Genetic and Clinical Factors Associated with Chronic Postsurgical Pain after Hernia Repair, Hysterectomy, and Thoracotomy: A Two-year Multicenter Cohort Study

SINCE the discussion of chronic postsurgical pain (CPSP) commenced in 1998,¹  this late complication has proven to be a frequent cause of persistent pain in the general population²  and has been linked to a wide range of surgical settings.^3,4  Risk factors, pathogenesis, and preventive strategies continue to be widely debated.^3–6  The main predictors described to date are female sex,⁷  age,⁸  psychosocial factors,^9,10  a history of pain in the region of surgery or other sites,^5,8,11  type of procedure,^3,4,12  nerve injury,¹³  and postoperative pain intensity.¹⁴  Additionally, genetic polymorphisms have been linked to varying sensitivity to pain,^15,16  susceptibility to certain painful conditions,¹⁷  and response to analgesics,^18–20  leading some to suggest that such factors might explain why some patients develop chronic pain and others do not.^3,4,21  Studies with sufficient power to confirm the relevance of single-nucleotide polymorphisms (SNPs) have yet to be published, although they are potentially of considerable interest.

We hypothesized that within patient populations sharing the same surgical contexts and clinical–demographic risk for CPSP, genetic factors would identify individuals at risk for this complication. Our aims were to analyze functional genetic polymorphisms related to CPSP risk or protection and clinical predictors at 4 months after three types of surgery—inguinal hernia repair, hysterectomy, and thoracotomy. We also sought to determine pain interference with daily living at 4 months; the incidence of CPSP and pain intensity at 4, 12, and 24 months; and the rate of neuropathic pain in patients with CPSP at 4 months.

This prospective multicenter cohort study enrolled patients scheduled for inguinal hernia repair, hysterectomy (vaginal or abdominal incision), or thoracotomy. Genetic associations in the subgroups of patients with and without CPSP (cases and controls) were compared.

Twenty-three Spanish hospitals (appendix) recruited patients from January 8, 2009, to December 31, 2010. Follow-up ended on December 31, 2012.

Candidates for inclusion were scheduled for inpatient or outpatient inguinal hernia repair (men), vaginal or abdominal hysterectomy (for nononcologic reasons or for cervical carcinoma in situ, but excluding other oncologic procedures), or thoracotomy (men) under general, regional, or local anesthesia with sedation (see table, Supplemental Digital Content 1, https://links.lww.com/ALN/B135, for patient distribution by diagnostic and surgical codes).

Candidates were excluded if they or their parents or grandparents had been born in the Canary Islands or outside Spain or if they were of Roma ethnicity. Candidates were also excluded if they were under 18 yr of age, needed reoperation, had a serious psychological disorder, were undergoing endoscopic or other procedures not requiring incision, or were relatives (parents, grandparents, children, grandchildren, or siblings) of patients already enrolled.

Cases were all patients with CPSP at 4 months; for the gene study, a control group was formed by selecting a block-randomized sample of CPSP-free patients from each surgical group.

The primary outcome was the incidence of CPSP confirmed by physical examination approximately 4 months after surgery based on the criteria of Macrae and Davies²²  published by the International Association for the Study of Pain. These criteria are as follows: (1) the pain should have developed after a surgical procedure; (2) the pain should be of at least two months’ duration; (3) other causes for the pain, such as continuing malignancy or chronic infection, should be excluded; and (4) the possibility that the pain is continuing from a preexisting problem should be explored and exclusion attempted. Although these criteria specified waiting at least 2 months before diagnosing CPSP, others later proposed waiting at least 3 months^23,24  because of the possibility of persisting inflammatory changes and neuropathic pain.²⁵  We therefore chose to modify the criteria slightly, cautiously waiting approximately 4 months before diagnosing CPSP.

The secondary outcomes were (1) the incidence of CPSP reported in telephone interviews at 12 and 24 months and (2) the percentage of patients with CPSP at 4 months whose pain had neuropathic characteristics.

Designated anesthesiologists in each hospital’s local research team attended training sessions on how to complete the clinical questionnaire and diagnose CPSP. Questionnaire variables and definitions are shown in Supplemental Digital Content 2, https://links.lww.com/ALN/B136. The following variables were collected before surgery and during hospitalization. Before surgery, the anesthesiologist administered the validated Spanish version²⁶  of the Hospital Anxiety and Depression Scale, which has proven useful for diagnosing anxiety or depression in patients without a prior history of psychiatric problems,²⁷  and version 2 of the Short Form Health Survey-12 (SF-12) questionnaire²⁸  to assess two components (physical and mental) of quality of life. Also recorded at this time were physical status according to the American Society of Anesthesiologists’ classification; the presence of prior pain in the area of surgery and in other parts of the body expressed on a verbal numerical rating scale (VNRS) of 0 to 10 (0 = no pain; 10 = the worst imaginable pain) and history of treatment with analgesics; concomitant diseases; and any history of substance addiction to street drugs, alcohol, or smoking. Surgical variables were procedure, duration, techniques of regional and local anesthesia, doses of opioids and antihyperalgesic agents, and intraoperative complications. For 24 h after surgery, analgesia and postoperative pain (VNRS) were recorded.

