Normative Functional Outcomes as a New Outcome Assessment Tool Following Hip Procedures

• Updated on: 2019-06-18
• Read original article here

Normative Functional Outcomes as a New Outcome Assessment Tool Following Hip Procedures
Liselore Maeckelbergh, MD; Tine Peeters, MD; Joseph Moskal, MD; Kristoff Corten, MD
Orthopedics. 2018;41(5):e663-e670
Abstract
Abstract
Patient-reported outcome measures play an important role in evaluating the functional outcome of surgical and nonsurgical treatments of the hip joint. One thousand healthy volunteers completed the modified Harris hip score, the University of California, Los Angeles score, the Hip Osteoarthritis Outcome Score, and the Western Ontario and McMaster Universities Osteoarthritis Index score. Between September 2010 and December 2015, a total of 127 periacetabular osteotomies were performed in 111 patients with symptomatic developmental dysplasia of the hip. Forty-two of these patients (10 male and 32 female) met inclusion criteria. Mean follow-up was 32 months (range, 13–59 months). Pre- and postoperative radiographic analysis of the lateral center-edge angle and the acetabular index was conducted in all cases with a proper pelvic anteroposterior radiograph. The patients completed the modified Harris hip score, the University of California, Los Angeles score, the Hip Osteoarthritis Outcome Score, and the Western Ontario and McMaster Universities Osteoarthritis Index score. The authors investigated the influence of various confounding variables to (1) obtain recommendations when outcome scores are being compared between 2 cohorts and (2) define a normative reference level of “hip-healthy” functionality. This normative level of functionality was used as a target level of functionality following a hip procedure such as periacetabular osteotomy. All functional outcome scores had significantly improved 1 year after periacetabular osteotomy; thus, patients were much better than preoperatively. However, only 55% achieved the 95% functionality of the normative population based on modified Harris hip score and University of California, Los Angeles score. The results were worse for the Hip Osteoarthritis Outcome Score subscales. This approach places the results of surgical procedures in a different but potentially more realistic perspective in terms of expectations and goals. [Orthopedics. 2018; 41(5):e663–e670.]
Full Text
Abstract
Patient-reported outcome measures play an important role in evaluating the functional outcome of surgical and nonsurgical treatments of the hip joint. One thousand healthy volunteers completed the modified Harris hip score, the University of California, Los Angeles score, the Hip Osteoarthritis Outcome Score, and the Western Ontario and McMaster Universities Osteoarthritis Index score. Between September 2010 and December 2015, a total of 127 periacetabular osteotomies were performed in 111 patients with symptomatic developmental dysplasia of the hip. Forty-two of these patients (10 male and 32 female) met inclusion criteria. Mean follow-up was 32 months (range, 13–59 months). Pre- and postoperative radiographic analysis of the lateral center-edge angle and the acetabular index was conducted in all cases with a proper pelvic anteroposterior radiograph. The patients completed the modified Harris hip score, the University of California, Los Angeles score, the Hip Osteoarthritis Outcome Score, and the Western Ontario and McMaster Universities Osteoarthritis Index score. The authors investigated the influence of various confounding variables to (1) obtain recommendations when outcome scores are being compared between 2 cohorts and (2) define a normative reference level of “hip-healthy” functionality. This normative level of functionality was used as a target level of functionality following a hip procedure such as periacetabular osteotomy. All functional outcome scores had significantly improved 1 year after periacetabular osteotomy; thus, patients were much better than preoperatively. However, only 55% achieved the 95% functionality of the normative population based on modified Harris hip score and University of California, Los Angeles score. The results were worse for the Hip Osteoarthritis Outcome Score subscales. This approach places the results of surgical procedures in a different but potentially more realistic perspective in terms of expectations and goals. [Orthopedics. 2018; 41(5):e663–e670.]
Patient-reported outcome measures play an important role in evaluating the functional results of surgical and nonsurgical treatments of the hip joint. 1–3 Functional outcome scores have been developed as an instrument to assess patients' opinion about the hip and associated problems. These scores aim to assess the functionality and disability of the hip. Frequently used scores are the modified Harris hip score (mHHS), the University of California, Los Angeles (UCLA) score, the Hip Osteoarthritis Outcome Score (HOOS), and the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) score. 4–7 Although they have been validated for clinical use, little is known about the variables that should be considered when interpreting these scores.
