Abstract
Introduction
Vaginal cuff dehiscence (CD) after hysterectomy is a rare but serious complication of robotic-assisted laparoscopic total hysterectomy (RLTH). The authors of this study aimed to compare the incidence and risk factors of CD following RLTH among patients with and without endometrial cancer.
Methods
This retrospective study included women aged 18 years or older who underwent RLTH by two surgeons at a single institution from 2013 to 2018. Patients with conversion to laparotomy, recent chemotherapy or radiation, or non-uterine malignancy were excluded. Data were abstracted from medical records.
Results
Of 950 patients meeting inclusion criteria, 50.7% had endometrial cancer. CD was reported in 2.5% of all patients. While adjusting for cancer status, age, sexual activity after surgery, distance from home to location of surgery, and time interval from surgery to loss to followup, obese patients were 25.1% less likely than non-obese patients to experience CD (62.5 vs. 37.5, p = 0.01). Surgeon A had a 2.8 times higher CD rate than surgeon B (70.8 vs. 29.2, p = 0.03). No other factors predicted CD.
Conclusions
Endometrial cancer patients were not at greater risk of experiencing CD compared to non-cancer patients. Surgeon differences and body mass index (BMI) were associated with CD risk, with normal BMI patients at higher risk.
Keywords: hysterectomy, surgical wound dehiscence, endometrial neoplasms, body mass index
INTRODUCTION
In gynecologic oncology, endometrial cancer is the most common indication for robotic-assisted laparoscopic total hysterectomy (RLTH).1 Vaginal cuff complications following RLTH, including dehiscence, have been reported.2,3 Vaginal cuff dehiscence (CD) is the separation of the anterior and posterior fibromuscular edges of the vaginal cuff, with or without bowel evisceration.3 CD has an incidence of 0.4% to 4.1% following RLTH, a considerable increase compared to other methods of total hysterectomy.2–7
Variances in vaginal cuff closure, including suture technique, material, and approach, may alter the risk of CD following RLTH.5,8–10 Risk of CD following RLTH also is associated with vaginal atrophy, tobacco use, obesity, and/or diabetes.11 Post-operative trauma to the healing cuff, such as sexual activity, also is known to be a potential precipitating factor for CD.10–12 Risk of CD following RLTH may also be associated with malignancy, specifically endometrial cancer.13 This study sought to determine if there was a difference in the incidence and risk factors of CD following RLTH among patients with endometrial cancer compared to patients without endometrial cancer.
METHODS
Authors of this retrospective review utilized data abstracted from electronic medical records at a single institution. This study was approved by The University of Kansas Medical Center Institutional Review Board (IRB). Patients 18 years or older who had undergone RLTH (CPT codes: 58570, 58571, 58572, 58573) performed by two gynecologic oncology surgeons from 2013 through 2018 were included in this study. Patients were excluded if they had undergone chemotherapy and/or radiation within a year before or after the RLTH, had primary malignancies other than endometrial cancer, or had conversion to laparotomy during the RLTH.
Demographics, comorbidities, cancer status, cancer descriptors, RLTH surgical information, CD occurrence, precipitating events, dehiscence characteristics, and method of repair were abstracted and managed using REDCap® electronic data capture tools hosted at the University of Kansas Medical Center.14,15
Statistical Analysis
SAS version 9.4 (SAS Int. Inc., Cary, NC) was used for data analysis. CD incidence after RLTH among endometrial cancer patients was 1.3%16 and among patients without endometrial cancer was 0.85%.17 With these percentages, the power was calculated to be 0.859, indicating a required sample size of 1,000 patients. Two proportions using the Pearson’s Chi-square approach were used, and normal approximation was assumed to approximate the sampling distribution of the difference in proportions between the two groups. This method assumed that the sampling distribution of the difference in proportions followed a normal distribution for the large sample size. The overall number of participants in the final analysis fell short of the intended sample size of 1,000 patients. Consequently, the power analysis was re-evaluated retrospectively using the sample size of 950, resulting in a power of 0.84, which met the statistical criteria for acceptability.
