j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e8
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.JournalofSurgicalResearch.com
Computed tomographic measurements predict component separation in ventral hernia repair Laurel J. Blair, MD, Samuel W. Ross, MD, MPH, Ciara R. Huntington, MD, John D. Watkins, Tanushree Prasad, MA, Amy E. Lincourt, PhD, MBA, Vedra A. Augenstein, MD, FACS, and B. Todd Heniford, MD, FACS* Division of Gastrointestinal and Minimally Invasive Surgery, Department of Surgery, Carolinas Medical Center, Charlotte, North Carolina
article info
abstract
Article history:
Background: Preoperative imaging with computed tomography (CT) scans can be useful in
Received 30 January 2015
preoperative planning. We hypothesized that CT measurements of ventral hernia defect
Received in revised form
size and abdominal wall thickness (AWT) would correlate with postoperative complica-
11 June 2015
tions and need for complex abdominal wall reconstruction (AWR).
Accepted 12 June 2015
Materials and methods: Patients who underwent open ventral hernia repair and had pre-
Available online xxx
operative abdominal CT imagining were identified from an institutional hernia-specific surgery outcomes database at our tertiary referral hernia center. Grade III and IV hernias
Keywords:
and biologic mesh cases were excluded. CT measures of defect size and AWT were
Ventral hernia repair
analyzed and correlated to complications and the need for AWR techniques using uni-
Complications
variate, multivariate, and principal component (PC) analyses. PC1 and PC2 used five AWT
Abdominal wall reconstruction
measures, hernia defect width, and body mass index to create a new component variable.
Panniculectomy
Results: There were 151 open ventral hernia repairs included in the study. Preoperative
Component separation
findings included 37.7% male; age 55.3 12.5 years; body mass index (BMI) 33.3 7.8 kg/m2;
CT scan
60.3% were recurrent hernias with average defect width 8.5 5.0 cm and area
Measurement
178.3 214 cm2; AWT at umbilicus 3.5 1.8 cm; and AWT at pubis 7.0 3.2. Component
Computed tomography
separation was performed in 24.0% of patients and panniculectomy in 34.4%. Wound
Defect size
complications occurred in 13.3% patients, and 2.7% had hernia recurrence. Increasing
Recurrence
defect width, length, and area as well as select AWT measurements were associated with increased need for component separation, concomitant panniculectomy, and higher rates of wound and total complications (all P < 0.05). Using multivariate regression, PC1 was associated with wound complications (odds ratio [OR], 1.08; 95% confidence interval [CI], 1.01e1.16); PC2 (hernia defect width) was associated with the need for component separation (OR, 1.16; 95% CI, 1.03e1.30). Hernia recurrence was not predicted by AWT or defect size (OR, 1.00; 95%CI, 0.87e1.15). Conclusions: Preoperative CT measurements of hernia defects and AWT predict wound complications and the need for complex AWR techniques. Obtaining preoperative CT imaging should be a consideration in preoperative planning and may help with patient counseling. ª 2015 Elsevier Inc. All rights reserved.
Presented at the Academic Surgical Congress, Las Vegas, Nevada, February 2015. * Corresponding author. Division of Gastrointestinal and Minimally Invasive Surgery, Department of Surgery, Carolinas Medical Center, 1025 Morehead Medical Drive, Suite 300, Charlotte, NC 28204. Tel.: þ1 704 355 3168; fax: þ1 704 355 4117. E-mail address: [email protected] (B.T. Heniford). 0022-4804/$ e see front matter ª 2015 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jss.2015.06.033
2
1.
j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e8
Introduction
Computed tomography (CT) as a method for measuring adipose volume of the human body has been validated and standardized since 1983 [1e3]. Studies have shown that abdominal fat, as measured by abdominal wall thickness (AWT) on CT, has a direct correlation with medical conditions associated with obesity in abdominal surgery patients; furthermore, there is an association with longer operative time and complexity in abdominal surgery. In pancreatic surgery patients, it is an independent risk factor for poor postoperative outcomes [4e7]. In colorectal surgery, amount of visceral fat and AWT increases the risk of postoperative wound complications [8]. However, no studies have evaluated the effect of AWT on patients undergoing ventral hernia repair thus far. Obesity is a known factor for hernia recurrence and the development of wound complications after hernia repair [9e11]. Although body mass index (BMI) is a common preoperative clinical tool for evaluating obesity, it fails to account for fat distribution and quantity across different body areas [4]. Previous studies have shown that increased hernia defect width, length and especially area, as measured at the time of operation, have been associated with the need for AWR techniques and entail worse postoperative outcomes including quality of life [12]. Additionally, one study has evaluated preoperative CT scan use to predict fascial closure after open ventral hernia repair (OVHR) with component separation techniques (CSTs) and found CT measured defect width and its relative ratio to abdominal circumference to correlate highly with ability to close the abdomen [13]. Over the past decade, several strategies have been proposed for managing ventral hernias in obese patients, including staged OVHR preceded by bariatric surgery, various CSTs, panniculectomy plus OVHR, and OVHR with CST and panniculectomy [14e17]. However, no study has evaluated the correlation of preoperative CT measurements with the need for complex abdominal wall reconstruction (AWR) techniques or to clinical outcomes. The authors hypothesized that CT measurements of AWT as well as hernia defect size may be used to predict the need for AWR procedures, wound complications, and hernia recurrence.
Patients who had a grade III or IV hernia as defined by the Ventral Hernia Working Group classification or repair with biologic mesh were excluded from the study given the significant confounding variables with contaminated wounds, open wounds, complications following, and higher recurrence rates [18]. This study was approved by the Institutional Review Board of Carolinas Medical Center, and all patients signed consent for the study.
2.2.
CT measurements
Each patient’s preoperative CT scan variables were measured by two blinded, independent reviewers. The CT measurements were performed by one trained reviewer and then assessed for accuracy by random selection of one in every 10 patients for remeasurement by the second reviewer. The measurements had excellent concordance for interval quality assessment of accuracy in results. Five key measurements of AWT were obtained for each patient (Fig. 1): AWT at the umbilicus (AWT umbilicaldmeasured length from rectus abdominis fascia to the abdominal skin at the level of the umbilicus), retrorenal AWT (AWT retrorenaldmeasured length from the lateral external oblique fascia to the abdominal skin at the level of the left renal calyx), retrorenal fat pad thickness (retrorenaldmeasured length from the kidney to the transversalis fascia at the level of the left renal calyx), AWT at the level of the pubis (AWT pubisdlength from the pubis to the skin at the level of the pubic symphysis), and hip girdle thickness (hip girdledmeasured length from the anterior superior iliac spine to the lateral skin). In addition, hernia defect width and length were obtained at their maximum distance for each patient, and area was calculated using the area of an ellipse as previously described (Fig. 1) [13,19].
2.3.
Study design and statistical analysis
2.
Methods
The primary outcomes of interest were overall complications, the need for component separation (CS), concomitant panniculectomy, wound complications, and hernia recurrence. Descriptive statistics including means and standard deviations, or counts and percentages, were used to describe the study population on all variables. CT measurements were compared between each of the dichotomous outcome groups using Wilcoxon two-sample tests. Statistical significance was determined at P < 0.05, using a two-tailed alpha.
2.1.
Patient population
2.4.
A prospectively enrolled and collected, institutional review boardeapproved, hernia-specific surgical outcomes database was queried for all consenting patients undergoing OVHR at a single tertiary hernia referral center from 2008e2012. Over 50% of the patients in this database were referred from out of state or international sources. Demographics, operative details, patient outcomes, complications, hernia recurrence, and quality of life were considered. Patients with a preoperative CT scan of the abdomen and pelvis were included in the study.
Univariate analysis of principal components
Principal component analysis (PCA) was used to create composite “principal component” (PC) scores using the sixkey CT measurements and BMI. PC analysis is an advanced statistical method, which allows one to combine any number of observed variables with high intervariable correlation (redundancy) into one composite variable, which can then be used for further analysis. All PCs contain varying degrees of the following seven variables listed in Table 1: AWT umbilical, AWT retrorenal, retrorenal, AWT
3
j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e8
Fig. 1 e CT measurements. A: AWT measured at the level of the umbilicus as the paramedian vertical distance between the left rectus abdominis fascia and the skin (AW umbilical). B: AWT measured at the level of the left posterior renal capsule and the junction of the abdominal wall and paraspinal musculature at the level of the left renal vein (AW retrorenal). C: Retrorenal visceral fat measurement (retrorenal). D: AWT as measured at the level of the pubis (AW pubis). E: Measurement of subcutaneous tissue at the hip girdle (hip girdle) as defined by the distance between the iliac plate and the skin at the level of the posterior superior iliac spine. (Color version of figure is available online.)
pubis, hip girdle, defect width, and BMI. Table 1 lists the amount of representation of each variable in each PC. Where a given row and column intersect, the percent correlation between the two corresponding variables is listed: PC2 consists almost completely of hernia defect width with a percent correlation of 99.5%. PC1 consists of a mixture of all seven measurements, with BMI demonstrating a 90% correlation, and AWT pubis 28.5%, and hip girdle, a 26.5% correlation. Because seven variables were entered into the model, seven PC loadings were generated; however, it is not necessary to use all of these PCs as PC1 explains 68.6% of
the variance and PC2 explains 20.4% of the variance leading to 89% of the variance observed being explained using only the first two variables. Therefore, the first two variables were then used in multivariate analysis.
