there was a low incidence of anxiety and depression. Thus, preoperatively 399 (95.0%) patients did not have depression, 13 (2.9%) had possible depression, and 9 (2.1%) had definite depression. Anxiety was not seen in 399 (90.2%) patients, possible in 23 (5.2%), and definite in 20 (4.5%). Both anxiety and depression was significantly binary correlated to the primary outcome of substantial PPP-related activity impairment (ρ= 0.1, P = 0.03 for depression, and ρ= 0.2, P < 0.001 for anxiety), but not in the logistic regression analysis. However, anxiety and depression were both correlated to preoperative AAS score (ρ > 0.12, P < 0.008 for both). The total PCS and subscores (rumination, magnification, or helplessness) were all significantly correlated to the preoperative AAS score (ρ > 0.1, P < 0.03); however, neither the total PCS score nor any of the subscores were correlated to the postoperative AAS score or the primary outcome (AAS ≥ 8.3) (ρ < 0.07, P > 0.2 for all).

Assessed by the lack of changes in sensory function on the lower forearm, the testing sequence was robust, that is, no systematic variation was seen between the preoperative and postoperative testing, implying that observed sensory changes in the groin area was due to surgical intervention (table 2). Thus, warmth and heat detection thresholds were identical in the groin and on the arm preoperatively as seen in table 2. However, at the postoperative examination, sensory function (WDT and HPT) in the groin was significantly impaired compared with the preoperative function (P < 0.001). The sensory changes primarily occurred in patients undergoing open hernia repair, and in laparoscopically operated patients, the only significant change was in HPT, which furthermore was significantly smaller than the HPT changes seen in open-operated patients (0.5 vs . 2.4 Δ°C, P < 0.001) (table 2), and clinically irrelevant. The changes in WDT were significantly higher in the 35 patients with increased postoperative AAS scores compared with preoperative scores, than in the 407 with less or equal impairment, supporting the relationship between nerve injury and PPP (table 6). The pain response to 47°C heat stimulation applied to the arm was closely correlated to the response from the groin (ρ= 0.77, P < 0.001). Preoperative pain was not associated with alterations in preoperative nociceptive function, evaluated by the correlation between preoperative pain matrix score11and response to the 47°C tonic heat stimulation on the arm or the groin (ρ < 0.08, P > 0.1 for both) or preoperative pain and WDT and HPT on the arm or groin (ρ < 0.08, P > 0.1 for all). Similarly, there was no significant correlation between other bodily pain than PPP and the pain response to 47°C (ρ=−0.03, P = 0.6).

Because several of the investigated risk factors mentioned earlier will not be available to the clinician before surgery (days 7 and 30 pain, changes in sensory function), we also fitted a model excluding these variables, resulting in the predictive model shown in table 7. The surgical intervention was now significantly independently correlated to the risk of persistent pain by a factor 0.45 in favor of laparoscopic surgery to reduce the risk of PPP. Thus, 16% of open versus  8.1% of laparoscopically operated patients had an impairment more than or equal to 8.3 AAS points (chi-square test = 6.3, P < 0.02). The information on distribution parameters ( appendix) of the significant preoperative factors together with stratified data on the type of surgical technique was used to construct risk plots (figs. 2 and 3). The specific preoperative risk for a patient to develop PPP-related functional impairment can be found by taking the NRS pain response to preoperative 47°C heat stimulation (x-axis) and the relevant AAS category with stratified risk for open and laparoscopic repair in figures 2 and 3, respectively; hereafter, the risk can be found on the y-axis. The analysis shows that low preoperative (∼< 2 NRS points) pain response to 47°C heat stimulation is related to a low (2–10%) risk of PPP—regardless of the surgical intervention. However, for patients with high preoperative pain responses to 47°C heat stimulation, the risk of PPP is higher after an open versus  a laparoscopic hernioplasty (figs. 2 and 3). Thus, in patients with severe preoperative pain-related activity impairment and maximum 10 NRS point pain during 47°C heat stimulation, the average risk for PPP was approximately 30% versus  approximately 70% (95% CI 0.1–0.6 vs . 0.35–1.0, P < 0.05) after laparoscopic versus  open repair, respectively. The bias-corrected Brier score for this model is 10.0%, and the C-statistic is 74.3%, indicating a fair predictive and discriminatory ability.

