{"paper_id":"79632af6-5bee-4728-9a07-4ad1eaa475cc","body_text":"Endometriosis is defined as the presence of endometrial‐like epithelial and stromal cells in extra‐uterine locations [ 1 ,  2 ,  3 ]. Endometriosis can be divided into three anatomical subtypes: deep infiltrating endometriosis (DIE), ovarian endometrioma (OMA), and superficial peritoneal endometriosis (SUP) [ 4 ]. Treatment includes hormonal suppression or surgical resection of endometriosis lesions. Surgical staging of endometriosis commonly utilizes the revised American Society for Reproductive Medicine (rASRM) classification, based on anatomical subtype amount and distribution, depth of endometriosis lesions, and presence of adhesions and OMAs [ 5 ]. The rASRM classification includes Stage I (minimal; score of 0–5), Stage II (mild; score of 6–15), Stage III (moderate; score of 16–40), and Stage IV (severe; score of >40) disease, where complete obliteration of the posterior cul‐de‐sac, commonly due to DIE, and large OMAs, are typically present [ 5 ,  6 ]. A diverse group of symptoms are associated with this condition including infertility, painful periods, sexual pain, and chronic pain. Despite ~190 million affected people worldwide, much remains to be understood about the fundamental molecular processes driving disease progression [ 4 ].\nThough endometriosis is a benign condition, malignant transformation occurs in approximately 1% of cases, largely restricted to the OMA subtype and leading to endometriosis‐associated ovarian cancers [ 7 ]. Somatic cancer‐driver mutations have been described in endometriosis adjacent to endometriosis‐associated ovarian cancers, and have been proposed to be necessary but not sufficient for malignant transformation [ 8 ]. A growing literature has also identified somatic cancer‐driver mutations in endometriosis not associated with cancer, across anatomic subtypes and restricted to endometriotic epithelium [ 9 ,  10 ,  11 ,  12 ,  13 ,  14 ]. Somatic activating  KRAS  mutations appear to be the most common somatic mutations currently reported in endometriosis, ranging from 19.4 to 46.7% of cases based on previous literature [ 12 ,  15 ,  16 ,  17 ]. Endometriosis can have tumor‐like qualities such as local invasiveness, proliferation, resistance to apoptosis, and spread, which leads to more anatomically severe disease [ 12 ,  16 ,  17 ,  18 ,  19 ,  20 ]. Only recently, studies have begun to explore somatic cancer‐driver mutations and endometriosis phenotypes [ 17 ].\nWe recently reviewed the literature on somatic mutations and other somatic genomic alterations in endometriosis, and proposed a protocol for clinical phenotyping to facilitate reproducibility [ 21 ]. This protocol emphasizes pathology review and tissue enrichment given that endometriotic epithelium only forms a minority of surgically excised tissue specimens. Proposed clinical annotation in the protocol includes anatomic features, pain and fertility measures, and prospective follow‐up of outcomes.\nIn this study, we describe a prospective longitudinal cohort of patients undergoing surgery for endometriosis, with somatic activating  KRAS  codon 12 mutation testing in excised endometriosis lesions after pathology review and tissue enrichment. Subjects were followed 5–9 years after surgery, with detailed clinical annotation. The objective of the study was to characterize the clinical phenotype of  KRAS  mutations in endometriosis. We hypothesized that  KRAS  mutations would be associated with greater anatomic disease burden in endometriosis (i.e. more severe anatomic subtypes and higher stage).\n\nThe setting of the study was the BC Centre for Pelvic Pain and Endometriosis, a tertiary referral center for endometriosis. Surgically excised endometriosis was prospectively biobanked beginning in 2013 (ENDOONC study; REB H11‐00536 and H14‐03040), while baseline and ongoing longitudinal follow‐up clinical data were collected as part of a prospective registry at our center (EPPIC registry, Clinicaltrials.gov #  NCT02911090 , REB H11‐02882, and H16‐00264) [ 22 ,  23 ]. The EPPIC registry systematically collected real‐time patient reported data (e.g. pain severities, ethnicity) and physician reported data (e.g. physical exam, recording of surgical procedures and findings). Standardized follow‐up in the EPPIC registry involved annual patient reported questionnaires to 2 years after baseline, as well as physician report of any repeat surgeries (re‐operations) at the center until study end in May 2022. Prospective consent was sought for both the biobank and the registry, which were linked to allow for correlation of somatic mutations with detailed phenotypic data.