Frequencies, proportion, means, and standard deviations were calculated. Associations between nominal or categorical variables were tested using Pearson’s Chi-square, likelihood ratio Chi-square, and Fisher’s exact test. The Mann-Whitney U test assessed differences in dehiscence levels over the surgery to follow-up time interval. In cases of non-normal distribution, the rank transformation approach combined ranking and general linear modeling. Univariate and multiple logistic regression models employing Firth’s bias-reduction penalized maximum likelihood estimation and explored the association between CD and explanatory variables. Fisher scoring, based on the expected information matrix, was used as an iterative algorithm. Statistical significance was set at p < 0.05 for all tests.
RESULTS
Out of 1,208 patients, 21.3% (n = 258) were excluded from the study. The remaining 950 patients were divided into two groups: those with endometrial cancer (50.7%, n = 482) and those without cancer (49.3%, n = 468).
The patients’ ages ranged from 23 to 95 years, with an average age of 56 years (SD = 13.3; Table 1). Most patients were White or Caucasian (87.6%, n = 832), post-menopausal (65.4%, n = 600), and lived less than 50 miles from the surgery location (59.8%, n = 569). Among patients with endometrial cancer, the majority had Stage IA (80.1%, n = 386), Grade 1 disease (67.0%, n = 323), had an endometrioid histologic type (90.9%, n = 438), and received pelvic lymph node dissection at the time of surgery (75.1%, n = 362).
Table 1.
Patient demographics and clinical characteristics.
| Characteristics | Endometrial Cancer Patients n = 482 |
Non-Cancer Patients n = 468 |
Chi-Square (df ) | P-Value |
|---|---|---|---|---|
| Age | 230.5 (1) | <0.001 | ||
| 18 to 49 Years | 48 (10) | 263 (56.2) | ||
| 50 Years or Older | 434 (90) | 205 (43.8) | ||
| Race | 10 (3) | 0.020 | ||
| White or Caucasian | 437 (90.6) | 395 (84.4) | ||
| Black or African American | 13 (2.7) | 22 (4.7) | ||
| Asian American | 4 (0.8) | 12 (2.6) | ||
| Other | 28 (5.8) | 39 (8.3) | ||
| Ethnicity | 3.8 (2) | 0.150 | ||
| Hispanic or Latino | 24 (5) | 30 (6.5) | ||
| Not Hispanic or Latino | 457 (95) | 433 (93.5) | ||
| Body Mass Index (BMI) | 75.8 (4) | <0.001 | ||
| Less than or equal to 18.5 | 1 (0.2) | 13 (2.8) | ||
| 18.6 to 24.9 | 46 (9.5) | 107 (22.9) | ||
| 25 to 29.9 | 77 (16) | 116 (24.8) | ||
| 30 to 39.9 | 194 (40.3) | 151 (32.3) | ||
| Greater Than or Equal to 40 | 164 (34) | 81 (17.3) | ||
| Menopause Status (n = 917) | 193.5 (1) | <0.001 | ||
| Postmenopausal | 407 (86.8) | 193 (43.1) | ||
| Premenopausal | 62 (13.2) | 255 (56.9) | ||
| Parity (n = 948) | 7.8 (4) | 0.100 | ||
| 0 | 113 (23.4) | 95 (20.4) | ||
| 1 | 53 (11) | 70 (15) | ||
| 2 | 136 (28.2) | 151 (32.4) | ||
| 3 | 110 (22.8) | 98 (21) | ||
| 4 or more | 70 (14.5) | 52 (11.2) | ||
| Hypertension Status | 55.5 (1) | <0.001 | ||
| No | 214 (44.4) | 320 (68.4) | ||
| Yes | 268 (55.6) | 148 (31.6) | ||
| Diabetes | 37.2 (3) | <0.001 | ||
| Not Diabetic | 361 (74.9) | 418 (89.3) | ||
| Diabetes Type 1 | 2 (0.4) | 4 (0.9) | ||
| Diabetes Type 2 | 116 (24.1) | 44 (9.4) | ||
| Diabetes Type Unknown | 3 (0.6) | 2 (0.4) | ||
| Smoking Status | 18.3 (3) | <0.001 | ||
| Never Smoker | 362 (75.9) | 306 (66.8) | ||
| Current Smoker (at time of surgery) | 31 (6.5) | 65 (14.2) | ||
| Former Smoker | 84 (17.6) | 87 (19) | ||
| Distance from Home to Surgical Center (n = 946) | 7.8 (3) | 0.049 | ||