In addition, normative values of the most frequently used scoring systems represent the normative level of functionality. These values can be used as baseline levels of functionality that could be aimed for after a hip procedure. Studies can be initiated with a predefined goal of functionality as a percentile of the normative population. This would allow assessment of the predictability of a hip procedure to reach a normative level of functionality. The Bernese periacetabular osteotomy (PAO) described by Ganz et al 8 is a hip joint preservative procedure that redirects the acetabulum in patients with a developmental dysplasia of the hip. 9 Although the improvement of functional outcomes following PAO has been well described, how predictably patients can achieve the 5th percentile of a normative level of functionality remains unknown. Such knowledge would allow realistic expectations to be set for patients undergoing this procedure.
First, the authors investigated which variables significantly influenced functional outcomes (mHHS, UCLA score, HOOS, and WOMAC score) in a population of 1000 healthy volunteers. Second, they compared the level of functionality of a cohort of 42 patients before and after PAO with that of the “normative” database of age- and sex-matched hip-healthy volunteers.
Materials and Methods
One thousand volunteers were asked to complete the mHHS, the UCLA score, the HOOS, and the WOMAC score. These individuals did not have the following: 1 or both hips operated on, restriction in range of motion or debilitating pain in the groin and trochanter region or elsewhere in the body, use of pain medication, treatment for hip or groin problems, a history of surgery at the lower limb or spine with residual symptoms or restrictions, joint replacement surgery, fibromyalgia, or systemic diseases such as lupus, rheumatoid arthritis, or psoriatic arthritis. They were asked about pain in the lower back, knee, ankle, and foot. They were also asked about “occasional but nondebilitating groin pain not requiring any further investigation, treatment, or pain medication.” The groin was indicated, with the use of a drawing, as the area extending from the inner thigh to the anterior border of the greater trochanter and from the anterior–superior iliac spine to the level of the base of the greater trochanter. Sex, age, body mass index, and the presence of pain from the lower back to the lower limb were considered predictors for functional outcome. A distinction was made among knee pain, pain in the foot or ankle, joint pain elsewhere, and occasional groin pain. A combined group of back and/or lower limb pain was also created if the individual had pain in at least 1 site (ie, the lower back, knee, foot, or ankle) but not in the groin region.
Between September 2010 and December 2015, a total of 127 PAOs were performed in 111 patients with symptomatic developmental dysplasia of the hip. The senior author (K.C.) performed the Bernese PAO as described by Ganz et al. 8 The modified Smith–Petersen approach was used, but the rectus femoris tendon was not tenotomized and the joint was not opened. 8 Fluoroscopy was used to evaluate the correction intraoperatively. Two to 4 cortical screws were used to achieve fixation. Screws were removed in all cases at a minimum of 3 months postoperatively. Patients without major intra- or postoperative complications (N=97) and with a minimum of 1 year of follow-up and no additional treatment of the involved or the contralateral hip were included for clinical and functional outcome assessment. The functional questionnaires were incomplete or absent for 10 patients. A total of 42 patients (10 male and 32 female) with PAO remained after the exclusion criteria were applied. Mean age at the time of surgery was 27 years (range, 14–50 years). Mean follow-up was 32 months (range, 13–59 months). Pre- and postoperative radiographic analysis of the lateral center-edge angle and the acetabular index was conducted in all cases with a proper pelvic anteroposterior radiograph as described by Sink et al. 10
The Spearman correlation and the Mann–Whitney U test were used to evaluate the association between each predictor and each functional outcome score or subscore. Ages 20 through 70 years were input in 10-year ranges. A multivariable linear regression model was used with the predictors for each outcome score separately. A model was fit separately on the subjects with and without occasional groin pain. The overall R2, which reflects the total percentage of variability explained by the considered predictor, of the multivariable model was calculated. For each predictor in the multivariable model, the semi-partial R2 was then calculated. This presents the percentage of variability in the functional score that is explained by the particular predictor after controlling for the other predictors (ie, the variability uniquely caused by the specific predictor). The Student's t test was used to compare values between sexes and groups of subjects with pain in the low back, knee, or groin region. P<.05 was considered significant. All analyses were performed using SAS version 9.2 software for Windows (SAS Institute Inc, Cary, North Carolina).