Data were collected with a structured telephone questionnaire between 1 and 1.5 months after surgery (see Supplemental Digital Content 2, https://links.lww.com/ALN/B136); all the interviews were done by the same investigator (J. Cantillo). Patients who reported pain at that time were telephoned again between 2.5 and 3.5 months after surgery and, if pain was still present, were given an appointment for clinical examination between 3.5 and 4.5 months after surgery; this visit, during which CPSP was diagnosed, took place at the hospital. The examiner at this time was an anesthesiologist expert in pain management who used the following instruments: Brief Pain Inventory (severity, analgesics, and interference with daily living), the Spanish SF-12 questionnaire, and the Douleur Neuropathique 4 questionnaire.²⁹  This third instrument assesses whether CPSP could be described as neuropathic, indicated by a positive response to 4 out of 10 items. The physical examination included determining the exact location of pain (noted in the Brief Pain Inventory) followed by testing for hypoesthesia (slight touch with a cotton swab, pinpricking with Von Frey filaments) as well as for dynamic allodynia (brushing) according to items specified in the Douleur Neuropathique 4 questionnaire. These tests were applied on both sides of the body. The patient also reported use of analgesics. Patients whose diagnosis of CPSP was confirmed at this time were interviewed by telephone again at 12 months and, if pain persisted, again at 24 months. If a patient was lost to follow-up, the National Health Service Death Register was checked.

The study was approved by the clinical research ethics committees of the leading center, Parc de Salut Mar (file reference CEIC-IMAS: 2008/3080/I) and all other centers (appendix). Patients signed informed consent statements for data collection, DNA analysis, follow-up telephone contact, and a hospital appointment for physical examination. Otherwise, patients received routine care.

We targeted examining the presence of strong associations with CPSP, some of which had previously been reported in the literature. Using standard procedures,³⁰  we estimated that a minimum sample of approximately 500 cases and 500 controls was needed to have greater than 90% power to detect a risk allele with an odds ratio (OR) of 1.5 for CPSP in a simple allelic test, assuming an incidence of at least 10% for this late complication and risk allele frequencies of 0.1 or larger.

Based on findings that CPSP develops after 10% of inguinal hernia repair procedures, 10% to 30% of hysterectomies, and 30% to 40% of thoracotomies³  and considering the numbers of these procedures recorded at the 23 participating hospitals in a previous epidemiological study in our area,³¹  we planned to recruit a sample of 600 patients with CPSP in 2 yr. A 20% loss to follow-up was expected.

DNA extraction was only performed in volunteering patients with confirmed CPSP and in selected control patients without CPSP who were matched to cases by age, surgical specialty, sex, domicile, and hospital recruitment. For each patient, peripheral blood (5 ml) was drawn in the operating room immediately before surgery and placed in an ethylene diamine tetraacetic acid–treated tube. Each blood sample was identified using adhesive barcode labels. Barcode digits were registered twice in succession in the database to avoid misidentification. In addition to the blood samples, each collaborating center also retained the consent forms and questionnaires. Labeled samples were stored in a refrigerator at 4 to 5°C until they were shipped to a central laboratory within 1 week. The blood and questionnaires were then forwarded to the clinical laboratory (Echevarne Clinical Laboratory^†), where they were stored in a freezer at −80°C.

DNA was extracted with the QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Germany) following the manufacturer’s specifications. Genotyping was carried out with the Illumina Golden Gate protocol with VeraCode technology (Illumina^‡) in the National Genotyping Centre (CEGEN, Barcelona, Spain). The selected SNPs were genotyped for each patient.

A total of 90 SNPs were included in the study (table 1). Eight-seven of the 90 SNPs were selected based on prioritizing functional genetic variants previously associated in the literature with pain sensitivity, chronic pain conditions, and related traits belonging to different genes whose protein products are linked to biological pathways that influence pain sensitivity.^{15,18,20,32–35}  Thus, SNPs with no proven influence on gene function at the time were not included. These 87 SNPs had minor allele frequencies in the general Caucasian population of up to 0.4 (National Center for Biotechnology Information^§) and a homogeneous distribution along the gene and location inside the exons or near them, with a minor allele frequency of 0.1 using data from HapMap.^‖ The process was carried out according to the suggestions published by Hoh et al.,³⁶  and TagSNPs with R² > 0.8 were selected according to Carlson et al.³⁷  These SNPs are related to two main functional categories:

• Type 1: Genes encoding proteins that mediate the transmission of pain signals by sensory nerve fibers and by central nervous system pathways that mediate the perception of pain.