| Less Than 50 Miles | 274 (57) | 295 (63.4) | ||
| 50 to 99 Miles | 77 (16) | 80 (17.2) | ||
| 100 to 199 Miles | 97 (20.2) | 67 (14.4) | ||
| 200 miles or more | 33 (6.9) | 23 (5) | ||
| Surgeon | 0.7 (1) | 0.400 | ||
| Surgeon A | 245 (50.8) | 225 (48.1%) | ||
| Surgeon B | 237 (49.2) | 243 (51.9%) | ||
| Endometrial Cancer Stage | ||||
| Stage IA | 386 (80.1) | |||
| Stage IB | 75 (15.6) | |||
| Stage II | 3 (0.6) | |||
| Stage III | 15 (3.1) | |||
| Stage IV | 2 (0.4) | |||
| Unknown/Not Recorded | 1 (0.2) | |||
| Endometrial Cancer Grade | ||||
| Grade 1 | 323 (67) | |||
| Grade 2 | 114 (23.7) | |||
| Grade 3 | 43 (8.9) | |||
| Unknown/Not Recorded | 2 (0.4) | |||
| Endometrial Histologic Type | ||||
| Carcinosarcoma | 2 (0.4) | |||
| Endometrioid | 438 (90.9) | |||
| Serous | 24 (5) | |||
| More Than One | 12 (2.5) | |||
| Other | 5 (1) | |||
| Unknown/Not Recorded | 1 (0.2) | |||
| Pelvic Lymph Node Dissection Performed at Surgery | ||||
| No | 120 (24.9) | |||
| Yes | 362 (75.1) | |||
Data are presented as n (%), unless otherwise indicated. Some totals may not add up to 100% due to missing values.
The suture type most frequently used for vaginal cuff closure by both surgeons was Polyglactin 9/10 (92.0%, n = 874). The most common suture technique for vaginal cuff closure (95.9%, n = 911) involved two sutures running from lateral to midline.
More non-cancer patients (57.1%) reported being sexually active after surgery than cancer patients (32.5%; χ2 [2, N = 947] = 51.5, p < 0.001). A greater proportion of non-cancer patients (63.4%) lived less than 50 miles from the surgery location compared to endometrial cancer patients (57.0%; χ2 [3, N = 946] = 7.8, p = 0.049). There was no significant difference in the number of endometrial cancer patients treated by surgeon A compared to surgeon B.
CD occurred in 2.53% (n = 24) of all patients. The incidence of CD among endometrial cancer patients was 1.9% (n = 9) and 3.2% (n = 15) among non-cancer patients (χ2 [1, N = 768] = 2.2, p = 0.139). Among patients diagnosed with CD, 29.2% (n = 7) reported a precipitating event, with intercourse being the most reported trigger (85.7%, n = 6). Of all CD occurrences, 16.7% (n = 4) involved overt evisceration of intra-abdominal contents, while 83.3% (n = 20) were classified as occult separation of the vaginal cuff, defined as a visibly or palpably thin membrane with peritoneal fluid or abdominal contents pushing against it, with or without a surrounding ring of scar tissue.
On average, patients were lost to follow-up 284 days after surgery. Patients living closer to the surgery location were lost to follow-up after a longer period (F[3, 942] = 3.5, p = 0.015) than those living further away. Patients living 50 miles or less were lost to follow-up 320 days after surgery, compared to 181 days for patients living 200 or more miles away. Endometrial cancer patients were lost to follow-up an average of 422 days post-surgery, compared to 140 days for non-cancer patients.
Among patients with CD, the condition was reported an average of 138 days post-surgery, with a median of 58 days post-surgery. Lastly, patients with CD were lost to follow-up 532 days post-surgery on average.
As shown in Table 2, CD was more common among patients with a normal BMI (41.7%) compared to overweight (25.0%) and obese patients (29.2%; χ2 [2, N = 768] = 13.2, p = 0.001). Of the 24 cases of CD, 70.8% (n = 17) were patients of surgeon A (χ2 [1, N = 768] = 6.7, p = 0.010).