\nInclusion criteria were consecutive subjects with index surgeries (see Table  1  for list of procedures) at the center between 2013 and 2017 who gave informed consent to both biobanking and the registry. Subjects were excluded if they were post‐menopausal (spontaneous or surgical) or had a history of cancer or co‐existing cancer; and tissue samples were excluded from the  KRAS  mutation assay if they had insufficient endometriotic epithelial cells. See supplementary material, Figure  S1  for the study flowchart.\nPatient and sample characteristics. Description of the study sample, including patient characteristics, findings at the index surgery, and procedures performed at the index surgery\nAge  n  = 121.\nThere were no cases requiring ureteric reimplantation, bowel resection, or bladder wall resection in this cohort.\nWe employed a selective sampling strategy of the endometriosis lesions in each subject for the  KRAS  mutation assay. For subjects with DIE, all available DIE lesions were sampled for the  KRAS  mutation assay. For subjects with OMA, one OMA was sampled, plus an OMA on the contralateral ovary was also sampled if available. For subjects with SUP, one SUP lesion was sampled, plus a second SUP lesion at a different anatomical location was also sampled if available. Note that each subject could have one, two, or three anatomic subtypes concurrently. Each anatomic subtype was diagnosed by the surgeon (e.g. >5 mm invasion for DIE).\nSamples were enriched by manual needle macrodissection or laser capture microdissection (LCM). Manual macrodissection was sufficient when the epithelial content of endometriosis lesions was sufficiently large (one or more clusters of epithelial glands appearing >1 mm 2  on visual estimation) and there was limited surrounding tissue (i.e. lesion was close to the edge of the sample). We observed no difference in the rate of detectable mutations between LCM and macrodissection enrichment (supplementary material, Table  S1 ). For each sample, the tissue was stained with hematoxylin and eosin (H&E) to identify endometriosis cells (i.e. pathology review). Sequential sections of a formalin‐fixed paraffin embedded block were then sectioned at 8 μm onto glass slides or polyethylene naphthalate membrane slides (Leica Microsystems Inc., Heerbrugg, Switzerland) for needle macrodissection or LCM, respectively. Tissue was deparaffinized with xylene and stained with 10% diluted H&E to enable capture with minimal surrounding fibrotic tissue. Manual macrodissection was done under a stereomicroscope with a 20‐gauge needle (Thermo Fisher Scientific, Waltham, MA, USA) [ 15 ]. LCM was performed on an LMD7000 (Leica Microsystems Inc., Switzerland) [ 12 ]. DNA was extracted using an Arcturus® PicoPure® DNA Extraction Kit (Thermo Fisher Scientific) and quantified using the Qubit 2.0 Fluorometer (Invitrogen; Thermo Fisher Scientific).\nDNA was pre‐amplified using primers flanking the codon 12 region (see  Supplementary materials and methods ). Droplet digital PCR (ddPCR), with a multiplex ddPCR screening assay to pre‐screen samples, was used to detect six  KRAS  codon 12 variants (c.34G>T (p.G12C), c.35G>A (p.G12D), c.34G>C (p.G12R), c.35G>T (p.G12V), c.35G>C (p.G12A), c.34G>A (p.G12S)). Empirical testing showed the combination of c.34G>T (p.G12C), c.35G>A (p.G12D), and c.34G>C (p.G12R) probes (Thermo Fisher Scientific) was sufficient to detect and resolve unique clusters for all  KRAS  variants (supplementary material, Figure  S2 ). Any positive variant was confirmed with a variant‐specific ddPCR assay that was also used to establish variant allele frequency ( Supplementary materials and methods ; and supplementary material, Table  S2 ). Note that enrichment of the endometriotic epithelium was not done (for LCM or macrodissection) and thus the variant allele frequency of the epithelial fraction is not known. Droplets were generated using a BioRad QX200 Automated Droplet Generator (BioRad Laboratories, Hercules, CA, USA), and quantified on a BioRad QX200 Droplet Reader (BioRad Laboratories). The mean detection thresholds and ranges for the individual  KRAS  assays in this study are listed in supplementary material, Table  S3 .\nFor  KRAS  mutation status, the subjects were divided into two groups (present versus absent): present was defined as at least one  KRAS  mutation in at least one sample assayed in a subject, and absent was defined as the absence of  KRAS  mutations in all samples assayed in a subject.