Statistical analysis comparing the outcomes of PAOs with those of the normative databases was performed using the SPSS 23 statistical software package (IBM, Armonk, New York). Gaussian distribution was tested using the Kolmogorov–Smirnov test. All outcomes of the subgroups were compared using the Wilcoxon signed rank test. To compare the results of the PAO population with those of the normative database, a binomial test was used to determine the proportion of patients with PAO who were below the 5th percentile of the normal population. Significance was set as P<.05.
Results
Variables Influencing Functional Outcomes
The functional scores of 37 subjects were discarded because either they were incomplete or the subject reported debilitating pain. This left 963 complete scores of 324 (34%) males and 639 (66%) females. Mean body mass index was 24 kg/m2 (range, 11–38 kg/m2), and 730 (76%) subjects were in the age categories of 20 to 60 years. In total, 386 (40%) subjects reported not having pain. The incidence of occasional groin pain was 21% in 144 (23%) female and 56 (17%) male subjects. Mean scores are presented in Table 1 .
Table 1:
Functional Scores
Multivariate analysis showed that groin pain was the most important predictor of worse results in 13 of 15 scores because it was associated with the highest semi-partial R2 values (Table 2 ). The mHHS, UCLA pain score, WOMAC score, and WOMAC stiffness score were especially highly influenced by the presence of groin pain. The UCLA activity score was influenced more by age (semi-partial R2=9.3%) than by groin pain (semi-partial R2=1.3%). Once groin pain was controlled for, age became the second most important predictor of variability in outcome scores. Except for the UCLA activity score, age did not influence the scores of subjects without pain and younger than 70 years. Body mass index was only a minor predictor of worse outcomes, having low semi-partial R2 values. Except for the UCLA activity (P=.009) and pain (P=.003) scores, for which females had worse results, sex did not significantly influence the outcomes. When groin pain was controlled for, only the WOMAC and HOOS subscores for pain were different between the sexes. Females with occasional groin pain had significantly worse results than males with groin pain for 9 of 15 scores.
Table 2:
Summary and Overview of the Most Important Predictors of Variability in Outcome Scores
Based on the multivariate analysis, it was clear that occasional groin pain and age were the 2 most important predictors of outcome variability. This indicated that age-matched scores of subjects without groin pain could be considered as the normative healthy scores because they controlled for the 2 most influential variables regarding functionality. Because groin pain influenced the outcome scores of females differently from those of males, the scores should also be matched for sex.
Comparing Periacetabular Osteotomy With the Normative Dataset
All patients had a pain-free range of motion at both hips, and all hips showed Tönnis grade 0 cartilage degenerative changes on pelvic radiograph. The mean lateral center-edge angle increased by 16°, from 17° preoperatively to 33° postoperatively. The acetabular index decreased by 14°, from 15° preoperatively to 1° postoperatively. Forty (95%) patients were satisfied with the procedure and reported being willing to undergo the same procedure again. All functional scores postoperatively improved significantly compared with preoperatively (Tables 3 – 6 ). On comparison of the scores with those of the normative data-set, 19 (45%) of 42 patients had worse age- and sex-matched mHHS and UCLA score than the worst 5th percentile normative results (P<.001) (Table 7 ). Similarly, significantly more patients with PAO had worse results than the worst 5th percentile normative outcome scores in all subscales of the HOOS (P<.0001) (Table 7 ). Between 36% and 48% of patients did not have the same level of functionality as age- and sex-matched healthy subjects without groin pain (P<.0001).
Table 3:
Normative Outcomes for Modified Harris Hip Score
Table 4:
Normative Outcomes for University of California, Los Angeles Score
Table 5:
Normative Outcomes for Western Ontario and McMaster Universities Osteoarthritis Index Score
Table 6:
Normative Outcomes for Hip Osteoarthritis Outcome Score
Table 7:
Preoperative and Postoperative Functional Scores for Periacetabular Osteotomy
Discussion
Functional outcome scores are used to assess the functional recovery of patients after a hip procedure. As such, they are useful tools for monitoring the changes in functionality from pre- to postoperatively. However, little is known about the confounding variables that should be considered when these outcome scores are being evaluated. In addition, it remains unclear what the target functional goals should be in the contemporary patient population. The authors considered the functionality of hip-healthy subjects as the normative functional goal to achieve with their hip procedures. This not only sets clear goals but also defines realistic expectations for patients. In this study, the authors sought to collect the most frequently used outcome scores from a large cohort of subjects who considered themselves hip healthy. The authors investigated the influence of various confounding variables to (1) obtain recommendations when outcome scores are being compared between 2 cohorts and (2) define a normative reference level of hip-healthy functionality. The latter was used as a target level of functionality following a hip procedure such as PAO.