• Type 2: Genes encoding proteins that mediate peripheral and central inflammatory responses related to tissue injury.

Finally, we included the three significant SNPs detected in the genome-wide association study of acute postsurgical pain in humans by Kim et al.,²⁰  bringing the total number of SNPs to 90.

Data are expressed as medians and 10th to 90th percentiles. Potential risk factors were evaluated for unadjusted bivariate association with CPSP occurrence based on the t test (continuous variables) or the Fisher exact test or chi-square test (categorical variables). Bivariate ORs and 95% CIs were also calculated. Collinearity between categorical variables was tested with the Cramer V test (between nominal variables) and Kendall tau-b coefficient (between ordinal variables).

A generalized linear mixed model (GLMM) with the variable recruitment center as a random factor was constructed using backward stepwise selection with CPSP as the dependent variable. Independent variables were selected for the model on the basis of the investigators’ consensus on relevant measurable preoperative variables, the results of previous studies,^3–7,9,38 the bivariate analysis (P < 0.05), and correlation between variables (Kendall tau-b). At each step, the likelihood ratio was used to evaluate a potential risk factor. The cutoff for variable removal was set at a significance level of 0.05, and the adjusted ORs and corresponding 95% CIs were calculated.

A bootstrap method was used for internal validation of the subset of factors. A total of 1,000 computer-generated samples, each including 2,834 individuals, were derived from the sample by random selection with replacement. Within each bootstrap sample, the β coefficient was calculated using all selected factors. The reliability of predictor variables in the final GLMM was estimated by the 95% CI of the β coefficient in the bootstrap samples. Reliable predictors were retained if the 80% CI of bootstrap samples indicated statistical significance (P < 0.05). To assess the model’s discrimination and predictive ability, we used the c-statistic expressed as a percentage (area under the receiver operating characteristic curve). GLMM calibration was assessed by the Hosmer–Lemeshow goodness-of-fit statistic as an estimate of agreement between observed and predicted outcomes.

For each SNP, allele and genotype frequency associations between the CPSP status and the presence of neuropathic pain were tested using SNPator.³⁹  In the genotype analysis, different inheritance models were tested in autosomic SNPs by comparing each genotype against the combination of the remaining two. Chi-square-based Pearson tests were applied to the resulting contingency tables to test for association. Allele frequency associations with pain intensity were also explored using the Wald test implemented in the PLINK suite.⁴⁰  Additional allele frequency testing was performed for CPSP status according to sex and type of surgery.

Haplotype blocks were defined by grouping genotyped SNPs by proximity, disallowing gaps greater than 50 kb. Haplotypes for each individual at each block were estimated using PHASE.⁴¹  For each block, the frequency of each estimated haplotype was compared in cases and controls against the aggregation of all other estimated haplotypes for that block using SNPator.³⁹ 

We report nominal P values for all statistical tests and performed multiple testing correction by means of a conservative Bonferroni strategy that considered all the tests in our analysis even if they are not independent of each other. Given that we performed allelic and genotypic tests for every marker and haplotypic tests for every gene, for a total of over 400 tests, we used a Bonferroni threshold of 10^–4.

To evaluate the quality of recruitment and data collection, independent observers audited the medical records of a random sample of 5% of the patients from 6 randomly chosen centers. Thus, 38 patient records (1.3% of the sample) were audited; the 102 items checked encompassed all variables directly involved in the predictive model plus others. This audit found 110 instances of error or missing data (2.8% of the data audited).

For a total of 3,890 recruited patients, we detected protocol violation in 1% of the cases and 23.7% were lost to follow-up for the recording of outcome variables. Thus, data for 2,929 patients (75.3% of those recruited) were analyzed. Eighty-seven patients (3.0%) were lost between the first follow-up visit and the 2-yr telephone interview. Figure 1 shows patient flow from recruitment through 2 yr. Table 2 shows patient characteristics according to surgical procedure. DNA samples for genotyping were available for 2,854 patients (97.4%).

Figure 2 shows the CPSP incidence after each procedure and each data collection time. Within a median (10th to 90th percentile) of 4.4 months (3.7 to 5.8), CPSP had developed in 527 patients (18.0%), in 13.6% of patients after hernia repair, 11.8% after vaginal hysterectomy, 25.1% after abdominal hysterectomy, and 37.6% after thoracotomy. The follow-up interviews to report CPSP were completed at a median of 14.6 months (12.5 to 16.1) and 26.3 months (23.8 to 29.4).

Table 3 shows the incidence of CPSP, including neuropathic pain, and life-interference data obtained at the follow-up visit distributed by surgical procedure. The thoracotomy group had the highest incidence of neuropathic pain (55.0%) as assessed by the Douleur Neuropathique 4 questionnaire, and the vaginal hysterectomy group had the lowest (24.5%). The percentages of patients with a VNRS pain score higher than 3 ranged from 52.7% (thoracotomy) to 38.0% (hernia repair). CPSP interfered with daily activities (scores of > 3 out of 10, Brief Pain Inventory) after thoracotomy in 30.5% and after hernia repair in 18%. For patients who still had CPSP pain at 2 yr, the intensity did not decrease, remaining between 3 and 5 on the VNRS (table 4).