\nThe primary analysis was for an association between  KRAS  mutation status and anatomic phenotyping at the index surgery based on the anatomic subtypes of DIE, OMA, and SUP that were each confirmed on pathology. Subjects were phenotyped by (1) the presence of one or more subtypes and (2) by rASRM staging of the subtypes. Significance was set at  α  = 0.05 for the hypothesis that  KRAS  mutation would be associated with more severe anatomic disease. A subanalysis was carried out in the subgroup of subjects with only one sample assayed for  KRAS  mutation.\nFor secondary exploratory analyses, we examined for associations between the  KRAS  mutation variable and five types of outcome: other anatomic findings at the index surgery; markers of surgical difficulty; fertility variables at baseline; demographics at baseline; and pain variables at baseline. Mean differences and relative risks (RRs) are reported with 95% confidence intervals (CIs). For re‐operation during the follow‐up period, time to re‐operation based on  KRAS  mutation status was analyzed using Kaplan–Meier survival analysis with log‐rank test. For longitudinal follow‐up of pain scores after surgery based on  KRAS  mutation status, we evaluated the change in pain scores and also used linear regression for follow‐up pain scores while controlling for baseline pain scores.\nAnalyses were done using IBM SPSS Statistics 25 (SPSS Inc, Chicago, IL, USA). Missing data were assumed to be missing at random and excluded based by pairwise deletion (available‐case analysis). Data analysis was performed by NLO, AA, YDL, and PJY.\n\nThis study included 122 subjects with a mean age of 34 ± 7 years. The distribution of subjects based on rASRM surgical stage was: Stage I – 23.8% (29/122), Stage II – 16.4% (20/122), Stage III – 22.1% (27/122), and Stage IV – 35.3% (43/122) (Table  1 ). The distribution of subjects based on anatomic subtype (each subtype confirmed as endometriosis on pathology) is illustrated in Figure  1 , with half (54.1%; 66/122) of subjects having more than one subtype. Additional demographics for the cases are in supplementary material, Table  S4 .\nCohort classification based on endometriosis anatomic subtype. Venn diagram of the 122 subjects in the study categorized by anatomic subtypes. Note that not all anatomic subtype lesions were sampled for  KRAS  mutation testing due to our selective sampling strategy (see  Materials and methods  section) and material availability constraints (lack of endometriotic epithelial cells).\nFrom the 122 subjects, a total of 262 endometriosis lesions were sampled for the  KRAS  mutation assay based on our selective sampling strategy: 105 DIE samples, 44 OMA samples, and 113 SUP samples (supplementary material, Table  S5 ). Most of the 122 subjects had one sample ( n  = 51) or two samples ( n  = 36) assayed for a mutation (range 1–7). In total, 52.5% (64/122) of the subjects had at least one endometriosis sample with a  KRAS  mutation, while 47.5% (58/122) had no  KRAS  mutations in any of the endometriosis samples assayed. The frequency of different codon 12 mutations was  KRAS  G12C ( n  = 4),  KRAS  G12D ( n  = 17),  KRAS  G12R ( n  = 5),  KRAS  G12V ( n  = 33),  KRAS  G12A ( n  = 11), and  KRAS  G12S ( n  = 7) – which includes 13 dual positives (i.e. individuals found to have more than one mutation). We observed the same  KRAS  G12 variant in more than one lesion in the same patient in eight (6.6%) patients, and  KRAS  G12D was more likely to be clonal in our cohort. The proportion of subjects with a  KRAS  mutation varied by which anatomic subtype(s) were present, with the lowest proportion in subjects with only SUP lesions (Figure  2 ).\nPrevalence of  KRAS  mutation based on anatomic subtypes. Number (percentage) of cases with  KRAS  mutation in the each of the following groups: subjects with DIE lesions only, OMA lesions only, SUP lesions only, and different combinations of anatomic subtypes.\nFor the primary outcome of anatomic phenotyping based on subtype, subjects with at least one  KRAS  mutation affected lesion were compared with subjects without mutation (Table  2 ). To meet assumptions of chi‐square testing, we categorized subjects as: (1) SUP only; (2) DIE only or OMA only; and (3) mixed (two or more) subtypes.  KRAS  mutation was present in 35.1% (13/37) of subjects with SUP lesions only, 57.9% (11/19) with DIE only or OMA only, and 60.6% (40/66) of those with mixed anatomic subtypes (chi‐square,  p  = 0.04) (Table  2 ). When the DIE or OMA only and the mixed subtypes were grouped together (versus SUP only cases),  KRAS  mutation had a RR of 1.36 (95% CI: 1.06–1.74) (Table  2 ), supporting an association between  KRAS  mutation and more anatomically severe disease.