2.5.
PCA and multivariate analysis
Because of a high degree of multicollinearity (correlation) between the CT measurements at multiple sites and BMI, PCA without rotation was performed, incorporating the five AWT measurements, defect width, and BMI to obtain new
Table 1 e PC loadings (eigenvectors) and proportion of explained variance.
AWT umbilical AWT retrorenal Retrorenal AWT pubis Hip girdle Defect width BMI Proportion of variance explained by PC
PC1
PC2
PC3
PC4
PC5
PC6
PC7
0.133966 0.108001 0.02647 0.285977 0.265382 0.060194 0.902139 0.6864
0.002668 0.002473 0.021065 0.013674 0.042744 0.995928 0.075209 0.2042
0.131301 0.114236 0.211614 0.744267 0.600296 0.023998 0.100327 0.0470
0.07097 0.112587 0.045702 0.50729 0.746962 0.056518 0.402131 0.0328
0.87668 0.035815 0.395355 0.260889 0.025504 0.016051 0.069803 0.0148
0.378252 0.349318 0.84916 0.037793 0.092691 0.018149 0.058998 0.0102
0.219378 0.916132 0.273145 0.193036 0.013999 0.012124 0.018995 0.0046
Each PC is a combined variable to act as a surrogate for closely correlated variables. The loadings (also called eigenvectors) signify the magnitude of the association of the new PC variable with each original variable.
4
j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e8
uncorrelated variables. This technique creates a function of the “component” variables that can be used as a surrogate. The PCs which explained the maximal variance in the data were retained for multivariate analyses. The associations between the PCs and primary outcomes were evaluated using multivariate logistic regression, controlling for age, sex, diabetes, tobacco use, and number of recurrent hernias.
2.6.
Determining CT measurement thresholds
In secondary analyses, cutoff points were established for defect area and defect width above which CST was predictable. Using logistic regression, the sensitivity and specificity of various cutoff points for defect area and defect width were calculated. To validate these cutoff points, the values were used to determine the receiver operating characteristics (ROCs) of defect area and defect width for identifying patients with and without the need for CST. The performance of the ROC test was evaluated by estimating the C statistic, which represents the area under the curve (AUC). An AUC of 0.7 was considered to have good discriminatory ability. All data analysis was conducted using the Statistical Analysis Software (SAS), version 9.4 (SAS Institute, Inc, Cary, NC).
3.
Results
3.1. Overall characteristics of patients, CT measurements, and perioperative outcomes There were 151 OVHR patients who met the inclusion criteria and also consented to be studied. The characteristics of the study subjects are listed in Table 2. In total, mean patient age was 55.3 12.5 years with an average BMI of 33.3 7.8 kg/m2. The study included 57 male (37.7%) and 94 female patients. Of 151 patients, 95.4% had a prior laparotomy, and 60.3% presented with recurrent ventral hernias. On average, patients with recurrent ventral hernias had at least two prior ventral hernia repairs. Average CT measurements are summarized in Table 3. Preoperative CT scan was obtained at a median of 84 d (minimum 1 d, maximum 363 d) before OVHR. The mean hernia defect width was 8.5 5.0 cm, length 7.3 5.6 cm, and area were 178.3 214 cm2. Average AWT at the umbilicus was 3.5 1.8 cm, mean retrorenal AWT was 2.6 1.3 cm, mean retrorenal thickness was 1.8 1.2 cm, mean AWT at the pubis was 7.0 3.2 cm, and mean hip girdle measurement was 7.6 3.0 cm. The mean operative time was 171 98 min, estimated blood loss was 116 127 mL, and mesh size measured 700.8 477 cm2. CS was performed in 24% of patients and a panniculectomy in 34.4%. The rate of wound and total complications were 13.25% and 39.1%, respectively. There were four recurrences (2.7%); three of which were associated with splitting lightweight mesh; the fourth recurrence and one of the lightweight recurrences was associated with infection. Mortality was 0.66%. Mean length of follow-up was 15 14.5 mo (Table 2).
Table 2 e Patient characteristics. Mean SD/n (%) Age (y) BMI (kg/m2) Gender Male Female Race Caucasian Other Previous abdominal surgery Ventral hernia repair No. of previous VHR Follow-up (mo) Operative characteristics Operative time (min) EBL (mL) Defect size (cm2) Mesh size (cm2) CS Panniculectomy Preperitoneal mesh placement Length of stay (d) Complications Cellulitis Wound infection Mesh infection Wound vac placement DVT Readmission Hernia recurrence Total wound complications Total complications Mortality
55.4 12.6 33.3 7.8 57 (37.7) 94 (63.3) 123 (81.5) 28 (18.5) 144 (95.4) 91 (60.3) 2.29 1.8 14.9 14.5 171 98 116.0 127 178.0 214 700.8 477 35 (23.97) 52 (34.4) 95 (63.8) 5.6 5.0 29 20 1 2 1 7 4 20 59 1
(19.2) (13.3) (0.66) (3.0) (0.66) (4.6) (2.7) (13.25) (39.1) (0.66)
SD ¼ standard deviation; EBL ¼ estimated blood loss; DVT ¼ deep venous thrombosis. Patient demographics, operative characteristics, and complications for the study population.
3.2. Univariate analysis: CT measurements with primary outcomes Associations between individual CT measurements and primary outcomes are summarized in Table 4. Larger hernia defect width, length, and area were associated with performance of a CS (all P < 0.001). However, CS was not associated
Table 3 e CT measurement characteristics. Mean SD AW umbilical (cm) AW retrorenal (cm) Retrorenal (cm) AW pubis (cm) Hip girdle (cm) Defect width (cm) Defect length (cm) Defect area (cm2) SD ¼ standard deviation. Mean CT measurements are presented here.