A total of 1,011 randomly selected samples (35.4% of the 2,854 available) were sent to be genotyped for 90 SNPs (505 cases, 506 controls); two samples were removed because of low genotyping success and four additional samples were removed because of incompatibilities between registered sex and sex imputed from genotypes. Thus, 1,005 samples (502 cases and 503 controls, table 5) were used in subsequent analyses. No significant deviations from Hardy–Weinberg equilibrium were found, and on comparing cases and controls (table 6), the subgroups were similar in all except two preoperative clinical variables (mental summary SF-12 score and preoperative pain in any nonsurgical area). The Bonferroni-corrected analysis showed no significant genetic differences in allele frequencies between patients with and without CPSP after any of the interventions studied (see table 7 and table 1, Supplemental Digital Content 3, https://links.lww.com/ALN/B137). Nonetheless, some allelic, genotypic, and haplotypic tests did show nominally significant P values for some SNPs or genes. In particular, two SNPs from DRD2 in chromosome 11 (rs12364283 and rs4648317) presented low P values in some tests but in no case went beyond a conservative Bonferroni threshold of approximately 10^–4.

After bivariate analysis of 31 independent variables (table 8), collinearity analysis (rejection of correlation coefficients higher than 0.25), 18 independent variables entered the GLMM. These candidate predictors were surgical procedure, body mass index (< 24.44, 24.44 to 28.08, > 28.08), anxiety (Hospital Anxiety and Depression Scale score ≥ 8), depression (Hospital Anxiety and Depression Scale score ≥ 8), substance addiction, chronic obstructive pulmonary disease, hypertension, neurologic disease, cancer, preoperative pain in the surgical area (VNRS score > 3), preoperative pain in other areas (pain score > 3), previous experiences of surgery-related pain, family history of surgery-related pain, type of anesthesia, intraoperative intravenous opioid use, age (< 51, 51 to 64, > 64 yr), SF-12 physical summary (0 to 33.5, 33.6 to 55.1, > 55.1), and SF-12 mental summary (0 to 44.8, > 44.8). (These cutoffs were determined by distributing the CPSP data in deciles.)

The multivariable GLMM selected six CPSP predictors (table 9), which were retained in more than 95% of the bootstrap subsamples. Table 9 shows the adjusted ORs for these variables, along with the 95% CIs after bootstrapping. This six-variable mixed model identified more than 73% of the patients who developed CPSP, with a c-statistic of 0.731 (0.705 to 0.755). The calibration was good according to the Hosmer–Lemeshow chi-square test (χ² = 4.02; P = 0.855). According to GLMM-derived β coefficients, an individual’s risk of CPSP might be calculated as follows: risk of CPSP = 1/(1 + e^{–linear predictor}) where the linear predictor comprising the six independent risk factors was as follows: −3.37 + 0.50 × surgery (abdominal hysterectomy) + 0.28 × surgery (hernia repair) + 1.88 × surgery (thoracotomy) + 1.13 × age (< 51 yr) + 0.48 × age (51 to 64 yr) + 0.86 × physical SF-12 (< 33.5) + 0.52 × physical SF-12 (33.5 to 55.1) + 0.51 × mental SF-12 (< 44.8) + 0.41 × preoperative pain in surgical area (VNRS > 3) + 0.37 × preoperative pain in other area (VNRS > 3).

Table 2, Supplemental Digital Content 3, https://links.lww.com/ALN/B137, shows the GLMM for five of the six predictors (excluding procedure type) applied to each of the four procedures (treating vaginal and abdominal hysterectomies separately). The c-statistics ranged from 0.731 (0.665 to 0.807) for vaginal hysterectomy to 0.645 (0.589 to 0.702) for thoracotomy. Table 3, Supplemental Digital Content 3, https://links.lww.com/ALN/B137, shows the GLMM for patients with neuropathic CPSP.

The overall incidences of CPSP confirmed on physical examination at 4 months are consistent with previously reported rates for the same procedures,^3,4  with the exception of vaginal hysterectomy. Previous authors reported similar CPSP rates for vaginal and abdominal hysterectomy,³⁸  but we observed different CPSP behavior after these procedures and consider them to be separate settings. The rates had decreased by approximately half 1 yr after surgery and by two thirds after 2 yr, but we found no long-term studies with which to compare that finding.

Also interesting was our finding of a lower 4-month neuropathic pain rate than reported by others^3,4,42 ; we attribute this difference to our reliance on exhaustive physical examination for diagnosis rather than postal questionnaires or patient charts. Finally, more than 20% of our CPSP patients reported moderate–intense pain at the diagnostic visit; that rate was similar to the 18.3% rate recorded in a population-based study of CPSP.²  For patients who were still experiencing pain 2 yr later, the intensity had not diminished. We emphasize that pain was responsible for moderate–intense interference with daily activities for 18% to 30%, with walking for 15% to 29%, and with mood for 10% to 34%.