The first goal of this study was to collect a large dataset of subjects who considered themselves hip healthy. The authors acknowledge that there were some limitations. First, although the authors included the scores of subjects who also had nondebilitating pain at other joints, they did not investigate the true origin of this pain. Multivariate analysis showed that this pain did not influence the functional outcomes to the same extent as occasional groin pain and age. Furthermore, the authors think that normative functional outcomes should take these minor confounding variables into account because 60% of their healthy subjects had some pain at other joints. Therefore, the investigated cohort can be considered a representative cohort for a normative age-matched functionality. Another limitation was that the authors did not investigate the origin of occasional but nondebilitating groin pain. Subjects with this might have had a hip problem. Subjects did not undergo a clinical examination or radiographs. To the best of the authors' knowledge, no correlations between functional outcome scores and clinical examination or radiographs have been reported. Of interest, 21% of the investigated population reported occasional groin pain that did not require further treatment or pain medication. The authors acknowledge that a more objective gradation of the pain level might have been better. However, this finding indicates that groin pain is not an infrequent symptom. The analysis showed that groin pain was the most important confounding variable for functional outcomes; therefore, the normative reference level was based on a large dataset of subjects without groin pain. In addition, this finding suggests that the functionality of 2 cohorts can be reliably compared only if patients with groin pain on the contralateral hip are excluded. The authors further acknowledge that they narrowed hip-related pain to groin pain. This was based on the study by Lavigne et al, 11 who found that 70% of subjects with hip complaints had preoperative groin pain. 12 Pain at the trochanter and the buttock might also be relevant hip-related symptoms, but there are no guidelines on their correlation with hip pathology. Finally, other scores also could have been used. The authors did not intend to detect the best score for functional outcome assessment; rather, they investigated the most frequently used outcome scores found in their literature search. 2,3,11,13–22 The authors' only aim was to optimize the utility of these scores by investigating their confounding variables.
All scores were influenced by the presence of any pain from the spine to the foot. However, multivariate analysis identified occasional groin pain and age as the most important predictors of variability in outcomes. Both should be controlled for as much as possible when functionality is compared between 2 procedures. This is in accordance with a study by Bin Nasser et al, 14 who compared the outcomes of hip resurfacing with and without groin pain. They found worse outcomes in the cohort with groin pain. 14 Similarly, Lavigne et al 11 found worse WOMAC and UCLA activity scores in those subjects with groin pain following hip replacement. 12 The current authors conclude that the pre- and postoperative incidence of groin pain at both sides should be reported because groin pain at the nonoperated hip significantly influences functional outcomes. Therefore, patients with contralateral groin pain should be excluded. When pain was controlled for, except for the UCLA activity score, age between 15 and 70 years did not influence the scores. The scores of subjects without pain represent the maximum achievable outcome scores following a hip procedure. However, 60% of the subjects had some pain in joints other than the hip; therefore, a representative population would be a population with pain at different joints. In these cases, age does become a predictor of variability in outcome and thus should be controlled.
The incidence of groin pain was higher in females than males, and females with groin pain had significantly worse outcomes than males. However, sex did not influence the scores in the pain-free population. That is, males and females should have similar outcome scores except for WOMAC and HOOS pain scores, which are slightly worse in females. If females and males have significantly different UCLA scores and WOMAC outcome subscores, this is most likely due to a difference in the incidence of groin pain. The mHHS is not sensitive enough to detect any difference between sexes in relation to groin pain.