3.5 2.6 1.8 7.0 7.6 8.5 7.3 178.3
1.8 1.3 1.2 3.2 3.0 5.0 5.6 214
5
j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e8
Table 4 e Association of CT measurements with surgical techniques and outcomes. Technique or outcome CS Yes No P value Panniculectomy Yes No P value Wound omplications Yes No P value Total complications Yes No P value Recurrence Yes No P value
AWT umbilical (cm)
AWT retrorenal (cm)
Retrorenal (cm)
AWT pubis (cm)
Hip girdle (cm)
Defect width (cm)
Defect length (cm)
Defect area (cm)
3.4 1.9 3.6 1.6 NS (0.46)
2.5 1.3 2.6 1.3 NS (0.80)
1.7 1.1 1.8 1.3 NS (0.92)
7.2 3.5 7.5 3.8 NS (0.64)
7.4 2.6 7.7 3.2 NS (0.90)
11.5 ± 5.2 7.6 ± 4.8 0.0002*
11.1 ± 5.9 6.0 ± 5.4 0.0008*
115 ± 93 49.4 ± 85 0.0006*
4.1 ± 2.0 3.3 ± 1.5 0.04*
3.1 ± 1.4 2.3 ± 1.1 0.0006*
2.0 1.4 1.7 1.2 NS(0.14)
9.3 ± 4.1 6.3 ± 2.8 0.0001*
8.7 ± 3.1 6.9 ± 2.7 0.003*
10.9 ± 5.4 7.2 ± 4.6 0.0001*
9.6 ± 6.6 6.4 ± 5.2 0.03*
109.5 ± 114 47.8 ± 70 0.007*
4.3 ± 2.2 3.2 ± 1.3 0.01*
2.9 ± 1.3 2.4 ± 1.2 0.007*
1.9 1.3 1.7 1.3 NS (0.19)
8.6 ± 3.3 6.9 ± 3.7 0.001*
8.1 3.2 7.3 2.9 NS (0.19)
10.0 ± 5.3 7.7 ± 5.0 0.01*
10.9 ± 6.1 6.3 ± 5.4 0.003*
116.4 ± 116 50.7 ± 73.5 0.005*
3.8 1.8 3.4 1.6 NS (0.25)
2.8 1.4 2.5 1.2 NS (0.20)
1.98 1.3 1.6 1.2 NS (0.07)
8.2 ± 3.8 6.8 ± 3.5 0.01*
7.99 3.0 7.3 2.96 NS (0.19)
10.2 ± 5.1 7.4 ± 4.9 0.0009*
9.5 ± 6.2 5.9 ± 5.2 0.005*
95.4 ± 104 46.5 ± 72.4 0.007*
1.7 0.6 3.6 1.7 NS (0.06)
2.5 0.9 2.6 1.3 NS (0.89)
1.7 0.7 1.8 1.3 NS (0.76)
7.7 1.9 7.3 3.6 NS (0.45)
9.2 2.9 7.5 2.9 NS (0.23)
10.8 4.8 8.4 5.1 NS (0.34)
4.5 7.5 5.9 NS (0.75)
51.2 66.8 91.1 NS (0.80)
NS ¼ not significant. Univariate association of CT measurements to predict complications or advanced AWR techniques in OVHR. Values set in bold are statistically significantly different (P < 0.05). * Represents significant P value 0.05 for all). Increased AWT measurements excluding retrorenal and hip girdle correlated with performance of a panniculectomy, and use of this technique was also associated with increased hernia defect width, length, and area (P < 0.05 for all). AWT measures (excluding retrorenal and hip girdle) and defect width, length, and area were associated with increased wound complications (all P < 0.05). Total complications were associated with increased AWT pubis and increased defect width, defect length, and defect area (P < 0.01). In this group of patients, recurrence was not associated with any of the CT measurements or hernia defect sizes (P > 0.05 for all), although increased AWT umbilical CT measurement approached statistical significance (P ¼ 0.06). The magnitude of association for the PCs, or “loadings” is summarized in Table 1. PC1 has high loadings on all variables except defect width, and PC1 explained almost 68% of the total variability among the variables. PC2 had the highest loading on hernia defect width and explained approximately 22% of the remaining variability among the variables. PC1 and PC2 jointly explained approximately 89% of the total variability among the variables. In univariate analyses, the composite variable PC1, a surrogate for CT measurements and correlated variables as described in Table 1, was associated with panniculectomy, wound complications, and total complications (all P < 0.05), but was not associated with CS. PC2 was not associated with wound complications, but did correlate with total complications, performance of CS and panniculectomy in univariate analyses (all P < 0.05).
3.3.
Multivariate regression using PCs
While controlling for confounding factors, the effects of calculated PC1 and PC2 on the primary outcomes are summarized in Table 5. Higher magnitude of PC1 was independently associated with increased wound complications (odds ratio [OR], 1.08; 95% confidence interval [CI], 1.01e1.16; P ¼ 0.024). None of the other outcomes including CS, panniculectomy, total complications, or hernia recurrence, had any significant association with PC1. PC2 (hernia defect width) was independently associated with CS on multivariate analysis (OR, 1.16; 95% CI, 1.03e1.30; P ¼ 0.013). Our multivariate analysis (age, sex, diabetes, tobacco use, and recurrent hernias) was selected on the basis of previously established and well-described clinical risk factors in an effort to control for confounding factors. Similar multivariate analyses have been published by our institution and others in the past, and these factors have been identified as attributing to confounding results between differing cohorts. This method of multivariate analysis has been used by other groups including Krpata et al. [20] when they used obesity, diabetes, chronic obstructive lung disease, smoking, and immunosuppression as covariates. In addition, a recent publication from our institution by Colavita et al. [21] used a similar multivariate analysis.
3.4. Clinical cut points for defect area and width predict CS Multiple cutoff values for hernia defect area and defect width to predict CS were examined. Table 6 summarizes the
6
j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e8
Table 5 e Multivariate logistic regression model: predictive factors for wound complications and CSTs. OR Outcome: wound complications PC1 1.080 PC2 1.038 Age 0.984 Male sex 0.735 Diabetes 0.995 Tobacco 1.828 Number of prior hernias 1.65 Outcome: need for CS PC1 0.960 PC2 1.159 Age 0.999 Male sex 0.477 Diabetes 0.998 Tobacco 1.85 Number of prior hernias 0.896 Outcome: recurrence PC1 1.00 PC2 1.00 Age 1.00 Male sex 1.00 Diabetes 1.00 Tobacco 1.00 Number of prior hernias 1.00
95% CI
P value
1.01e1.160 0.933e1.155 0.94e1.16 0.24e2.23 0.29e3.41 0.48e6.92 1.08e2.53
0.024 NS (0.490) NS (0.462) NS (0.587) NS (0.994) NS (0.375) 0.021
0.89e1.04 1.03e1.30 0.95e1.05 0.14e1.61 0.28e3.53 0.49e6.95 0.58e1.40
NS (0.303) 0.013 NS (0.959) NS (0.232) NS (0.998) NS (0.362) NS (0.626)
0.77e1.29 0.66e1.51 0.84e1.19 0.014e71.78 0.008e118.43 0.006e163.39 0.23e4.42
NS (1.00) NS (1.00) NS (1.00) NS (1.00) NS (1.00) NS (1.00) NS (1.00)
NS ¼ not significant. Multivariate regression revealed that PC1 and number of prior hernias significantly increase risk for wound complications in this population. Number of prior hernias, CS, and PC2 were evaluated as continuous variables. Values set in bold are statistically significantly different (P < 0.05).
sensitivity, specificity, and AUC for various cutoff values for defect area and defect width. The ROC curves are displayed in Figure 2. Defect widths >8.3 cm typically required CS (sensitivity 74%, specificity 63%, AUC ¼ 0.72). In addition, hernias >164 cm2 were more likely to be repaired using a CS (sensitivity 89%, specificity 73%, AUC ¼ 0.83).
4.
Discussion
In this study, the preoperative CT measurements of 151 ventral hernia patients were analyzed, and AWT measurements and hernia defect size correlated with postoperative complications and utilization of AWR techniques. Although recurrence did not correlate with any CT measurements of the abdominal wall or hernia defect size, increasing AWT in the umbilical region, at the level of the retrorenal measurement, and at the pubic symphysis was associated with increased risk of wound complications. Increasing hernia defect width, length, and area were associated with wound complications, total complications, CS, and panniculectomy. The retrorenal measurement was not strongly correlated with surgical outcomes in this study. On multivariate analysis, the composite variables PC1 and PC2 were associated with wound complications and CS, respectively. In addition, we used advanced statistical methods to attempt to establish a cutoff point after which patients would be more likely to require CS. We found
Table 6 e Predictors of CS. Predictor
Cutoff value
Sensitivity (%)
Specificity (%)
AUC
Defect area
150 cm2 157 cm2 164 cm2 170 cm2 183 cm2 7.2 cm 7.8 cm 8.3 cm 8.8 cm 9.4 cm
92.9 92.9 89.3 85.7 75 80 80 74.3 71.4 60
69.3 70.5 72.7 73.9 76.1 59 61 62.9 66.7 70.5
0.82 0.83 0.83 0.82 0.81 0.70 0.71 0.72 0.70 0.66
Defect width
Cutoff values and diagnostics of defect area and defect width in correlation with CS. Bolddcutoff values of defect area and defect width with best tradeoff between sensitivity and specificity and high AUC.
that defect areas >164 cm2 or width >8.3 cm were associated with performance of a CS. This finding would be useful for surgeons in preoperative planning and counseling. CS can be complex, requires expertise in AWR, and it also carries associated unique risks and benefits for the patient including increased risk of wound infection, dehiscence, ischemia, and hernia recurrence [22]. Our study references and is somewhat similar to the Franklin et al. [13] article, published in September 2013 in Annals of Plastic Surgery; however, the previous study is a small series of only 54 patients. The authors write: “No cutoff value was able to be calculated because of small sample sizes in each group.” Our larger cohort of 151 patients allowed for this calculation and further predicted cutoff point of defect widths >8.3 cm and hernia area >164 cm2 needing CS. This study adds to the body of literature and provides objective evidence of a cutoff point that is clinically applicable. Although we have not yet applied these cutoff points to an external data set for validation, we plan to do that in the near future to verify the conclusions. Our findings in this ventral hernia study population are in line with prior findings published in the colorectal and urology literature, which found that wound and total complications are increased with increasing AWT [5,7,8]. Patients with large ventral hernias (classified as hernia area >100 cm2) are at increased risk of seroma formation and worse long-term quality of life outcomes according to recently published data [12]. Patients in this cohort had average defect widths of 8.5 5 cm and defect areas of 178 214 cm2 and are at-risk patients. Another recent study evaluating the likelihood of abdominal closure after CS using CT measurements stated that there is a paucity of data concerning hernia-specific variables related to recurrence, although defect area was a factor increasing recurrence rates [13]. The advanced statistical techniques used here are useful for application to the ventral hernia population and can potentially impact current preoperative planning practices. Preoperative use of CT scans can help predict outcomes and type of procedures needed for improving results. Many patients have CT scans before surgery, especially if they have a recurrent ventral hernia, to examine the hernia contents, presence and placement of previous mesh, and to determine
j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e8
7
Fig. 2 e ROC curves of defect area and defect width for predicting CS. Measurements are recorded in centimeters. (Color version of figure is available online.)