The comparison between 502 patients with confirmed CPSP and 503 selected controls without CPSP showed that a strong effect of genetic profile on this late complication is unlikely. Under the CPSP diagnostic criteria we applied at 4 months, and with case–control sample sizes that rendered a power of approximately 99% for detecting ORs higher than 1.9 in individual allelic tests, we conclude that any potential effects of the tested SNPs would be weaker than that threshold. We cannot, of course, formally exclude associations with SNPs that were not selected for tagging in this study.

Even though none of the studied SNPs survived multiple test correction, we note that the lowest allelic P values were for the association between CPSP and the dopamine D2 receptor (DRD2) gene SNPs rs12364283 and rs4648317. Both have been associated with enhanced DRD2 expression⁴³  and several substance addictions—such as nicotine dependence (rs4648317⁴⁴ )—as well as with inhibition and impulsivity related to d-amphetamine response effects on stop-task performance and mood (rs12364283⁴⁵ ). These are but two of many SNPs linked to dopamine pathway dysregulation, which has also been observed in chronic pain unrelated to substance addiction.⁴⁶  Given the weak associations observed for these DRD2 SNPs, we believe that they might still be candidates for more complex polygenic and multifactorial modeling. One recent study demonstrated an association between the HLA DQB1*03:02 allele and higher CPSP risk after one of the procedures we included (inguinal hernia repair) and after lumbar disk herniation.²¹  We did not analyze HLA DQB1*03:02 because this pathway had not been directly linked to pain pathogenesis, but we think this new finding encourages further exploration of pathways not covered in the current study.

Thus, although we did not find any association between the 90 analyzed SNPs and CPSP, we cannot completely exclude the role of genetics in the development of CPSP. Our reasons are first, because our study was only powered to detect strong associations (OR > 1.9); second, because we selected SNPs to cover certain genes and did not exhaustively tag for all variations in every studied gene; and third, because reduced (or incomplete) penetrance, variable expressivity, and meiotic or mitotic epigenetic factors can contribute to the maintenance of CPSP. We think, however, that our findings do indicate that the positive results reported in other studies should probably be revisited critically until replicated.^21,47,48 

The CPSP risk model identified 73% of the patients with CPSP based on the following clinical predictors: (1) surgical procedure, (2) age, (3) physical health (SF-12 score), (4) mental health (SF-12 score), (5) preoperative pain in the surgical field, and (6) preoperative pain in another area. Although the discriminative power of the model is moderate, to our knowledge it is the first to offer some promise of assessing CPSP risk preoperatively, at least in the surgical settings studied. When we applied five of the six predictors (excluding procedure type) to each of the four procedures (treating vaginal and abdominal hysterectomies separately), we found that the model remained valid in each setting, although its predictive value is more robust in a mixed surgical population.

A clinical scoring system, based on the six easily recorded variables the model identifies, therefore merits external validation to test transportability to other settings. In contrast with a recent study by Althaus et al.,⁴⁹  who studied CPSP risk in a cohort of 150 patients undergoing a range of surgical procedures, we did not find that the presence of moderate or intense postsurgical pain substantially increased the predictive ability of the model. Thus, this factor was excluded for statistical reasons. However, we also emphasize our interest in identifying predictors available before surgery, such as psychological traits on which clinicians may be able to intervene. Such factors are probably related to patient hypervigilance⁵⁰  and are potentially related to certain gene polymorphisms in the dopamine pathway affecting pain perception.⁵¹  Of the five CPSP predictors identified by Althaus et al.,⁴⁹  our findings are consistent with two: preoperative pain in the operating field and other preoperative pain. The relevance of preoperative pain in another area of the body is possibly attributable to poor functioning of endogenous pain inhibition mechanisms, as has been demonstrated in patients who develop CPSP after thoracotomy¹⁴  and in several chronic pain settings.⁵² 

Factors related to surgery and anesthetic technique were not predictors of CPSP (table 4, Supplemental Digital Content 3, https://links.lww.com/ALN/B137). The lack of statistical relevance of anesthetic and analgesic variables may be attributable to the certain degree of variability in the execution of techniques in this study, reflecting a routine practice setting. In other words, an observational design, even when prospective, may make it difficult to identify factors as predictors if they are subject to small clinical variations. However, we did detect six other clinical risk factors for CPSP, and we think it is reasonable to suggest that anesthetic and analgesic factors may be less important to the development of CPSP in clinical circumstances than randomized trials might lead us to believe. Finally, although we detected a higher level of preoperative anxiety in patients with CPSP, anxiety was excluded from the model because it was strongly associated with results for the mental component of the SF-12, which was a stronger predictor.