The need for contemporary normative levels of functional outcomes as a new target for postoperative functionality can be illustrated by means of the mHHS. The mHHS, which has been validated by Byrd and Jones, 15 takes the same functional parameters into consideration but discards the clinical assessment. In the original article, 15 a HHS of 100 to 90 was arbitrarily categorized as excellent, 90 to 80 as good, 80 to 70 as fair, and less than 70 as poor. The current authors stratified subjects without groin pain into sex-matched age groups, which represents the normative dataset. They assessed the numbers of subjects within the 4 HHS categories, finding that only 84% of female subjects between 30 and 50 years old attained an excellent score. In addition, only 63% of females between 50 and 60 years old attained an excellent score, and 37% of healthy subjects had a good score. Thus, the arbitrary categories excellent and good can be questioned in the contemporary population and are related to age and sex. Furthermore, it seems unrealistic that a greater number of subjects had excellent scores in clinical studies than in the normative dataset. It would be better to compare the functional results of a surgical procedure with those of a normative dataset of contemporary age- and sex-matched subjects. The authors propose an alternative method by using normative levels of functionality that they aim to achieve. They suggest using a postoperative target functionality that is based on the normal contemporary population. An age- and sex-matched functionality similar to 95% of the normative population without groin pain can be targeted. For example, to define this functionality, the authors assessed per age group the minimum mHHS score that 95% of subjects without groin pain achieved. In the age group of 50 to 60 years, 95% of subjects had a minimum mHHS of 77 and 81 for females and males, respectively. Five percent had a worse mHHS. If an incidence of worse outcomes higher than 5% is found for patients within the same age category, it can be concluded that the procedure is not able to reproducibly achieve the normative level of functionality. Similarly, more stringent criteria such as 10th or 25th percentile could also be used.
This new approach sets realistic expectations for patients and health care providers. For example, all functional outcome scores significantly improved 1 year after PAO. Patients were thus much better than preoperatively, but only 55% achieved the 95% functionality of the normative population based on the mHHS and the UCLA score. The results were even worse for the HOOS subscales. This approach places the results of surgical procedures in a different but potentially more realistic perspective in terms of expectations and goals. In addition, such an approach also bypasses the ceiling effect that is found in most of the functional outcome scores. Finally, using normative outcome scores also allows comparison of the predictability in expected outcome of 2 surgical procedures relative to the normative reference dataset.
Conclusion
This approach permits assessment of how predictable a surgical technique is in achieving a predefined goal. It also allows comparison of the predictability of 2 techniques to achieve a goal. If the level of functionality is not within the normative level, the result can be considered as not normal. That is, normative data allow definition of an evidence-based goal of functionality that surgeons want to achieve with their surgical procedures; the authors want to achieve the same level of functionality after surgery as that among age-matched hip-healthy peers. Although these normative data may not necessarily be compatible with the expectations of some patients, they at least represent a predefined and realistic minimum goal that the authors aim to achieve.
References
Ashby E, Grocott MP, Haddad FS. Outcome measures for orthopaedic interventions on the hip. J Bone Joint Surg Br. 2008; 90(5):545–549. doi:10.1302/0301-620X.90B5.19746 [CrossRef]
Marx RG, Jones EC, Atwan NC, Closkey RF, Salvati EA, Sculco TP. Measuring improvement following total hip and knee arthroplasty using patient-based measures of outcome. J Bone Joint Surg Am. 2005; 87(9):1999–2005. doi:10.2106/JBJS.D.02286 [CrossRef]
Terwee CB, Bot SD, de Boer MR, et al. Quality criteria were proposed for measurement properties of health status questionnaires. J Clin Epidemiol. 2007; 60(1):34–42. doi:10.1016/j.jclinepi.2006.03.012 [CrossRef]
de Groot IB, Reijman M, Terwee CB, et al. Validation of the Dutch version of the Hip disability and Osteoarthritis Outcome Score. Osteoarthritis Cartilage. 2009; 17(1):132. doi:10.1016/j.joca.2008.05.014 [CrossRef]
Gosens T, Hoefnagels NH, de Vet RC, et al. The “Oxford Heup Score”: the translation and validation of a questionnaire into Dutch to evaluate the results of total hip arthroplasty. Acta Orthop. 2005; 76(2):204–211. doi:10.1080/00016470510030580 [CrossRef]
Roorda LD, Jones CA, Waltz M, et al. Satisfactory cross cultural equivalence of the Dutch WOMAC in patients with hip osteoarthritis waiting for arthroplasty. Ann Rheum Dis. 2004; 63(1):36–42. doi:10.1136/ard.2002.001784 [CrossRef]
Wamper KE, Sierevelt IN, Poolman RW, Bhandari M, Haverkamp D. The Harris hip score: do ceiling effects limit its usefulness in orthopedics?Acta Orthop. 2010; 81(6):703–707. doi:10.3109/17453674.2010.537808 [CrossRef]
Ganz R, Klaue K, Vinh TS, Mast JW. A new periacetabular osteotomy for the treatment of hip dysplasias: technique and preliminary results. Clin Orthop Relat Res. 1988; 232:26–36.