\nPrimary analysis. Bivariate analyses for associations between  KRAS  mutation and the primary outcome of anatomic phenotyping of endometriosis\nChi‐square test for 2 × 3 table (anatomic subtype) and 2 × 4 table (stage).\nRR for  KRAS  mutation as the ‘exposure’.\nStage  n  = 119.\nSimilarly, a  KRAS  mutation was present in 27.6% (8/29) of rASRM Stage I, 65.0% (13/20) of Stage II, 63.0% (17/27) of Stage III, and 58.1% (25/43) of Stage IV cases (chi‐square,  p  = 0.02) (Table  2 ). When Stage II–IV cases were grouped together (versus Stage I cases),  KRAS  mutation had a RR of 1.40 (95% CI = 1.12–1.75) (Table  2 ), again supporting an association between  KRAS  mutation and more anatomically severe (advanced stage) disease.\nA subanalysis was performed in those subjects that had only one sample assayed for mutation ( n  = 51). In this subgroup,  KRAS  mutation had an RR of 1.60 (95% CI = 0.98–2.60) for DIE or OMA only or mixed subtypes grouped together (versus SUP only cases). Likewise, in this subgroup,  KRAS  mutation had an RR of 1.60 (95% CI: 1.07–2.38) for Stage II–IV cases grouped together (versus Stage I cases).\nSecondary exploratory analyses are summarized in Table  3  for other anatomic findings at the index surgery, markers of surgical difficulty, fertility variables, demographics, and pain scores at baseline.  KRAS  mutation was associated with a higher risk of ureterolysis (retroperitoneal dissection of the ureter) being required during the index surgery (RR = 1.47, 95% CI: 1.02–2.12) (Table  3 ). We were limited in stratification of ethnicities with our cohort being predominantly Caucasian, and thus the sample was categorized into Caucasian versus other ethnicities (East and Southeast Asian ethnicities being the most prevalent [57.1%]). Subjects of Caucasian ethnicity were less likely to have a  KRAS  mutation (RR = 0.64, 95% CI: 0.47–0.89). There was no evidence of associations with other secondary outcomes (Table  3 ).\nSecondary analyses. Bivariate analyses for associations between  KRAS  mutation, and the following categories of variables: anatomic findings at the index surgery, difficulty of the index surgery, fertility variables, demographics, and pain scores\nMD, mean difference.\nRR for  KRAS  mutation as the ‘exposure’ (except for ethnicity, where Caucasian was considered the ‘exposure’). MD for continuous or 0–10 variables.\nLeft OMA  n  = 33, right OMA  n  = 34, index surgery time  n  = 114, parity  n  = 121, infertility  n  = 121, age  n  = 121, dysmenorrhea  n  = 110 (excludes those with no menses at baseline), deep dyspareunia  n  = 117 (excludes those not sexually active).\nOther (non‐Caucasian) ethnicities included: East or Southeast Asian ( n  = 16), South Asian ( n  = 6), Hispanic ( n  = 5), Other ( n  = 1). Mixed ethnicities group excluded.\nFor the assessment of re‐operation at the center, duration of prospective longitudinal follow‐up ranged from 5 to 9 years. The rate of re‐operation (for endometriosis) was low in the cohort overall (13.9%; 17/122), with 17.2% (11/64) of subjects with  KRAS  mutation and 10.3% (6/58) of subjects without  KRAS  mutation having a re‐operation (RR = 1.66, 95% CI: 0.66–4.21). The curve of re‐operation free survival appeared lower for subjects with  KRAS  mutation compared with subjects without  KRAS  mutation but with no statistically significant difference on Kaplan–Meier survival analysis log‐rank testing (supplementary material, Figure  S3 ). For the 17 patients who have re‐operation data within our center, no patients have developed malignancies.\nPain scores (11‐point numeric rating scale [ 24 ]) were evaluated longitudinally after surgery at 2‐year follow‐up (or 1‐year follow‐up, if 2‐year data not available). Overall, pain scores decreased over time after surgery: dysmenorrhea decreased from 7.3 ± 2.6 to 3.6 ± 3.1, deep dyspareunia decreased from 6.2 ± 3.1 to 4.3 ± 3.2, dyschezia decreased from 4.9 ± 3.2 to 2.7 ± 2.5, and chronic pelvic pain decreased from 6.3 ± 2.7 to 3.5 ± 3.1. These changes over time were not associated with  KRAS  mutation status (supplementary material, Figure  S4 ). Similarly, using linear regression with follow‐up pain score as the outcome (controlling for baseline pain), no associations with  KRAS  mutation were identified (supplementary material, Table  S6 ).