One major strength of this study was its prospective, population-based, multicenter design with physician-diagnosed CPSP. We collected data for a representative random sample of surgical patients undergoing routine anesthetic procedures in a large genetically homogeneous population. This study was also the first to follow patients for 2 yr (range, 22.2 to 30.0 months) and to prospectively include all the variables considered relevant to chronic pain at the time of design.^53,54 

A potential limitation was the large number of data collectors (164 anesthesiologists) in 23 centers; however, we used a structured questionnaire and conducted three training sessions to prevent inconsistent collection that might have led to a center effect.⁵⁵  A second limitation was that we did not use a scale to analyze pain catastrophizing because the predictive value of this factor^10,56  had not yet been established when our study was designed. This attribute, defined as a tendency to misinterpret and exaggerate situations that may be threatening, has recently been shown to confer risk for CPSP.⁵⁶  A potential limitation with regard to hernia repair was that the type of mesh used for this procedure was not considered as a possible surgery-related risk factor. The final limitation relates to gender in the genetic analysis. We chose to include only men in the hernia repair and thoracotomy groups because of the difficulty in balancing gender in these samples: based on a previous descriptive study of surgical populations in our geographic setting, we estimated that women would only account for 28% and 22% of these groups, respectively.³¹  In order to analyze the genetic factor in relation to clinical characteristics within surgical specialties, while enrolling large but not vast numbers of patients, we balanced the all-male thoracotomy and hernia repair groups against the two hysterectomy groups.

We conclude that the lack of unequivocal confirmation of genetic factors predisposing certain patients to CPSP necessitates our continued reliance on scoring clinical factors—particularly procedure, age, and preoperative quality of life and experience of pain—to guide interventions or vigilance against the development of this late complication. A surgical team’s understanding of CPSP risk stratification has many applications in large healthcare systems or the management of individual cases. For benign conditions, high risk should lead to reassessment of surgery and deference to other treatment options,⁵⁷  especially in younger patients with concomitant pain or psychological comorbidity. Our model can facilitate trials of preventive strategies so that ineffective treatments that can have adverse effects or entail inconvenience can be avoided.^58,59  We encourage the further development of valid, transportable scoring systems to predict CPSP risk based on clinical factors in other surgical settings while the search for genetic and clinical interactions continues through more detailed multifactorial study.

The authors thank Joan García, M.Inf. (Dinamic Computer, Malgrat de Mar, Barcelona, Spain), for his technical assistance on the research group’s Web site and Mary Ellen Kerans, M.A. (Freelance Editor, Barcelona, Spain), who revised the English usage in some versions of the article.

This study was supported by grant 071210-2007 from Fundació La Marató de TV3 (Catalan public television network marathon foundation, Barcelona, Spain).

The authors declare no competing interests.

A. The GENDOLCAT Study Investigators by Center (n = 23). Centers are listed alphabetically; all cities are in Spain. PI refers to principal investigator; SC, steering committee.

Althaia, Xarxa Assistencial de Manresa, Manresa

Consuelo Ruiz (PI), M.D., Ph.D.; Carme Font, M.D.; Josep Delgado, M.D.; Lissette del M. Jiménez, M.D.; Mª Meritxell Sabrià, M.D.

Consorci Sanitari del Maresme, Hospital de Mataró, Mataró

Paloma Ricós (PI), M.D.; Antoni Pérez, M.D.; Carme Subirà, M.D.; Daniel Neira, M.D.; Francisco José Añez, M.D.; Gladys Margarita Hernández, M.D.; Isabel Cabré, M.D.; Laura Silberman, M.D.; Manuel Eduardo López, M.D.; María Garolera, M.D.; María Teresa Guerrero, M.D.; Marina García, M.D.; Montserrat Yuste, M.D.; Noemí Pou, M.D.; Rosa Calatayud, M.D.

Consorci Sanitari de Terrassa, Terrassa

Carmen Martín (PI), M.D., Ph.D.; Ester Lombán, M.D.; Gisela Egido, M.D.; José Antonio Bernia, M.D., Ph.D.; Ricardo Leiro, M.D.; Xavier García, M.D.

Corporació Sanitària Parc Taulí, Sabadell

Jordi Troy (PI), M.D.; Antonia Bassols, M.D., Ph.D.; Carme Colilles, M.D.; Cristina Tremps, M.D.; Diana Fernández, M.D., Ph.D.; Joan Blázquez, M.D.; Josep Planell, M.D.; Magdalena Serra, M.D.; Martí Solà, M.D.; Mercedes Rosas, M.D.; Montserrat Cañellas, M.D., Ph.D.

Fundació Privada Hospital Asil de Granollers

Victor Espiga (PI), M.D.; Fernando Martínez, M.D.; María Teresa Vilalta, M.D.