Table 2.
Testing association between CD and patient factors.
| Patient Factors | Cuff Dehiscence | |||
|---|---|---|---|---|
| No | Yes | Chi-Square (df) | P-value | |
| Age | 0.7 (1) | 0.390 | ||
| 18 to 49 years | 247 (33.2) | 10 (41.7) | ||
| 50 years or older | 497 (66.8) | 14 (58.3) | ||
| BMI | 13.2 (3) | 0.001 | ||
| Less than or equal to 18.5 | 11 (1.5) | 1 (4.2) | ||
| 18.5 to 24.9 | 114 (15.3) | 10 (41.7) | ||
| 25 to 29.9 | 154 (20.7) | 6 (25) | ||
| Greater than or equal to 30 | 465 (62.5) | 7 (29.2) | ||
| Diabetes | 3.1 (1) | 0.080 | ||
| Patients with diabetes | 134 (18) | 1 (4.2) | ||
| Patients without diabetes | 610 (82) | 23 (95.8) | ||
| Sexually active after surgery | 0.100 | |||
| No | 357 (54.7) | 9 (37.5) | ||
| Yes | 296 (45.3) | 15 (62.5) | ||
| Surgeon | 6.7 (1) | 0.010 | ||
| A | 329 (44.2) | 17 (70.8) | ||
| B | 415 (55.8) | 7 (29.2) | ||
| Distance from home to surgery location | 1.7 (1) | 0.190 | ||
| Less than 50 miles | 459 (61.7) | 18 (75) | ||
| Greater than or equal to 50 miles | 285 (38.3) | 6 (25) | ||
| Patient type | 2.2 (1) | 0.140 | ||
| Endometrial cancer patient | 393 (52.8) | 9 (37.5) | ||
| Non-cancer patient | 351 (47.2) | 15 (62.5) | ||
Data are presented as n (%).
While controlling for other predictor variables, Table 3 illustrates that obese patients were 25.1% less likely than patients with a normal BMI to experience CD (p = 0.011). Additionally, CD was 2.8 times more likely to be reported when surgery was performed by surgeon A compared to surgeon B (p = 0.027). The remaining predictors in the model (cancer status, age, sexual activity after surgery, distance from home to the surgery location, and time interval from surgery to loss to follow-up) were not associated with CD.
Table 3.
Logistic regression model of patient characteristic predictors of CD.
| Predictors | Wald χ2 | P-value | Odds Ratio | 95% Confidence Interval |
|---|---|---|---|---|
| Cancer status | ||||
| Endometrial cancer patient | 0.80 | 0.37 | 0.6 | (0.20, 1.83) |
| Non-cancer patient | (Ref) | (Ref) | (Ref) | (Ref) |
| Age | ||||
| 50 years or older | 0.18 | 0.67 | 1.25 | (0.44, 3.57) |
| Younger than 50 years | (Ref) | (Ref) | (Ref) | (Ref) |
| BMI | ||||
| Obese (greater than 30) | 6.49 | 0.01 | 0.25 | (0.09, 0.73) |
| Overweight (25 to 29.9) | 1.62 | 0.20 | 0.50 | (0.17, 1.46) |
| Underweight (less than 18.5) | 0.10 | 0.76 | 1.43 | (0.15, 13.20) |
| Normal weight (18.5 to 24.9) | (Ref) | (Ref) | (Ref) | (Ref) |
| Surgeon | ||||
| Surgeon A | 4.87 | 0.03 | 2.80 | (1.12, 7.00) |
| Surgeon B | (Ref) | (Ref) | (Ref) | (Ref) |
| Sexually active after surgery | ||||
| No | 0.93 | 0.33 | 0.64 | (0.25, 1.59) |
| Yes | (Ref) | (Ref) | (Ref) | (Ref) |
| Distance | ||||
| Less than 50 miles | 0.67 | 0.41 | 1.50 | (0.57, 3.98) |
| 50 miles or more | (Ref) | (Ref) | (Ref) | (Ref) |
| Time interval from surgery to loss to follow-up | 3.43 | 0.06 | 1.00 | (0.99, 1.01) |
DISCUSSION
Our study reports a 2.5% incidence of CD after RLTH, with incidences of 1.9% in endometrial cancer patients, and 3.2% in non-cancer patients. We found no increased risk of CD among endometrial cancer patients compared to non-cancer patients. Additionally, the current study suggests that BMI and surgeon are the only variables associated with incidence of CD.