the size of the defect. If surgeons are familiar with the data presented herein, measurements easily obtained from CT scans can predict wound complications and the need for components separation. Counseling patients preoperatively is important and identifying tools to do so can be very helpful. Despite potential applications for this data, several limitations exist. This is a single-institution study conducted at a highly specialized hernia center where fellowship-trained surgeons practice advanced AWR very regularly. Advanced experience and training in AWR may underestimate complications that might be seen in the general population. Grade III and IV hernia patients were excluded from this study, and therefore, caution should be employed before applying the results of this study, especially that of wound complications, to that population. In addition, biologic mesh cases were excluded, and we cannot extrapolate complications or techniques for the patients whose hernias are repaired with biologic mesh. Although hernia recurrence did not correlate to any AWT measurements or hernia defect size, the number of hernia recurrences was very low (n ¼ 4, 2.7%); Therefore, a type 2 error could exist where an actual clinical difference exists but the study is underpowered to detect it. Three of the recurrences were associated with splitting lightweight mesh; the fourth recurrence and one of the lightweight mesh recurrences had infectious complications. Three of the four had 2.6 average failed prior ventral hernia repairs, only one of the patients underwent CS. Preoperative planning to optimize outcomes, predict and minimize risks and cost, and allocate resources appropriately is important for improving patient quality of life and overall efficiency of the health care system. This study uses preoperative CT measurements to predict the use of advanced complex AWR for patients with ventral hernias. CT scanning preoperatively helps predict which patients may be at increased risk for wound and total complications. PC1 was associated very strongly with BMI in this study (90.2% correlation). Further research will be necessary before this composite variable can be used in the clinical setting; however, at
this time, it is clear that increased BMI (obesity) leads to an increased risk of wound complications. This has been demonstrated by other groups as well, such as Krpata et al. [20], who reported that patients undergoing open repair for grade II hernias had a 16% increased risk of developing a surgical site occurrence when they had comorbidities such as obesity, chronic obstructive lung disease, diabetes, and smoking history. This ability to identify at-risk patients has the potential to identify and modify patients’ risk factors preoperatively and impact operative techniques and outcomes. Larger multicenter studies will better define the impact of CT measurements for patients undergoing ventral hernia repair.
5.
Conclusions
Preoperative CT measurements of ventral hernia defects and AWT predict postoperative complications and the use of complex AWR techniques. In addition, increasing AWT correlates with wound complications and increasing defect width correlates with likelihood of performance of a CS. Analyzing CT measurements may play an important role in the future as these may be used to calculate risk of postoperative complications at the time of the preoperative clinic visit and predict the operative techniques likely required.
Acknowledgment Authors’ contributions: L.J.B. contributed significantly to the writing of the manuscript; S.W.R. and C.R.H. contributed to the writing and study design; J.D.W. contributed to the study design and data gathering; T.P. contributed to the statistical analysis and data gathering and article writing; A.E.L. contributed to the study design; V.A.A. was involved in the study design implementation, data collection, article writing,
8
j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e8
and editing; and B.T.H was involved in the study design, article writing, and editing.
Disclosure B.T.H., V.A.A., and A.E.L. have previously been awarded surgical research and education grants from W.L. Gore and Associates, Ethicon, Novadaq, Bard/Davol, and LifeCell Inc. All other authors have no potential conflicts or disclosures relevant to this work.
references
[1] Tokunaga K, Matsuzawa Y, Ishikawa K, Tarui S. A novel technique for the determination of body fat by computed tomography. Int J Obes 1983;7:437. [2] Sampaio LR, Simoes EJ, Assis AM, Ramos LR. Validity and reliability of the sagittal abdominal diameter as a predictor of visceral abdominal fat. Arq Bras Endocrinol Metabol 2007;51: 980. [3] Yoshizumi T, Nakamura T, Yamane M, et al. Abdominal fat: standardized technique for measurement at CT. Radiology 1999;211:283. [4] Balentine CJ, Marshall C, Robinson C, et al. Validating quantitative obesity measurements in colorectal cancer patients. J Surg Res 2010;164:18. [5] Hagiwara M, Miyajima A, Hasegawa M, et al. Visceral obesity is a strong predictor of perioperative outcome in patients undergoing laparoscopic radical nephrectomy. BJU Int 2012; 110:E980. [6] Anderson KM, Lindler TU, Lamberton GR, Baron PW, Ojogho OK, Baldwin DD. Laparoscopic donor nephrectomy: effect of perirenal fat upon donor operative time. J Endourol 2008;22:2269. [7] Shimizu A, Tani M, Kawai M, et al. Influence of visceral obesity for postoperative pulmonary complications after pancreaticoduodenectomy. J Gastrointest Surg 2011;15:1401. [8] Fujii T, Tsutsumi S, Matsumoto A, et al. Thickness of subcutaneous fat as a strong risk factor for wound infections in elective colorectal surgery: impact of prediction using preoperative CT. Dig Surg 2010;27:331. [9] Heniford BT, Park A, Ramshaw BJ, Voeller G. Laparoscopic repair of ventral hernias: nine years’ experience with 850 consecutive hernias. Ann Surg 2003;238:391. discussion 399e400.
[10] Hughes KC, Weider L, Fischer J, et al. Ventral hernia repair with simultaneous panniculectomy. Am Surg 1996;62:678. [11] Fekkes JF, Velanovich V. Amelioration of the effects of obesity on short-term postoperative complications of laparoscopic and open ventral hernia repair. Surg Laparosc Endosc Percutan Tech 2015;25:151. [12] Wormer BA, Walters AL, Bradley JF 3rd, et al. Does ventral hernia defect length, width, or area predict postoperative quality of life? Answers from a prospective, international study. J Surg Res 2013;184:169. [13] Franklin BR, Patel KM, Nahabedian MY, Baldassari LE, Cohen EI, Bhanot P. Predicting abdominal closure after component separation for complex ventral hernias maximizing the use of preoperative CT. Ann Plast Surg 2013;71:261. [14] Newcomb WL, Polhill JL, Chen AY, et al. Staged hernia repair preceded by gastric bypass for the treatment of morbidly obese patients with complex ventral hernias. Hernia 2008;12: 465. [15] Reid RR, Dumanian GA. Panniculectomy and the separationof-parts hernia repair: a solution for the large infraumbilical hernia in the obese patient. Plast Reconstr Surg 2005;116: 1006. [16] Borud LJ, Grunwaldt L, Janz B, Mun E, Slavin SA. Components separation combined with abdominal wall plication for repair of large abdominal wall hernias following bariatric surgery. Plast Reconstr Surg 2007;119:1792. [17] Chang EI, Foster RD, Hansen SL, Jazayeri L, Patti MG. Autologous tissue reconstruction of ventral hernias in morbidly obese patients. Arch Surg 2007;142:746. discussion 749e51. [18] Ventral hernia working group. Incisional ventral hernias: review of the literature and recommendations regarding the grading and technique of repair. Surgery 2010;148:544. [19] Muysoms FE, Miserez M, Berrevoet F, et al. Classification of primary and incisional abdominal wall hernias. Hernia 2009; 13:407. [20] Krpata DM, Blatnik JA, Novitsky YW, Rosen MJ. Evaluation of high-risk, comorbid patients undergoing open ventral hernia repair with synthetic mesh. Surgery 2013;153:120. http://dx. doi.org/10.1016/j.surg.2012.06.003. Epub 2012 Aug 3. [21] Colavita PD, Tsirline VB, Belyansky I, et al. Prospective, longterm comparison of quality of life in laparoscopic versus open ventral hernia repair. Ann Surg 2012;256:714. [22] Jensen KK, Henriksen NA, Jorgensen LN. Endoscopic component separation for ventral hernia causes fewer wound complications compared to open components separation: a systematic review and meta-analysis. Surg Endosc 2014;28:3046.