Hospital Clínic de Barcelona, Barcelona

Ana Bogdanovich (PI), M.D.; Pinar de Santos, M.D., Ph.D.; Teresa Anglada, M.D., Ph.D.

Hospital Comarcal Sant Bernabé, Berga

Anna Vidal (PI), M.D.; Josep Mª Canudas, M.D.

Hospital General Universitario de Alicante, Alicante

Ana Mª Peiró (PI), M.D., Ph.D.; Luis Gómez, M.D.; Yolanda Sastre, M.D.

Hospital Lluís Alcanyís, Xàtiva

Vicente Domingo (PI), M.D., Ph.D.; Antonio Antolí, M.D.; Belén Bardisa, M.D.; Blanca Moro, M.D.; Gerardo Presencia, M.D.; Julián García, M.D.; Mª Teresa Crespo, M.D.; Ramón José Ferri, M.D., Ph.D.; Vicente Roselló, M.D.

Hospital Municipal de Badalona, Badalona

Dolors Sintes (PI), M.D.; Ana Sapé, M.D.; Candi Giralt, M.D.; Eduardo Esteva, M.D.; Elena Barceló, M.D.; Fernando Rey, M.D.; Joan Fornaguera, M.D.; Luis Martínez, M.D.; Montserrat Pijoan, M.D.; Raquel Mansilla, M.D.

Hospital Sant Rafael, Barcelona

Jose Luis Casbas (PI), M.D.; Antonio Monso, M.D., Ph.D.; Inma Garrido, M.D.; Joaquín Torres, M.D.; Laura Mahillo, M.D.

Hospital Santa María, Lleida

Gregorio Marco (PI), M.D., Ph.D.; Dolors Del Pozo, M.D., Ph.D.; Montserrat Torra, M.D.; Reis Drudis, M.D.

Hospital Universitari de Bellvitge, Hospitalet de Llobregat

Ancor Serrano (PI), M.D.; Antoni Sabaté, M.D., Ph.D.; Carmen Ribes, M.D.; Eugenia Martínez, M.D.; María Valcárcel, M.D.; Resurrección Sanzol, M.D.; Víctor Mayoral, M.D.

Hospital Universitari de Girona Dr. Josep Trueta, Girona

Josep Vilaplana (PI), M.D.; Antonio Villalonga, M.D., Ph.D.; Antoni Arxer, M.D.; Joaquim Pujol, M.D.

Hospital Universitari de la Santa Creu i Sant Pau, Barcelona

Elena Català (PI), M.D., Ph.D.; Ana Bueno, M.D.; Bartomeu Cantallops, M.D.; Immaculada India, M.D.; Mª Angeles Gil de Bernabé, M.D.; Mª Carmen Unzueta, M.D., Ph.D.; Mª Teresa Rivilla, M.D.; Marc Bausili, M.D.; Miren Revuelta, M.D.

Hospital Universitari Germans Trias i Pujol, Badalona

Gisela Roca (PI), M.D., Ph.D.; Agnes Marti, M.D.; Alicia Melero, M.D.; Anna Flo, M.D.; Daniel Samper, M.D.; Jaume Canet, M.D., Ph.D.; Jordi Busquets, M.D.; Josefina Bassons, M.D., Ph.D.; Mar Monerris, M.D.; Marta Hinojosa, M.D.; Pere Vila, M.D., Ph.D.; Rosa Rincón, M.D.; Roser Garcia, M.D., Ph.D.

Hospital Universitari i Politècnic la Fe, Valencia

José María Loro (PI), M.D.; Aitana Lluch, M.D.; Ana Tur, M.D.; Antonio Abengochea, M.D., Ph.D.; Juan Gallego, M.D.; Manuel Barbera, M.D., Ph.D.; Maria Soledad Peláez, M.D.

Hospital Universitari Mútua de Terrassa, Terrassa

Domingo Jorge Echevarria (PI), M.D.

Hospital Universitari Sant Joan de Reus, Reus

Pilar Prieto (PI), M.D., Ph.D.; Dolors Pérez, M.D.; Ildefons Sardà, M.D.; Jesús Cuenca, M.D.; Lida Gonza, M.D.

Hospital Universitari Vall d’Hebron, Barcelona

Mª Victòria Ribera (PI), M.D., Ph.D.; Angela Mesas, M.D.; Anna González, M.D.; Carme Bosch, M.D.; Conxita Muñoz, M.D.; Cristina Aguilera, M.D.; Daniel Ruiz, M.D.; Dolors Mateo, M.D.; Gloria Agreda, M.D.; Imma Vives, M.D.; Jaume Roigé, M.D., Ph.D.; Mª Carmen Suescun, M.D.; Mª Dolores Dorda, M.D.; Patricia Bascuñana, M.D.; Patricia Ciurana, M.D.; Pilar Cortiella, M.D.; Susana Manrique, M.D.