This is the first detailed large-scale prospective study to investigate the relative role of several factors assumed to be related to persistent postsurgical pain in general and groin herniotomy in specific. Thus, by applying a logistic regression analysis, we found four factors to be independently correlated to PPP-related impairment: preoperative pain-related functional impairment intensity (AAS score), preoperative pain response to heat, intraoperative nerve injury (6-month sensory disturbances), and early (day 30) postoperative pain intensity. The current study has several strengths, including a large patient population and a uniform detailed study protocol in the two high volume hernia centers. The number of patients with PPP and the pain intensity distribution seen in our study was similar to what has been reported previously.4,7,17The primary outcome of substantial PPP-related activity impairment, assessed by the validated AAS score,10may be a relevant clinical outcome closely correlated to the pain intensity and frequency matrix, supporting the validity of the AAS and allowing comparison with other trials. Our choice of introducing a new item into the original AAS questionnaire without revalidating the entire questionnaire could be of potential concern, but because of the simplicity of the question, we do not believe that overall reproducibility has been seriously altered. Furthermore, our choice of multiple logistic regression for the statistical analysis allows us to assess the effect of the predictors while adjusting for other confounding factors, in contrast to simple correlation analysis. As shown in the results section, several potential predictors of PPP were correlated, and the multiple regression was, thus, applied to identify the strongest predictors. Thus, our results show that several of the previously suggested factors for persistent postoperative pain, such as age, psychometrics, and other bodily pain,1do not have a significant role in the prediction of PPP in our study.

Despite these obvious strengths, the study has weaknesses. Patients were not randomized or stratified for Lichtenstein or TAPP hernioplasty because we wanted data from each type of surgical intervention from established high volume and expertise centers. This may explain why a few patient characteristics were different between centers, as shown in table 1. However, most of the factors were only significantly different because of the large study size with small numerical differences. For instance, the preoperative difference in WDT was less than 0.5°C, assumed to be without clinical relevance. Overall, the sensory function tests were homogeneous between centers, and the testing was robust as seen by the almost identical results from the groin area preoperatively, and the control arm area preoperatively and postoperatively. Furthermore, several factors expected to adversely affect the pain-related outcome was worse in the group with the best outcome (TAPP), for instance psychologic factors. Finally, the multiple logistic regression model adjusted for these variations. Despite the large patient material, our study may have overfitted the data by including too many variables when considering the low incidence of the primary outcome. To assess the predictive ability of our reduced models, we computed bias-corrected Brier scores and C-statistics of 9.8 and 77.0% and 10.0 and 74.3%, respectively, for the two reduced models, showing a fair predictive ability. The uncorrected numbers were 9.3 and 80.0%, 9.7 and 76.5%, indicating that the final models perform well in replications.

Similar to previous reports on the correlation between preoperative pain and PPP,8,18–21we found preoperative pain at the surgical site to be an independent significant risk factor for development of PPP. Furthermore, we were able to quantify the relative role of preoperative pain-related impairment, showing that patients with severe impairment had a 5-fold higher risk of persistent impairment compared with patients without preoperative impairment. However, there was no correlation between sensory thresholds or pain response to heat and the level of preoperative pain-related impairment, supporting the previous finding11that local preoperative neuroplastic changes may not be important for the development of PPP.