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.JournalofSurgicalResearch.com
Computed tomographic measurements predict component separation in ventral hernia repair Laurel J. Blair, MD, Samuel W. Ross, MD, MPH, Ciara R. Huntington, MD, John D. Watkins, Tanushree Prasad, MA, Amy E. Lincourt, PhD, MBA, Vedra A. Augenstein, MD, FACS, and B. Todd Heniford, MD, FACS* Division of Gastrointestinal and Minimally Invasive Surgery, Department of Surgery, Carolinas Medical Center, Charlotte, North Carolina
article info
abstract
Article history:
Background: Preoperative imaging with computed tomography (CT) scans can be useful in
Received 30 January 2015
preoperative planning. We hypothesized that CT measurements of ventral hernia defect
Received in revised form
size and abdominal wall thickness (AWT) would correlate with postoperative complica-
11 June 2015
tions and need for complex abdominal wall reconstruction (AWR).
Accepted 12 June 2015
Materials and methods: Patients who underwent open ventral hernia repair and had pre-
Available online xxx
operative abdominal CT imagining were identified from an institutional hernia-specific surgery outcomes database at our tertiary referral hernia center. Grade III and IV hernias
Keywords:
and biologic mesh cases were excluded. CT measures of defect size and AWT were
Ventral hernia repair
analyzed and correlated to complications and the need for AWR techniques using uni-
Complications
variate, multivariate, and principal component (PC) analyses. PC1 and PC2 used five AWT
Abdominal wall reconstruction
measures, hernia defect width, and body mass index to create a new component variable.
Panniculectomy
Results: There were 151 open ventral hernia repairs included in the study. Preoperative
Component separation
findings included 37.7% male; age 55.3 12.5 years; body mass index (BMI) 33.3 7.8 kg/m2;
CT scan
60.3% were recurrent hernias with average defect width 8.5 5.0 cm and area
Measurement
178.3 214 cm2; AWT at umbilicus 3.5 1.8 cm; and AWT at pubis 7.0 3.2. Component
Computed tomography
separation was performed in 24.0% of patients and panniculectomy in 34.4%. Wound
Defect size
complications occurred in 13.3% patients, and 2.7% had hernia recurrence. Increasing
Recurrence
defect width, length, and area as well as select AWT measurements were associated with increased need for component separation, concomitant panniculectomy, and higher rates of wound and total complications (all P < 0.05). Using multivariate regression, PC1 was associated with wound complications (odds ratio [OR], 1.08; 95% confidence interval [CI], 1.01e1.16); PC2 (hernia defect width) was associated with the need for component separation (OR, 1.16; 95% CI, 1.03e1.30). Hernia recurrence was not predicted by AWT or defect size (OR, 1.00; 95%CI, 0.87e1.15). Conclusions: Preoperative CT measurements of hernia defects and AWT predict wound complications and the need for complex AWR techniques. Obtaining preoperative CT imaging should be a consideration in preoperative planning and may help with patient counseling. ª 2015 Elsevier Inc. All rights reserved.
Presented at the Academic Surgical Congress, Las Vegas, Nevada, February 2015. * Corresponding author. Division of Gastrointestinal and Minimally Invasive Surgery, Department of Surgery, Carolinas Medical Center, 1025 Morehead Medical Drive, Suite 300, Charlotte, NC 28204. Tel.: þ1 704 355 3168; fax: þ1 704 355 4117. E-mail address: [email protected] (B.T. Heniford). 0022-4804/$ e see front matter ª 2015 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jss.2015.06.033
2
1.
j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e8
Introduction
Computed tomography (CT) as a method for measuring adipose volume of the human body has been validated and standardized since 1983 [1e3]. Studies have shown that abdominal fat, as measured by abdominal wall thickness (AWT) on CT, has a direct correlation with medical conditions associated with obesity in abdominal surgery patients; furthermore, there is an association with longer operative time and complexity in abdominal surgery. In pancreatic surgery patients, it is an independent risk factor for poor postoperative outcomes [4e7]. In colorectal surgery, amount of visceral fat and AWT increases the risk of postoperative wound complications [8]. However, no studies have evaluated the effect of AWT on patients undergoing ventral hernia repair thus far. Obesity is a known factor for hernia recurrence and the development of wound complications after hernia repair [9e11]. Although body mass index (BMI) is a common preoperative clinical tool for evaluating obesity, it fails to account for fat distribution and quantity across different body areas [4]. Previous studies have shown that increased hernia defect width, length and especially area, as measured at the time of operation, have been associated with the need for AWR techniques and entail worse postoperative outcomes including quality of life [12]. Additionally, one study has evaluated preoperative CT scan use to predict fascial closure after open ventral hernia repair (OVHR) with component separation techniques (CSTs) and found CT measured defect width and its relative ratio to abdominal circumference to correlate highly with ability to close the abdomen [13]. Over the past decade, several strategies have been proposed for managing ventral hernias in obese patients, including staged OVHR preceded by bariatric surgery, various CSTs, panniculectomy plus OVHR, and OVHR with CST and panniculectomy [14e17]. However, no study has evaluated the correlation of preoperative CT measurements with the need for complex abdominal wall reconstruction (AWR) techniques or to clinical outcomes. The authors hypothesized that CT measurements of AWT as well as hernia defect size may be used to predict the need for AWR procedures, wound complications, and hernia recurrence.
Patients who had a grade III or IV hernia as defined by the Ventral Hernia Working Group classification or repair with biologic mesh were excluded from the study given the significant confounding variables with contaminated wounds, open wounds, complications following, and higher recurrence rates [18]. This study was approved by the Institutional Review Board of Carolinas Medical Center, and all patients signed consent for the study.
2.2.
CT measurements
Each patient’s preoperative CT scan variables were measured by two blinded, independent reviewers. The CT measurements were performed by one trained reviewer and then assessed for accuracy by random selection of one in every 10 patients for remeasurement by the second reviewer. The measurements had excellent concordance for interval quality assessment of accuracy in results. Five key measurements of AWT were obtained for each patient (Fig. 1): AWT at the umbilicus (AWT umbilicaldmeasured length from rectus abdominis fascia to the abdominal skin at the level of the umbilicus), retrorenal AWT (AWT retrorenaldmeasured length from the lateral external oblique fascia to the abdominal skin at the level of the left renal calyx), retrorenal fat pad thickness (retrorenaldmeasured length from the kidney to the transversalis fascia at the level of the left renal calyx), AWT at the level of the pubis (AWT pubisdlength from the pubis to the skin at the level of the pubic symphysis), and hip girdle thickness (hip girdledmeasured length from the anterior superior iliac spine to the lateral skin). In addition, hernia defect width and length were obtained at their maximum distance for each patient, and area was calculated using the area of an ellipse as previously described (Fig. 1) [13,19].
2.3.
Study design and statistical analysis
2.
Methods
The primary outcomes of interest were overall complications, the need for component separation (CS), concomitant panniculectomy, wound complications, and hernia recurrence. Descriptive statistics including means and standard deviations, or counts and percentages, were used to describe the study population on all variables. CT measurements were compared between each of the dichotomous outcome groups using Wilcoxon two-sample tests. Statistical significance was determined at P < 0.05, using a two-tailed alpha.
2.1.
Patient population
2.4.
A prospectively enrolled and collected, institutional review boardeapproved, hernia-specific surgical outcomes database was queried for all consenting patients undergoing OVHR at a single tertiary hernia referral center from 2008e2012. Over 50% of the patients in this database were referred from out of state or international sources. Demographics, operative details, patient outcomes, complications, hernia recurrence, and quality of life were considered. Patients with a preoperative CT scan of the abdomen and pelvis were included in the study.
Univariate analysis of principal components
Principal component analysis (PCA) was used to create composite “principal component” (PC) scores using the sixkey CT measurements and BMI. PC analysis is an advanced statistical method, which allows one to combine any number of observed variables with high intervariable correlation (redundancy) into one composite variable, which can then be used for further analysis. All PCs contain varying degrees of the following seven variables listed in Table 1: AWT umbilical, AWT retrorenal, retrorenal, AWT
3
j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e8
Fig. 1 e CT measurements. A: AWT measured at the level of the umbilicus as the paramedian vertical distance between the left rectus abdominis fascia and the skin (AW umbilical). B: AWT measured at the level of the left posterior renal capsule and the junction of the abdominal wall and paraspinal musculature at the level of the left renal vein (AW retrorenal). C: Retrorenal visceral fat measurement (retrorenal). D: AWT as measured at the level of the pubis (AW pubis). E: Measurement of subcutaneous tissue at the hip girdle (hip girdle) as defined by the distance between the iliac plate and the skin at the level of the posterior superior iliac spine. (Color version of figure is available online.)
pubis, hip girdle, defect width, and BMI. Table 1 lists the amount of representation of each variable in each PC. Where a given row and column intersect, the percent correlation between the two corresponding variables is listed: PC2 consists almost completely of hernia defect width with a percent correlation of 99.5%. PC1 consists of a mixture of all seven measurements, with BMI demonstrating a 90% correlation, and AWT pubis 28.5%, and hip girdle, a 26.5% correlation. Because seven variables were entered into the model, seven PC loadings were generated; however, it is not necessary to use all of these PCs as PC1 explains 68.6% of
the variance and PC2 explains 20.4% of the variance leading to 89% of the variance observed being explained using only the first two variables. Therefore, the first two variables were then used in multivariate analysis.