Hospital de Viladecans, Viladecans

Dolors Pelegrí (PI), M.D.; Anna Abad, M.D.; Anna Faura, M.D.; Domingo Blanco, M.D., Ph.D.; Elisenda Izquierdo, M.D.; Mª Carmen López, M.D.; Mª José Linares, M.D., Ph.D.; Paco Nebot, M.D.; Robert Rodríguez, M.D.

Parc Hospitalari Martí i Julià, Hospital Santa Caterina, Salt

Santiago Beltrán (PI), M.D.; Benet Casagran, M.D. Ph.D.; Claudia Liliana Charry, M.D.; Olga Pineda, M.D.

Parc de Salut MAR, Barcelona

Susana Fernández (PI), M.D., Ph.D.; Alejandro Pérez, M.D.; Amelia Rojo, M.D.; Antonio Montes, M.D., Ph.D.; Carlos Alberto García, M.D.; Fernando Escolano, M.D., Ph.D.; Luis Moltó, M.D.; Nuria Baldomá, M.D.; Raquel Arroyo, M.D.; Ricardo López, M.D.; Saida Sánchez, M.D.

B. The GENDOLCAT Committees, with contributorships other than principal authors; listed alphabetically within sections. An asterisk (*) denotes a member of the writing team; some writers served on more than one committee.

Steering committee:

Ana Bogdanovich, M.D., Department of Anesthesiology, Hospital Clínic, Barcelona, Spain.

*Jaume Canet, M.D., Ph.D., Department of Anesthesiology, Hospital Universitari Germans Trias I Pujol, Universitat Autònoma de Barcelona, Badalona, Spain.

Elena Català, M.D., Ph.D., Department of Anesthesiology, Hospital Universitari de la Santa Creu i Sant Pau, Barcelona, Spain.

*Jose Ignacio Lao, M.D., Ph.D., Medical Director, Genomic Genetics International, Barcelona, Spain.

Victor Mayoral, M.D., Department of Anesthesiology, Hospital Universitari de Bellvitge, Hospitalet de Llobregat, Barcelona, Spain.

*Antonio Montes, M.D., Ph.D., Department of Anesthesiology, Parc de Salut MAR, Institut Municipal d’Investigació Mèdica, Universitat Autònoma de Barcelona, Barcelona, Spain.

Dolors Pelegrí, M.D., Department of Anesthesiology, Hospital de Viladecans, Viladecans, Spain.

Mª Victòria Ribera, M.D., Ph.D., Department of Anesthesiology, Hospital Universitari Vall d’Hebron, Barcelona, Spain.

*Gisela Roca, M.D., Ph.D., Department of Anesthesiology, Hospital Universitari Germans Trias i Pujol, Universitat Autònoma de Barcelona, Badalona, Spain.

*Sergi Sabaté, M.D., Ph.D., Department of Anesthesiology, Fundació Puigvert, Barcelona, Spain.

Jordi Troy, M.D., Department of Anesthesiology, Corporació Sanitària Parc Taulí, Sabadell, Spain.

Statistical and data coordination:

*Jordi Cantillo, B.Sc.(Statistics), Department of Anesthesiology, Parc de Salut MAR, Institut Municipal d’Investigació Mèdica, Universitat Autònoma de Barcelona, Barcelona, Spain.

Carlos Morcillo, M.D., Ph.D., Institute of Evolutionary Biology (Universitat Pompeu Fabra-CSIC), National Institute for Bioinformatics, Institució Catalana de Recerca I Estudis Avançats, Barcelona, Spain.

*Arcadi Navarro, Ph.D., Department of Experimental and Health Sciences (Universitat Pompeu Fabra), and Institute of Evolutionary Biology (Universitat Pompeu Fabra-CSIC), Centre de Regulació Genòmica, National Institute for Bioinformatics, Institució Catalana de Recerca I Estudis Avançats, Barcelona, Spain.

*Sergi Sabaté, M.D., Ph.D., Department of Anesthesiology, Fundació Puigvert, Barcelona, Spain.

Design, analysis, and interpretation of genetic study:

*Jose Ignacio Lao, M.D., Ph.D., Medical Director, Genomic Genetics International, Barcelona, Spain.

Carlos Morcillo, M.D., Ph.D., Institute of Evolutionary Biology (Universitat Pompeu Fabra-CSIC), National Institute for Bioinformatics, Institució Catalana de Recerca I Estudis Avançats, Barcelona, Spain.

*Arcadi Navarro, M.D., Ph.D., Institute of Evolutionary Biology (Universitat Pompeu Fabra-CSIC), Centre de Regulació Genòmica, National Institute for Bioinformatics, Institució Catalana de Recerca I Estudis Avançats, Barcelona, Spain.

Advice on health-related quality-of-life assessment and measurement:

Gemma Vilagut, M.D., Ph.D., Health Services Research Unit, IMIM (Hospital del Mar Medical Research Institute), Barcelona, Spain; CIBER for Epidemiology and Public Health (CIBERESP), Madrid, Spain.