\n\nIn this prospective longitudinal study of 122 subjects who underwent endometriosis surgery at a tertiary referral center, somatic  KRAS  codon 12 cancer‐driver mutations were associated with more anatomically severe endometriosis (i.e. those with DIE or OMA only or mixed subtypes, and higher rASRM stage).\nWe detected  KRAS  mutation in 58–65% of the more anatomically severe cases. This figure is higher than previously reported for OMA (38.5–46.7%) or for DIE (19.4–33.3%) [ 12 ,  15 ,  16 ,  17 ,  25 ], likely due to sampling more lesions per case than in previous work or perhaps related to the patient population at our tertiary center likely representing more severe cases. Our mutation prevalence is also higher than prior studies utilizing lower resolution detection methods or without enrichment for endometriosis cells [ 9 ,  26 ].\nThe association between  KRAS  mutations and greater anatomic disease burden suggests that these activating mutations may play a role in lesion growth, invasion, or spread [ 13 ,  21 ]. We also noted an association between  KRAS  mutation and ureterolysis as a marker of surgical difficulty. This association with surgical difficulty is likely influenced by the association between  KRAS  mutation and advanced stage disease, as the latter often requires ureterolysis to separate the ureter from surrounding adhesions and fibrosis, prior to excision of endometriosis lesions.\nKRAS  mutations were less common among Caucasian subjects and more common among subjects of other ethnicities (predominantly East and Southeast Asian). Anatomically severe endometriosis was previously reported to be more common in East and Southeast Asian individuals in our registry [ 27 ]. It is possible that this epidemiological observation is accounted for by a higher rate of somatic events in  KRAS  in non‐Caucasian individuals with endometriosis, at least in our population. While intriguing, larger studies in other settings are needed to validate this possible association and investigate potential mechanisms.\nNo trends were observed for  KRAS  mutation status and baseline pain scores or change in pain scores over time. The lack of association is consistent with the marginal correlation between anatomic severity of endometriosis and pain symptoms, as the pathophysiology of endometriosis‐associated pain is multifactorial and can involve central sensitization [ 2 ,  3 ,  4 ,  28 ].\nRe‐operation rates were overall low at this tertiary referral center for endometriosis (13.9% over 5–9 years of follow‐up). This may reflect the experience of high‐volume endometriosis surgeons at our center, though we cannot rule out the possibility that patients returned to and sought re‐operation in the community. A nonsignificant reduction in re‐operation free survival time after the index surgery was observed in those having a  KRAS  mutation. Since surgery involves excision of visible endometriosis disease, it is conceivable that microscopic residual endometriosis cells harboring  KRAS  mutations may be more likely to cause recurrent disease. However, confirmation of this difference would require a future study with larger sample size. Curiously, subjects with  KRAS  mutant endometriosis appeared to be less likely to have a prior surgery before the index surgery (Table  3 ). This may be an artefact of referral of suspected higher stage cases to our tertiary center, though biological influence cannot be ruled out and improved outcomes in some Ras‐harboring cancers have been reported [ 29 ,  30 ].\nA major strength of this study is its prospective longitudinal design, with surveillance of re‐operation up to 5–9 years of follow‐up. Furthermore, registry data were entered in real time, ensuring higher accuracy and avoiding risk of bias from prior knowledge of mutation status. Moreover, highly sensitive ddPCR testing was employed, ensuring detection of subclonal alterations that have been reported previously, and enabling application of moderate enrichment methods that are operationally more feasible than laser capture or single‐cell molecular assays (e.g. macrodissection). However, we recognize that somatic alterations in  ARID1A, PIK3CA , and other oncogenes and tumor suppressors have been reported and may coexist with our observed  KRAS  mutations contributing to clinical phenotypes and heterogeneity [ 12 ,  15 ,  17 ,  31 ]. Likewise, we tested only for the most common hot‐spot,  KRAS  codon 12 variants [ 21 ]. While reports of  KRAS  mutations outside of codon 12 alteration in endometriosis (and malignancies) are moderately rare [ 21 ] we cannot exclude the possibility that these may be present and not accounted for in our population. Whole genome or exome approaches [ 12 ,  14 ,  31 ] may also reveal complex genomic landscapes associated with clinical phenotypes. Due to enrichment via macrodissection in the majority of specimens in our study, we could not reliably test for associations between (increasing/decreasing) allele frequency and clinicopathological features. In particular, the use of macrodissection results in apparent low mutant allele frequencies (supplementary material, Table  S5 ); allele frequencies would be higher if endometriotic epithelium were isolated by LCM.