BMI was significantly associated with CD occurrence. Our study suggests obese patients were less likely to experience CD. These findings are supported by a prior study suggesting obesity to be a protective factor against CD after total laparoscopic hysterectomy or RLTH.18 Intercourse was the inciting event in six of the seven cases with identifiable precipitating events. Donnellan et al.18 hypothesized differences in positioning during intercourse may confer decreased force to the vaginal cuff among obese women compared to underweight or normal weight women. Therefore, women with a greater BMI may be at less of a risk for CD when resuming sexual activity after RLTH compared to women with a lesser BMI.
Both surgeons used the same suture material and nearly always used the same closure technique, suggesting there may have been a confounding factor. Some possible factors include patient demographics and sexual activity. Both surgeons used the same method of colpotomy for every surgery, so we were unable to determine if there was an association between CD and colpotomy technique.
Limitations
We excluded patients undergoing chemotherapy and/ or radiation within a year before or after RLTH to avoid the confounding negative effects that these treatments could potentially have on wound healing and tissue integrity.13 However, by excluding these patients, we consequently excluded those with high stage/grade endometrial cancer, which may be a limitation of this study. Additionally, our sample size did not meet the original power calculation requirements. Retrospectively, however, our power was still statistically acceptable.
Because many patients in this study reside outside a 50-mile radius of the surgery center, we anticipated a risk of potentially missed CD diagnoses due to follow-up examinations at outside facilities. We attempted to account for this by considering this variable in our data analyses. Despite this, CD may be underreported in the group living further from the surgery center.
The timing and frequency of resumed post-operative intercourse could not be determined or controlled for, making it difficult to analyze how intercourse affected cuff healing in this study. This limitation could also explain the difference in CD rates between surgeons A and B.
Finally, endometrial cancer patients had a longer follow-up period than non-cancer patients, likely due to the necessity of surveillance for relapse or spread of disease, which would be unnecessary in patients with benign indications for RLTH. This could affect the reported incidence of CD in non-cancer patients, as CD is less likely to be seen and diagnosed in patients with limited follow-up. Similarly, we anticipate that there are patients who have undetected CD or may eventually experience symptoms of CD in the future.
CONCLUSIONS
Our findings suggest that the identifiable incidence of CD after RLTH is 2.5%. Endometrial cancer patients were not at greater risk of experiencing CD compared to non-cancer patients. Surgeon differences and BMI were the only variables associated with the incidence of CD, with patients having a normal BMI being most likely to report experiencing this complication. Nearly one-third of patients with CD reported a precipitating event, with sexual intercourse being the most reported.
ACKNOWLEDGEMENTS
We would like to thank Elizabeth Ablah, Ph.D., MPH, CPH for her valuable expertise and guidance in the development of this manuscript. Her willingness to give her time so generously is greatly appreciated.