2.5.
PCA and multivariate analysis
Because of a high degree of multicollinearity (correlation) between the CT measurements at multiple sites and BMI, PCA without rotation was performed, incorporating the five AWT measurements, defect width, and BMI to obtain new
Table 1 e PC loadings (eigenvectors) and proportion of explained variance.
AWT umbilical AWT retrorenal Retrorenal AWT pubis Hip girdle Defect width BMI Proportion of variance explained by PC
PC1
PC2
PC3
PC4
PC5
PC6
PC7
0.133966 0.108001 0.02647 0.285977 0.265382 0.060194 0.902139 0.6864
0.002668 0.002473 0.021065 0.013674 0.042744 0.995928 0.075209 0.2042
0.131301 0.114236 0.211614 0.744267 0.600296 0.023998 0.100327 0.0470
0.07097 0.112587 0.045702 0.50729 0.746962 0.056518 0.402131 0.0328
0.87668 0.035815 0.395355 0.260889 0.025504 0.016051 0.069803 0.0148
0.378252 0.349318 0.84916 0.037793 0.092691 0.018149 0.058998 0.0102
0.219378 0.916132 0.273145 0.193036 0.013999 0.012124 0.018995 0.0046
Each PC is a combined variable to act as a surrogate for closely correlated variables. The loadings (also called eigenvectors) signify the magnitude of the association of the new PC variable with each original variable.
4
j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e8
uncorrelated variables. This technique creates a function of the “component” variables that can be used as a surrogate. The PCs which explained the maximal variance in the data were retained for multivariate analyses. The associations between the PCs and primary outcomes were evaluated using multivariate logistic regression, controlling for age, sex, diabetes, tobacco use, and number of recurrent hernias.
2.6.
Determining CT measurement thresholds
In secondary analyses, cutoff points were established for defect area and defect width above which CST was predictable. Using logistic regression, the sensitivity and specificity of various cutoff points for defect area and defect width were calculated. To validate these cutoff points, the values were used to determine the receiver operating characteristics (ROCs) of defect area and defect width for identifying patients with and without the need for CST. The performance of the ROC test was evaluated by estimating the C statistic, which represents the area under the curve (AUC). An AUC of 0.7 was considered to have good discriminatory ability. All data analysis was conducted using the Statistical Analysis Software (SAS), version 9.4 (SAS Institute, Inc, Cary, NC).
3.
Results
3.1. Overall characteristics of patients, CT measurements, and perioperative outcomes There were 151 OVHR patients who met the inclusion criteria and also consented to be studied. The characteristics of the study subjects are listed in Table 2. In total, mean patient age was 55.3 12.5 years with an average BMI of 33.3 7.8 kg/m2. The study included 57 male (37.7%) and 94 female patients. Of 151 patients, 95.4% had a prior laparotomy, and 60.3% presented with recurrent ventral hernias. On average, patients with recurrent ventral hernias had at least two prior ventral hernia repairs. Average CT measurements are summarized in Table 3. Preoperative CT scan was obtained at a median of 84 d (minimum 1 d, maximum 363 d) before OVHR. The mean hernia defect width was 8.5 5.0 cm, length 7.3 5.6 cm, and area were 178.3 214 cm2. Average AWT at the umbilicus was 3.5 1.8 cm, mean retrorenal AWT was 2.6 1.3 cm, mean retrorenal thickness was 1.8 1.2 cm, mean AWT at the pubis was 7.0 3.2 cm, and mean hip girdle measurement was 7.6 3.0 cm. The mean operative time was 171 98 min, estimated blood loss was 116 127 mL, and mesh size measured 700.8 477 cm2. CS was performed in 24% of patients and a panniculectomy in 34.4%. The rate of wound and total complications were 13.25% and 39.1%, respectively. There were four recurrences (2.7%); three of which were associated with splitting lightweight mesh; the fourth recurrence and one of the lightweight recurrences was associated with infection. Mortality was 0.66%. Mean length of follow-up was 15 14.5 mo (Table 2).
Table 2 e Patient characteristics. Mean SD/n (%) Age (y) BMI (kg/m2) Gender Male Female Race Caucasian Other Previous abdominal surgery Ventral hernia repair No. of previous VHR Follow-up (mo) Operative characteristics Operative time (min) EBL (mL) Defect size (cm2) Mesh size (cm2) CS Panniculectomy Preperitoneal mesh placement Length of stay (d) Complications Cellulitis Wound infection Mesh infection Wound vac placement DVT Readmission Hernia recurrence Total wound complications Total complications Mortality
55.4 12.6 33.3 7.8 57 (37.7) 94 (63.3) 123 (81.5) 28 (18.5) 144 (95.4) 91 (60.3) 2.29 1.8 14.9 14.5 171 98 116.0 127 178.0 214 700.8 477 35 (23.97) 52 (34.4) 95 (63.8) 5.6 5.0 29 20 1 2 1 7 4 20 59 1
(19.2) (13.3) (0.66) (3.0) (0.66) (4.6) (2.7) (13.25) (39.1) (0.66)
SD ¼ standard deviation; EBL ¼ estimated blood loss; DVT ¼ deep venous thrombosis. Patient demographics, operative characteristics, and complications for the study population.
3.2. Univariate analysis: CT measurements with primary outcomes Associations between individual CT measurements and primary outcomes are summarized in Table 4. Larger hernia defect width, length, and area were associated with performance of a CS (all P < 0.001). However, CS was not associated
Table 3 e CT measurement characteristics. Mean SD AW umbilical (cm) AW retrorenal (cm) Retrorenal (cm) AW pubis (cm) Hip girdle (cm) Defect width (cm) Defect length (cm) Defect area (cm2) SD ¼ standard deviation. Mean CT measurements are presented here.