\nIt should be noted that we did not sample every lesion in each subject, but utilized a selective sampling strategy based on anatomic subtype. By definition, cases with more anatomically severe endometriosis will have more lesions to be sampled, and one would expect a greater probability of finding at least one mutation by chance with a higher number of samples assayed. For this reason, we performed a subanalysis in the subgroup with only one lesion sampled according to our sampling strategy, and observed increased RRs that were similar to the whole group analysis. Optimally, complete sampling of all lesions within each patient in future research would enable a per lesion analysis of mutation rate (versus the per subject analysis in this study). Validation of the findings in other cohorts is also required to confirm generalizability. In addition, our cases are from a tertiary referral center and thus the results cannot be reliably extrapolated to the general population.\nWe hypothesize that somatic activating  KRAS  codon 12 mutations may contribute to constitutive activation in downstream pathways that results in proliferation, local invasion, and metastatic spread resulting in advanced stage disease with elevated surgical complexity [ 17 ,  21 ]. Models of endometriosis, in particular whole‐animal models, will be necessary to investigate putative mechanisms [ 32 ,  33 ,  34 ,  35 ,  36 ,  37 ]. While targeting of  KRAS  has historically been challenging, new opportunities have been presented including variant‐specific agents and synthetic lethal strategies [ 38 ,  39 ,  40 ]. Work presented here suggests the  KRAS  axis may be a relevant target to reduce the invasiveness, spread, or burden of endometriosis.\nFurthermore, somatic mutations such as in  KRAS  may inform a novel molecular classification of endometriosis. This future molecular classification may incorporate other somatic genomic alterations, germline polymorphisms, RNA expression changes, immunohistochemistry markers, together integrated with clinical variables. Ultimately, a clinically useful molecular classification for endometriosis should demonstrate correlations with baseline phenotype and response to treatment. In the future, molecular subtypes of endometriosis could be built into clinical trial design and subtype‐specific treatments could be incorporated into care.\n\nKRAS  somatic‐cancer driver mutations were associated with greater anatomic disease burden in endometriosis and thus more surgical complexity.  KRAS  mutations may serve as a nonhormonal therapeutic target and contribute to a molecularly informed classification in endometriosis.\n\nNLO, AFL, MAB, MSA and PJY conceptualized the study. NLO, AL, JH, CLI, JS and TMN carried out experiments. NO, TMN, AL, KL, MSA and PJY designed the methodology and curated the data. NOL, AA, YDL and PJY analyzed the data. AFL, MAB, MSA and PJY acquired funding. CA, CW, MAB, MSA and PJY provided study resources. HN led project administration. NLO, MSA and PJY wrote the original draft. All authors were involved in reviewing, editing and final approval of the manuscript.\n\nSupplementary materials and methods\nFigure S1.  Study flowchart\nFigure S2.  Multiplex assay with all six  KRAS  codon 12 positive controls, each with a unique clustering region\nFigure S3.  Kaplan–Meier survival analysis\nFigure S4.  Change in pain scores between baseline and follow‐up, based on mutation status\nTable S1.  Comparison between manual microdissection and laser capture microdissection\nTable S2.  Cell line controls\nTable S3.  Mean detection thresholds for the individual  KRAS  assays\nTable S4.  Demographics\nTable S5.  Summary of mutation calls\nTable S6.  The association between follow‐up pain scores and  KRAS  mutation status, controlling for baseline pain severity – linear regression\nClick here for additional data file.","source_license":"CC0","license_restricted":false}