The preoperative pain response to a standardized noxious input, in this case a series of 5 s at 47°C, was a significant independent predictor of PPP, with an approximately 30% increased risk per NRS point (table 7; figs. 2 and 3). This is in accordance with previous studies on the predictive value of preoperative pain responses to various noxious stimuli and acute postoperative pain.22–25Although we found only the pain response to the heat exposure in the groin to be significantly correlated to PPP in the logistic regression analysis, the preoperative pain response to heat stimulation reflected a general pain hypersensitivity not restricted to the groin because the arm stimulation results was highly correlated to the groin results (ρ= 0.77, P < 0.001). Despite this close correlation, the most likely cause why the arm heat pain response did not emerge as a predictive factor in the logistic analysis may be that they reflect the same sensory modality and the quantitatively larger effect from the groin test, thus overpowering the arm results. Recently, the preoperative function of the endogenous analgesia system assessed by the diffuse noxious inhibitory control system was shown to predict persistent postthoracotomy pain26but not the suprathreshold pain responses to heat as we have found. The diffuse noxious inhibitory control was assessed by rating pain during two identical noxious stimuli: one delivered alone, the next concomitantly with another “conditioning” remote noxious stimulus. These differences may be explained by our study having a more uniform and a far larger sample size (442 vs . 62) and with more detailed variables, but it cannot be excluded that diffuse noxious inhibitory control may be an important predictor of persistent postoperative pain. However, simple noxious heat stimulation is easier to perform than diffuse noxious inhibitory control testing and may, therefore, have more practical clinical impact.

Similar to previous studies in hernia patients14,27–30and other surgical procedures,31–33we found that nerve injury (6 months sensory disturbances) was common in pain and pain-free patients, and our study furthermore shows that the risk of pain increases with increased sensory loss independent of the surgical procedure (table 6). Beldi et al .30in a smaller study (n = 96) found a similar effect with increased PPP with increased hypoesthesia, but only in laparoscopically operated patients and not open herniotomy. However, their study suffers from less-detailed sensory testing and major differences in follow-up period between the surgical groups, thereby limiting interpretation. In contrast to previous studies, we decided not to include a full quantitative sensory testing protocol,14,34and large nerve fiber injury/regeneration may, therefore, not have been detected. However, we have recently shown that loss of one sensory modality is strongly correlated to loss of other sensory modalities.35Furthermore, a full quantitative sensory testing protocol would not have been feasible in the current large study.

In accordance with previous but less detailed studies,6,36,37we found a significantly reduced incidence of PPP after laparoscopic versus  open herniotomy. For the first time with detailed quantitative sensory testing, the reduced risk of PPP was found to be correlated to a significantly lower incidence of nerve injury (sensory disturbances) from the laparoscopic TAPP technique with glue fixation and implantation of a lightweight mesh. In fact, the correlation between the type of surgical technique and the nerve injury was so strong that nerve injury as a single factor in the analysis overshadowed the effect from the surgical technique per se , because of the fact that there are more levels in the nerve injury assessments and thereby higher sensitivity compared with the dichotomous surgical technique factor in the analysis. Nevertheless, the effect was clearly seen when only preoperative available factors, including choice of surgical technique, was analyzed showing that laparoscopy significantly reduced the risk for substantial PPP with an OR of 0.45 and particularly in patients with the highest risk for substantial PPP based on preoperative pain-related activity impairment and pain response to heat stimulation (figs. 2 and 3). Besides direct intraoperative nerve trauma because of the operative technique, nerves may be affected by the inflammatory reaction produced by the mesh. In TAPP, a lightweight mesh with large pores (36 g/m²polypropylene; pore size 1 mm) was implanted, which entails less foreign body reaction and consecutively less pain than conventional heavyweight meshes (> 80 g/m²).38Furthermore, in open surgery, a direct contact between the mesh and nerves may occur,39which may be avoided in TAPP due to a thin fascial layer protecting the nerves located in the parietal compartment of the preperitoneal space, whereas the mesh is placed in the visceral compartment without any direct touch with the nerves.

Of specific interest is dysejaculation seen in approximately 1.5% similar to earlier studies.40,41Although dysejaculation was seen in equal number of patients preoperatively and postoperatively, it was a new phenomenon in five patients and more frequently after open surgery. However, the scale of the current study does not allow for identification of specific dysejaculation risk factors.