3.5 2.6 1.8 7.0 7.6 8.5 7.3 178.3
1.8 1.3 1.2 3.2 3.0 5.0 5.6 214
5
j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e8
Table 4 e Association of CT measurements with surgical techniques and outcomes. Technique or outcome CS Yes No P value Panniculectomy Yes No P value Wound omplications Yes No P value Total complications Yes No P value Recurrence Yes No P value
AWT umbilical (cm)
AWT retrorenal (cm)
Retrorenal (cm)
AWT pubis (cm)
Hip girdle (cm)
Defect width (cm)
Defect length (cm)
Defect area (cm)
3.4 1.9 3.6 1.6 NS (0.46)
2.5 1.3 2.6 1.3 NS (0.80)
1.7 1.1 1.8 1.3 NS (0.92)
7.2 3.5 7.5 3.8 NS (0.64)
7.4 2.6 7.7 3.2 NS (0.90)
11.5 ± 5.2 7.6 ± 4.8 0.0002*
11.1 ± 5.9 6.0 ± 5.4 0.0008*
115 ± 93 49.4 ± 85 0.0006*
4.1 ± 2.0 3.3 ± 1.5 0.04*
3.1 ± 1.4 2.3 ± 1.1 0.0006*
2.0 1.4 1.7 1.2 NS(0.14)
9.3 ± 4.1 6.3 ± 2.8 0.0001*
8.7 ± 3.1 6.9 ± 2.7 0.003*
10.9 ± 5.4 7.2 ± 4.6 0.0001*
9.6 ± 6.6 6.4 ± 5.2 0.03*
109.5 ± 114 47.8 ± 70 0.007*
4.3 ± 2.2 3.2 ± 1.3 0.01*
2.9 ± 1.3 2.4 ± 1.2 0.007*
1.9 1.3 1.7 1.3 NS (0.19)
8.6 ± 3.3 6.9 ± 3.7 0.001*
8.1 3.2 7.3 2.9 NS (0.19)
10.0 ± 5.3 7.7 ± 5.0 0.01*
10.9 ± 6.1 6.3 ± 5.4 0.003*
116.4 ± 116 50.7 ± 73.5 0.005*
3.8 1.8 3.4 1.6 NS (0.25)
2.8 1.4 2.5 1.2 NS (0.20)
1.98 1.3 1.6 1.2 NS (0.07)
8.2 ± 3.8 6.8 ± 3.5 0.01*
7.99 3.0 7.3 2.96 NS (0.19)
10.2 ± 5.1 7.4 ± 4.9 0.0009*
9.5 ± 6.2 5.9 ± 5.2 0.005*
95.4 ± 104 46.5 ± 72.4 0.007*
1.7 0.6 3.6 1.7 NS (0.06)
2.5 0.9 2.6 1.3 NS (0.89)
1.7 0.7 1.8 1.3 NS (0.76)
7.7 1.9 7.3 3.6 NS (0.45)
9.2 2.9 7.5 2.9 NS (0.23)
10.8 4.8 8.4 5.1 NS (0.34)
4.5 7.5 5.9 NS (0.75)
51.2 66.8 91.1 NS (0.80)
NS ¼ not significant. Univariate association of CT measurements to predict complications or advanced AWR techniques in OVHR. Values set in bold are statistically significantly different (P < 0.05). * Represents significant P value 0.05 for all). Increased AWT measurements excluding retrorenal and hip girdle correlated with performance of a panniculectomy, and use of this technique was also associated with increased hernia defect width, length, and area (P < 0.05 for all). AWT measures (excluding retrorenal and hip girdle) and defect width, length, and area were associated with increased wound complications (all P < 0.05). Total complications were associated with increased AWT pubis and increased defect width, defect length, and defect area (P < 0.01). In this group of patients, recurrence was not associated with any of the CT measurements or hernia defect sizes (P > 0.05 for all), although increased AWT umbilical CT measurement approached statistical significance (P ¼ 0.06). The magnitude of association for the PCs, or “loadings” is summarized in Table 1. PC1 has high loadings on all variables except defect width, and PC1 explained almost 68% of the total variability among the variables. PC2 had the highest loading on hernia defect width and explained approximately 22% of the remaining variability among the variables. PC1 and PC2 jointly explained approximately 89% of the total variability among the variables. In univariate analyses, the composite variable PC1, a surrogate for CT measurements and correlated variables as described in Table 1, was associated with panniculectomy, wound complications, and total complications (all P < 0.05), but was not associated with CS. PC2 was not associated with wound complications, but did correlate with total complications, performance of CS and panniculectomy in univariate analyses (all P < 0.05).
3.3.
Multivariate regression using PCs
While controlling for confounding factors, the effects of calculated PC1 and PC2 on the primary outcomes are summarized in Table 5. Higher magnitude of PC1 was independently associated with increased wound complications (odds ratio [OR], 1.08; 95% confidence interval [CI], 1.01e1.16; P ¼ 0.024). None of the other outcomes including CS, panniculectomy, total complications, or hernia recurrence, had any significant association with PC1. PC2 (hernia defect width) was independently associated with CS on multivariate analysis (OR, 1.16; 95% CI, 1.03e1.30; P ¼ 0.013). Our multivariate analysis (age, sex, diabetes, tobacco use, and recurrent hernias) was selected on the basis of previously established and well-described clinical risk factors in an effort to control for confounding factors. Similar multivariate analyses have been published by our institution and others in the past, and these factors have been identified as attributing to confounding results between differing cohorts. This method of multivariate analysis has been used by other groups including Krpata et al. [20] when they used obesity, diabetes, chronic obstructive lung disease, smoking, and immunosuppression as covariates. In addition, a recent publication from our institution by Colavita et al. [21] used a similar multivariate analysis.
3.4. Clinical cut points for defect area and width predict CS Multiple cutoff values for hernia defect area and defect width to predict CS were examined. Table 6 summarizes the
6
j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e8
Table 5 e Multivariate logistic regression model: predictive factors for wound complications and CSTs. OR Outcome: wound complications PC1 1.080 PC2 1.038 Age 0.984 Male sex 0.735 Diabetes 0.995 Tobacco 1.828 Number of prior hernias 1.65 Outcome: need for CS PC1 0.960 PC2 1.159 Age 0.999 Male sex 0.477 Diabetes 0.998 Tobacco 1.85 Number of prior hernias 0.896 Outcome: recurrence PC1 1.00 PC2 1.00 Age 1.00 Male sex 1.00 Diabetes 1.00 Tobacco 1.00 Number of prior hernias 1.00
95% CI
P value
1.01e1.160 0.933e1.155 0.94e1.16 0.24e2.23 0.29e3.41 0.48e6.92 1.08e2.53
0.024 NS (0.490) NS (0.462) NS (0.587) NS (0.994) NS (0.375) 0.021
0.89e1.04 1.03e1.30 0.95e1.05 0.14e1.61 0.28e3.53 0.49e6.95 0.58e1.40
NS (0.303) 0.013 NS (0.959) NS (0.232) NS (0.998) NS (0.362) NS (0.626)
0.77e1.29 0.66e1.51 0.84e1.19 0.014e71.78 0.008e118.43 0.006e163.39 0.23e4.42
NS (1.00) NS (1.00) NS (1.00) NS (1.00) NS (1.00) NS (1.00) NS (1.00)
NS ¼ not significant. Multivariate regression revealed that PC1 and number of prior hernias significantly increase risk for wound complications in this population. Number of prior hernias, CS, and PC2 were evaluated as continuous variables. Values set in bold are statistically significantly different (P < 0.05).
sensitivity, specificity, and AUC for various cutoff values for defect area and defect width. The ROC curves are displayed in Figure 2. Defect widths >8.3 cm typically required CS (sensitivity 74%, specificity 63%, AUC ¼ 0.72). In addition, hernias >164 cm2 were more likely to be repaired using a CS (sensitivity 89%, specificity 73%, AUC ¼ 0.83).
4.
Discussion
In this study, the preoperative CT measurements of 151 ventral hernia patients were analyzed, and AWT measurements and hernia defect size correlated with postoperative complications and utilization of AWR techniques. Although recurrence did not correlate with any CT measurements of the abdominal wall or hernia defect size, increasing AWT in the umbilical region, at the level of the retrorenal measurement, and at the pubic symphysis was associated with increased risk of wound complications. Increasing hernia defect width, length, and area were associated with wound complications, total complications, CS, and panniculectomy. The retrorenal measurement was not strongly correlated with surgical outcomes in this study. On multivariate analysis, the composite variables PC1 and PC2 were associated with wound complications and CS, respectively. In addition, we used advanced statistical methods to attempt to establish a cutoff point after which patients would be more likely to require CS. We found
Table 6 e Predictors of CS. Predictor
Cutoff value
Sensitivity (%)
Specificity (%)
AUC
Defect area
150 cm2 157 cm2 164 cm2 170 cm2 183 cm2 7.2 cm 7.8 cm 8.3 cm 8.8 cm 9.4 cm
92.9 92.9 89.3 85.7 75 80 80 74.3 71.4 60
69.3 70.5 72.7 73.9 76.1 59 61 62.9 66.7 70.5
0.82 0.83 0.83 0.82 0.81 0.70 0.71 0.72 0.70 0.66
Defect width
Cutoff values and diagnostics of defect area and defect width in correlation with CS. Bolddcutoff values of defect area and defect width with best tradeoff between sensitivity and specificity and high AUC.
that defect areas >164 cm2 or width >8.3 cm were associated with performance of a CS. This finding would be useful for surgeons in preoperative planning and counseling. CS can be complex, requires expertise in AWR, and it also carries associated unique risks and benefits for the patient including increased risk of wound infection, dehiscence, ischemia, and hernia recurrence [22]. Our study references and is somewhat similar to the Franklin et al. [13] article, published in September 2013 in Annals of Plastic Surgery; however, the previous study is a small series of only 54 patients. The authors write: “No cutoff value was able to be calculated because of small sample sizes in each group.” Our larger cohort of 151 patients allowed for this calculation and further predicted cutoff point of defect widths >8.3 cm and hernia area >164 cm2 needing CS. This study adds to the body of literature and provides objective evidence of a cutoff point that is clinically applicable. Although we have not yet applied these cutoff points to an external data set for validation, we plan to do that in the near future to verify the conclusions. Our findings in this ventral hernia study population are in line with prior findings published in the colorectal and urology literature, which found that wound and total complications are increased with increasing AWT [5,7,8]. Patients with large ventral hernias (classified as hernia area >100 cm2) are at increased risk of seroma formation and worse long-term quality of life outcomes according to recently published data [12]. Patients in this cohort had average defect widths of 8.5 5 cm and defect areas of 178 214 cm2 and are at-risk patients. Another recent study evaluating the likelihood of abdominal closure after CS using CT measurements stated that there is a paucity of data concerning hernia-specific variables related to recurrence, although defect area was a factor increasing recurrence rates [13]. The advanced statistical techniques used here are useful for application to the ventral hernia population and can potentially impact current preoperative planning practices. Preoperative use of CT scans can help predict outcomes and type of procedures needed for improving results. Many patients have CT scans before surgery, especially if they have a recurrent ventral hernia, to examine the hernia contents, presence and placement of previous mesh, and to determine
j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e8
7
Fig. 2 e ROC curves of defect area and defect width for predicting CS. Measurements are recorded in centimeters. (Color version of figure is available online.)