REFERENCES
- 1.Bouquet de Joliniere J, Librino A, Dubuisson JB, et al. Robotic surgery in gynecology. Front Surg. 2016;3:26. doi: 10.3389/fsurg.2016.00026. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Uccella S, Ghezzi F, Mariani A, et al. Vaginal cuff closure after minimally invasive hysterectomy: Our experience and systematic review of the literature. Am J Obstet Gynecol. 2011;205(2):119.e111–112. doi: 10.1016/j.ajog.2011.03.024. [DOI] [PubMed] [Google Scholar]
- 3.Kho RM, Akl MN, Cornella JL, Magtibay PM, Wechter ME, Magrina JF. Incidence and characteristics of patients with vaginal cuff dehiscence after robotic procedures. Obstet Gynecol. 2009;114(2 Pt 1):231–235. doi: 10.1097/AOG.0b013e3181af36e3. [DOI] [PubMed] [Google Scholar]
- 4.Nick AM, Lange J, Frumovitz M, et al. Rate of vaginal cuff separation following laparoscopic or robotic hysterectomy. Gynecol Oncol. 2011;120(1):47–51. doi: 10.1016/j.ygyno.2010.08.035. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Uccella S, Ceccaroni M, Cromi A, et al. Vaginal cuff dehiscence in a series of 12,398 hysterectomies: Effect of different types of colpotomy and vaginal closure. Obstet Gynecol. 2012;120(3):516–523. doi: 10.1097/AOG.0b013e318264f848. [DOI] [PubMed] [Google Scholar]
- 6.Fuchs Weizman N, Einarsson JI, Wang KC, Vitonis AF, Cohen SL. Vaginal cuff dehiscence: Risk factors and associated morbidities. JSLS. 2015;19(2) doi: 10.4293/JSLS.2013.00351. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Kashani S, Gallo T, Sargent A, Elsahwi K, Silasi DA, Azodi M. Vaginal cuff dehiscence in robotic-assisted total hysterectomy. JSLS. 2012;16(4):530–536. doi: 10.4293/108680812X13462882736817. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Cronin B, Sung VW, Matteson KA. Vaginal cuff dehiscence: Risk factors and management. Am J Obstet Gynecol. 2012;206(4):284–288. doi: 10.1016/j.ajog.2011.08.026. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Hur HC, Guido RS, Mansuria SM, Hacker MR, Sanfilippo JS, Lee TT. Incidence and patient characteristics of vaginal cuff dehiscence after different modes of hysterectomies. J Minim Invasive Gynecol. 2007;14(3):311–317. doi: 10.1016/j.jmig.2006.11.005. [DOI] [PubMed] [Google Scholar]
- 10.Iaco PD, Ceccaroni M, Alboni C, et al. Transvaginal evisceration after hysterectomy: Is vaginal cuff closure associated with a reduced risk? Eur J Obstet Gynecol Reprod Biol. 2006;125(1):134–138. doi: 10.1016/j.ejogrb.2005.08.009. [DOI] [PubMed] [Google Scholar]
- 11.Nezhat C, Kennedy Burns M, Wood M, Nezhat C, Nezhat A, Nezhat F. Vaginal cuff dehiscence and evisceration: A review. Obstet Gynecol. 2018;132(4):972–985. doi: 10.1097/AOG.0000000000002852. [DOI] [PubMed] [Google Scholar]
- 12.Ceccaroni M, Berretta R, Malzoni M, et al. Vaginal cuff dehiscence after hysterectomy: A multicenter retrospective study. Eur J Obstet Gynecol Reprod Biol. 2011;158(2):308–313. doi: 10.1016/j.ejogrb.2011.05.013. [DOI] [PubMed] [Google Scholar]
- 13.Payne WG, Naidu DK, Wheeler CK, et al. Wound healing in patients with cancer. Eplasty. 2008;8:e9. [PMC free article] [PubMed] [Google Scholar]
- 14.Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG. Research electronic data capture (REDCap)--a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform. 2009;42(2):377–381. doi: 10.1016/j.jbi.2008.08.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Harris PA, Taylor R, Minor BL, et al. The REDCap consortium: Building an international community of software partners. J Biomed Inform. 2019;95:103208. doi: 10.1016/j.jbi.2019.103208. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Park DA, Lee DH, Kim SW, Lee SH. Comparative safety and effectiveness of robot-assisted laparoscopic hysterectomy versus conventional laparoscopy and laparotomy for endometrial cancer: A systematic review and meta-analysis. Eur J Surg Oncol. 2016;42(9):1303–1314. doi: 10.1016/j.ejso.2016.06.400. [DOI] [PubMed] [Google Scholar]
- 17.Dauterive E, Morris GT. Incidence and characteristics of vaginal cuff dehiscence in robotic-assisted and traditional total laparoscopic hysterectomy. J Robot Surg. 2012;6(2):149–154. doi: 10.1007/s11701-011-0285-4. [DOI] [PubMed] [Google Scholar]
- 18.Donnellan NM, Mansuria S, Aguwa N, Lum D, Meyn L, Lee T. Obesity and older age as protective factors for vaginal cuff dehiscence following total hysterectomy. Gynecol Surg. 2015;12(2):89–93. doi: 10.1007/s10397-015-0882-8. [DOI] [PMC free article] [PubMed] [Google Scholar]