Acute postoperative pain (day 30) was a significant independent predictor for PPP, similiar to the findings in several other studies both on PPP42and other surgical procedures.1,2Preoperative pain response to 47°C was also significantly correlated to acute pain on days 7 and 30, suggesting that increased pain susceptibility may result in high and persistent acute pain, as described previously,22and subsequent persistent pain. Furthermore, patients with high pain on days 7 and 30 were found to have increased WDT (nerve damage) at 6 months, suggesting a role for neuropathy in the early postoperative phase. This may be also be the reason why pain intensity and frequency was higher on day 30 than on day 7, because increased pain may be the result of developing neuropathy during the regenerative process.

An important finding in the current study is that several of the previously suspected factors were not found to be independently correlated to PPP. Thus, despite a prevalence of other bodily pain similar to previous reports,43we did not find a significant effect of pain from other body regions on the risk of PPP, which has otherwise been suggested as a marker for pain susceptibility in previous studies from various procedures.18,32,41,44Furthermore, no correlation between other bodily pain and pain response to 47°C pain was seen, suggesting that other bodily pain may not alone be the result of increased nociceptive function but encompasses a wide range of syndromes (headache, low-back pain, arthrosis, and others) with various etiology (genetics and trauma), which thereby explains the demonstrated low specificity for persistent pain at least in herniotomy. In contrast to other studies,1,2,27we did not find age to be an independent factor for PPP, although age was correlated to decreased sensory function assessed both as reduced preoperative warmth and heat detection and lower pain response to a heat stimulus, similar to previous data.45Thus, previous findings of age and PPP relations may have been proxy correlates of the herein found correlations between detailed assessments of nociceptive function and PPP. Psychologic factors such as anxiety46,47and depression48or catastrophizing49have been shown to be associated with acute postoperative pain. However, data on psychologic factors and persistent postoperative pain are sparse50and inconclusive despite a positive correlation between preoperative scores in catastrophizing and limb amputation and pain 2 yr postoperatively.51Other studies have suggested that psychologic factors do not predict postoperative pain but rather persist or aggravate in case of a negative outcome.52,53Our findings supports these latter studies in the sense that we did not find anxiety, depression, or pain-coping strategies to be significant predictive factors for PPP, despite anxiety being correlated to acute pain. However, in herniotomy patients, the incidence of psychometric disturbances was low (5.2%), resulting in a wide confidence interval for the OR of the psychometric variables (for instance OR 95% CI 0.43–4.03 for depression, table 4) hindering a more detailed analysis in relation to PPP. For future research, it could be noted that for finding a significant difference between a high risk group of say 20% risk and a low risk group of 5% risk with 80% certainty without correcting for other confounders, a sample of only 146 persons in total would be needed if the two groups were even sized, whereas a total of 762 would be needed if the low risk group outnumbered the high risk group by 19 to 1.

In conclusion, this is the first large scaled multifactorial study that shows that PPP is the result of both patient and surgical factors. Independent factors for PPP-related activity impairment are preoperative AAS score, increased pain to preoperative heat stimulation, nerve injury, and early postoperative pain. Preoperative data on AAS score and response to heat stimulation can help clinicians in guiding high-risk patients to laparoscopic (TAPP) surgery with reduced risk for PPP-related activity impairment. Finally, the results serve as a basis for improved design of future studies on the pathogenesis and prevention of the transition from acute to persistent postsurgical pain.

The authors thank Jørgen Malmstrøm, M.D., and Torsten Asmussen, M.D. (Physicians, Hernia Section, Hørsholm Hospital, Hørsholm, Denmark), for their surgical assistance.

The estimated risks in figure 2were obtained from the logistic regression model with the formula:

where I(—) denotes an indicator function that is 1 if the indicated condition is present and 0 otherwise, and expit(x) = exp(x)/(1 + exp(x)). In the formula, –5.21 is the intercept term, and the values multiplied with the predictors are the regression coefficients.