the size of the defect. If surgeons are familiar with the data presented herein, measurements easily obtained from CT scans can predict wound complications and the need for components separation. Counseling patients preoperatively is important and identifying tools to do so can be very helpful. Despite potential applications for this data, several limitations exist. This is a single-institution study conducted at a highly specialized hernia center where fellowship-trained surgeons practice advanced AWR very regularly. Advanced experience and training in AWR may underestimate complications that might be seen in the general population. Grade III and IV hernia patients were excluded from this study, and therefore, caution should be employed before applying the results of this study, especially that of wound complications, to that population. In addition, biologic mesh cases were excluded, and we cannot extrapolate complications or techniques for the patients whose hernias are repaired with biologic mesh. Although hernia recurrence did not correlate to any AWT measurements or hernia defect size, the number of hernia recurrences was very low (n ¼ 4, 2.7%); Therefore, a type 2 error could exist where an actual clinical difference exists but the study is underpowered to detect it. Three of the recurrences were associated with splitting lightweight mesh; the fourth recurrence and one of the lightweight mesh recurrences had infectious complications. Three of the four had 2.6 average failed prior ventral hernia repairs, only one of the patients underwent CS. Preoperative planning to optimize outcomes, predict and minimize risks and cost, and allocate resources appropriately is important for improving patient quality of life and overall efficiency of the health care system. This study uses preoperative CT measurements to predict the use of advanced complex AWR for patients with ventral hernias. CT scanning preoperatively helps predict which patients may be at increased risk for wound and total complications. PC1 was associated very strongly with BMI in this study (90.2% correlation). Further research will be necessary before this composite variable can be used in the clinical setting; however, at
this time, it is clear that increased BMI (obesity) leads to an increased risk of wound complications. This has been demonstrated by other groups as well, such as Krpata et al. [20], who reported that patients undergoing open repair for grade II hernias had a 16% increased risk of developing a surgical site occurrence when they had comorbidities such as obesity, chronic obstructive lung disease, diabetes, and smoking history. This ability to identify at-risk patients has the potential to identify and modify patients’ risk factors preoperatively and impact operative techniques and outcomes. Larger multicenter studies will better define the impact of CT measurements for patients undergoing ventral hernia repair.
5.
Conclusions
Preoperative CT measurements of ventral hernia defects and AWT predict postoperative complications and the use of complex AWR techniques. In addition, increasing AWT correlates with wound complications and increasing defect width correlates with likelihood of performance of a CS. Analyzing CT measurements may play an important role in the future as these may be used to calculate risk of postoperative complications at the time of the preoperative clinic visit and predict the operative techniques likely required.
Acknowledgment Authors’ contributions: L.J.B. contributed significantly to the writing of the manuscript; S.W.R. and C.R.H. contributed to the writing and study design; J.D.W. contributed to the study design and data gathering; T.P. contributed to the statistical analysis and data gathering and article writing; A.E.L. contributed to the study design; V.A.A. was involved in the study design implementation, data collection, article writing,
8
j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e8
and editing; and B.T.H was involved in the study design, article writing, and editing.
Disclosure B.T.H., V.A.A., and A.E.L. have previously been awarded surgical research and education grants from W.L. Gore and Associates, Ethicon, Novadaq, Bard/Davol, and LifeCell Inc. All other authors have no potential conflicts or disclosures relevant to this work.
references
[1] Tokunaga K, Matsuzawa Y, Ishikawa K, Tarui S. A novel technique for the determination of body fat by computed tomography. Int J Obes 1983;7:437. [2] Sampaio LR, Simoes EJ, Assis AM, Ramos LR. Validity and reliability of the sagittal abdominal diameter as a predictor of visceral abdominal fat. Arq Bras Endocrinol Metabol 2007;51: 980. [3] Yoshizumi T, Nakamura T, Yamane M, et al. Abdominal fat: standardized technique for measurement at CT. Radiology 1999;211:283. [4] Balentine CJ, Marshall C, Robinson C, et al. Validating quantitative obesity measurements in colorectal cancer patients. J Surg Res 2010;164:18. [5] Hagiwara M, Miyajima A, Hasegawa M, et al. Visceral obesity is a strong predictor of perioperative outcome in patients undergoing laparoscopic radical nephrectomy. BJU Int 2012; 110:E980. [6] Anderson KM, Lindler TU, Lamberton GR, Baron PW, Ojogho OK, Baldwin DD. Laparoscopic donor nephrectomy: effect of perirenal fat upon donor operative time. J Endourol 2008;22:2269. [7] Shimizu A, Tani M, Kawai M, et al. Influence of visceral obesity for postoperative pulmonary complications after pancreaticoduodenectomy. J Gastrointest Surg 2011;15:1401. [8] Fujii T, Tsutsumi S, Matsumoto A, et al. Thickness of subcutaneous fat as a strong risk factor for wound infections in elective colorectal surgery: impact of prediction using preoperative CT. Dig Surg 2010;27:331. [9] Heniford BT, Park A, Ramshaw BJ, Voeller G. Laparoscopic repair of ventral hernias: nine years’ experience with 850 consecutive hernias. Ann Surg 2003;238:391. discussion 399e400.
[10] Hughes KC, Weider L, Fischer J, et al. Ventral hernia repair with simultaneous panniculectomy. Am Surg 1996;62:678. [11] Fekkes JF, Velanovich V. Amelioration of the effects of obesity on short-term postoperative complications of laparoscopic and open ventral hernia repair. Surg Laparosc Endosc Percutan Tech 2015;25:151. [12] Wormer BA, Walters AL, Bradley JF 3rd, et al. Does ventral hernia defect length, width, or area predict postoperative quality of life? Answers from a prospective, international study. J Surg Res 2013;184:169. [13] Franklin BR, Patel KM, Nahabedian MY, Baldassari LE, Cohen EI, Bhanot P. Predicting abdominal closure after component separation for complex ventral hernias maximizing the use of preoperative CT. Ann Plast Surg 2013;71:261. [14] Newcomb WL, Polhill JL, Chen AY, et al. Staged hernia repair preceded by gastric bypass for the treatment of morbidly obese patients with complex ventral hernias. Hernia 2008;12: 465. [15] Reid RR, Dumanian GA. Panniculectomy and the separationof-parts hernia repair: a solution for the large infraumbilical hernia in the obese patient. Plast Reconstr Surg 2005;116: 1006. [16] Borud LJ, Grunwaldt L, Janz B, Mun E, Slavin SA. Components separation combined with abdominal wall plication for repair of large abdominal wall hernias following bariatric surgery. Plast Reconstr Surg 2007;119:1792. [17] Chang EI, Foster RD, Hansen SL, Jazayeri L, Patti MG. Autologous tissue reconstruction of ventral hernias in morbidly obese patients. Arch Surg 2007;142:746. discussion 749e51. [18] Ventral hernia working group. Incisional ventral hernias: review of the literature and recommendations regarding the grading and technique of repair. Surgery 2010;148:544. [19] Muysoms FE, Miserez M, Berrevoet F, et al. Classification of primary and incisional abdominal wall hernias. Hernia 2009; 13:407. [20] Krpata DM, Blatnik JA, Novitsky YW, Rosen MJ. Evaluation of high-risk, comorbid patients undergoing open ventral hernia repair with synthetic mesh. Surgery 2013;153:120. http://dx. doi.org/10.1016/j.surg.2012.06.003. Epub 2012 Aug 3. [21] Colavita PD, Tsirline VB, Belyansky I, et al. Prospective, longterm comparison of quality of life in laparoscopic versus open ventral hernia repair. Ann Surg 2012;256:714. [22] Jensen KK, Henriksen NA, Jorgensen LN. Endoscopic component separation for ventral hernia causes fewer wound complications compared to open components separation: a systematic review and meta-analysis. Surg Endosc 2014;28:3046.