Biedermann R, Donnan L, Gabriel A, Wachter R, Krismer M, Behensky H. Complications and patient satisfaction after periacetabular pelvic osteotomy. Int Orthop. 2008; 32(5):611–617. doi:10.1007/s00264-007-0372-3 [CrossRef]
Sink EL, Leunig M, Zaltz I, Gilbert JC, Clohisy JAcademic Network for Conservational Hip Outcomes Research Group. Reliability of a complication classification system for orthopaedic surgery. Clin Orthop Relat Res. 2012; 470(8):2220–2226. doi:10.1007/s11999-012-2343-2 [CrossRef]
Lavigne M, Laffosse JM, Ganapathi M, Girard J, Vendittoli P. Residual groin pain at a minimum of two years after metal-on-metal THA with a twenty-eight-millimeter femoral head, THA with a large-diameter femoral head, and hip resurfacing. J Bone Joint Surg Am. 2011; 93(suppl 2):93–98. doi:10.2106/JBJS.J.01711 [CrossRef]
Vendittoli PA, Ganapathi M, Roy AG, Lusignan D, Lavigne M. A comparison of clinical results of hip resurfacing arthroplasty and 28 mm metal on metal total hip arthroplasty: a randomised trial with 3–6 years follow-up. Hip Int. 2010; 20(1):1–13. doi:10.1177/112070001002000101 [CrossRef]
Bartelt RB, Yuan BJ, Trousdale RT, Sierra RJ. The prevalence of groin pain after metal-on-metal total hip arthroplasty and total hip resurfacing. Clin Orthop Relat Res. 2010; 468(9):2346–2356. doi:10.1007/s11999-010-1356-y [CrossRef]
Bin Nasser A, Beaulé PE, O'Neill M, Kim PR, Fazekas A. Incidence of groin pain after metal-on-metal hip resurfacing. Clin Orthop Relat Res. 2010; 468(2):392–399. doi:10.1007/s11999-009-1133-y [CrossRef]
Byrd JW, Jones KS. Prospective analysis of hip arthroscopy with 10-year followup. Clin Orthop Relat Res. 2010; 468(3):741–746 doi:10.1007/s11999-009-0841-7 [CrossRef]
Gastwirth JL. Percentile modifications of two sample rank tests. J Am Stat Assoc. 1965; 60(312):1127–1141. doi:10.1080/01621459.1965.10480856 [CrossRef]
Harris WH. Traumatic arthritis of the hip after dislocation and acetabular fractures: treatment by mold arthroplasty. An end-result study using a new method of result evaluation. J Bone Joint Surg Am. 1969; 51(4):737–755. doi:10.2106/00004623-196951040-00012 [CrossRef]
Howie DW, Beck M, Costi K, Pannach SM, Ganz R. Mentoring in complex surgery: minimising the learning curve complications from peri-acetabular osteotomy. Int Orthop. 2012; 36(5):921–925. doi:10.1007/s00264-011-1347-y [CrossRef]
Liebs TR, Herzberg W, Rüther W, Haasters J, Russlies M, Hassenpflug J. Ergometer cycling after hip or knee replacement surgery: a randomized controlled trial. J Bone Joint Surg Am. 2010; 92(4):814–822. doi:10.2106/JBJS.H.01359 [CrossRef]
Martin R, Clayson PE, Troussel S, Fraser BP, Docquier PL. Anterolateral minimally invasive total hip arthroplasty: a prospective randomized controlled study with a follow-up of 1 year. J Arthroplasty. 2011; 26(8):1362–1372. doi:10.1016/j.arth.2010.11.016 [CrossRef]
McElduff F, Cortina-Borja M, Chan SK, Wade A. When t-tests or Wilcoxon-Mann-Whitney tests won't do. Adv Physiol Educ.2010; 34(3):128–133. doi:10.1152/advan.00017.2010 [CrossRef]
Thas O, Clement L, Rayner JC, Carvalho B, Van Criekinge W. An omnibus consistent adaptive percentile modified Wilcoxon rank sum test with applications in gene expression studies. Biometrics. 2012; 68(2):446–454. doi:10.1111/j.1541-0420.2012.01750.x [CrossRef]
Table 1