INTRODUCTION

Crohn's disease (CD) is a chronic inflammatory disease that can affect any segment of the gastrointestinal tract (1,2). Surgery is indicated in patients with obstruction and penetrating disease because of the accumulation of bowel inflammation and destruction. Approximately 80% of patients with CD may undergo surgery during their lifetime (3).

Postoperative complications in CD undergoing bowel resection are not rare. The incidence of postoperative complications ranges from 20% to 40% according to literatures (4–7). It had been recognized that the preoperative nutrition and immunologic status were strongly associated with postoperative complications of patients with CD (4). Tzivanakis et al studied 207 patients undergoing ileocecal or ileocolic resections for CD. The result demonstrated that the steroid usage and the presence of preoperative abscess formation were identified as independent predictors of anastomotic-associated complications, which almost doubles the risk of postoperative complications (14% vs 6%) (8). Therefore, establishing a comprehensive prognostic evaluation system is a key point to reduce the occurrence of postoperative complications and improve outcomes for patients with CD who will undergo bowel resection.

The prognostic nutritional index (PNI) has been considered a preoperative prognostic marker of various malignancies (9–12). Kanda et al conducted a retrospective study including 268 patients who underwent resection for adenocarcinoma of the pancreas and found that the PNI was associated with overall survival and postoperative complications, in particular pancreatic fistula, in patients with pancreatic cancer (13). In recent years, the importance of PNI has been expanded to predict surgical complications in the field of benign disease (14–16). Okita et al suggested that a poor PNI may be a significant predictor of postoperative infectious complications in patients with ulcerative colitis undergoing proctectomy with ileal pouch-anal anastomosis (16). However, limited evidence has been reported in patients with CD.

Thus, the aim of this study was to investigate whether the PNI could be a useful predictor of postoperative complications in patients with CD and to develop a nomogram for predicting major complications in patients who underwent CD-related bowel surgery.

METHODS Study design

Patients who underwent CD-related bowel surgery for the first time at the Department of Gastrointestinal Surgery of Ruijin Hospital from January 2013 to October 2019 were retrospectively analyzed. CD-related bowel surgery was defined as a surgical procedure to cope with major complications, such as obstruction, leakage, and refractory abscess, in CD. This study was approved by the Institutional Review Board of Ruijin Hospital, and the requirement for informed consent was waived because of the retrospective nature of this study.

The exclusion criteria were as follows: (i) computed tomography data not available 90 days before surgery or 30 days after surgery; (ii) patients with severe comorbidity, important organ insufficiency, malignancy, or HIV infection; (iii) a history of abdominal surgery; and (iv) perianal surgery.

Perioperative workup and management

We performed a retrospective review of these patients' records to retrieve specific data. Demographic data included age, sex, marriage status, body mass index (BMI), smoking history, and alcohol use. Normal BMI, overweight, obesity, and malnutrition were defined according to World Health Organization criteria as 18.5–25, 25–30, >30, and <18.5 kg/m2, respectively. Clinical data included disease duration, Montreal classification (17), and preoperative medication. Disease activity was assessed using the Harvey-Bradshaw index (HBI) (18). Laboratory test results, including serum albumin level, proalbumin concentration, white blood cell (WBC) count, hemoglobin, and platelet count, were routinely recorded before surgery.

Peripheral blood samples were collected 1 day before the operation to measure albumin (ALB) and total lymphocyte count (TLC) for PNI calculation. The PNI was calculated using the following formula: 10 × ALB (g/dL) + 0.005 × TLC (per mL). The probable cutoff of the PNI was determined by analyzing the receiver-operating characteristic (ROC) curve, and the most optimal cutoff value was used for further analysis.

The skeletal muscle mass index (SMI) of each patient was rated on the basis of the skeletal muscle mass measured through abdominal and pelvic computed tomography (19). The total muscle cross-sectional area (cm2) at the L3 vertebra was used for the segmentation of the skeletal muscle (including the psoas, paraspinal, and abdominal wall muscles). The threshold range for the skeletal muscle was from −30 to +150 Hounsfield units in accordance with reports (see Supplementary Figure 1, Supplementary Digital Content, https://links.lww.com/CTG/A905). Myopenia was identified on the basis of SMI (cm2/m2), which is the ratio of the skeletal muscle area (cm2) to the height squared (m2). According to the study by Martin L, myopenia was identified once a patient fulfilled one of the following criteria: (i) SMI <41 cm2/m2 in women, (ii) SMI<43 cm2/m2 in men with BMI <25 kg/m2, and (iii) SMI<53 cm2/m2 in men with BMI ≥25 kg/m2 (20).

Postoperative complications were registered mainly as skin or soft-tissue infections, major intra-abdominal leak, postoperative sepsis, postoperative thrombosis, and organ space infection. The length of postoperative hospital admission was also recorded.

Statistical analyses

The mean value and SD were calculated for quantitative and qualitative variables. Data between groups were compared using the Student t test for normally distributed values, and categorical data were compared using the χ2 or Fisher exact test as appropriate. The Kaplan-Meier curve was applied to estimate the effect of risk of postoperative complications on the length of hospital stay. P < 0.05 was considered statistically significant.

The least absolute shrinkage and selection operator (LASSO) logistic regression algorithm, which can efficiently analyze high-dimensional data, was applied to select the most significant predictive factors. A nomogram was constructed on the basis of the results of the LASSO logistic regression model. The Harrell concordance index (C-index) was used to measure the performance of this major complication nomogram. The C-index was then corrected by bootstrapping validation (1,000 bootstrap resamples). According to decision curve analysis, the clinical usefulness of the major complication nomogram was evaluated by quantifying the net benefits at different threshold probabilities. Statistical analyses were performed using the R statistical software (version 3.4.3; R foundation for Statistical Computing, Vienna, Austria; http://www.R-project.org/).

RESULTS Clinical characteristics of the patients

A total of 124 CD patients with initial CD-related bowel surgery in Ruijin Hospital from January 2013 to October 2019 were enrolled. The baseline characteristics of the patients are summarized in Table 1. Of these patients, 74 (59.7%) were males, and the mean age of the study cohort was 37.06 ± 13.08 years. The mean disease duration was 55.81 ± 51.67 months. Among them, 12 (9.7%) had a history of smoking and 34 (27.4%) were diagnosed with myopenia. Most of the patients (28.5%) had ileocolonic involvement, and the other patients had isolated ileal or colonic disease. Over half of the patients were receiving an immunomodulator or infliximab at the time of surgery. The median PNI was 37.2 (range: 15.3–55.5).

Table 1. - Basic characteristics of the patients enrolled in this study Characteristics Mean ± SD or n (%) Age, yr 37.06 ± 13.08 Male 74 (59.7) Disease duration, mo 55.81 ± 51.67 Alcohol 7 (5.6) Smoker 12 (9.7) Disease location  Ileal 19 (26.0)  Colonic 18 (14.5)  Ileocolonic 35 (28.5) Medications before surgery  No therapy 22 (17.7)  Steroid 20 (16.1)  5-ASA 24 (19.4)  AZA 36 (29.0)  MTX 4 (3.2)  IFX 18 (14.5) Penetrating behavior 35 (28.2) Myopenia 34 (27.4) PNI, median (range) 37.2 (15.3–55.5)

5-ASA, 5-aminosalicylate; AZA, azathioprine; MTX, methotrexate; IFX, infliximab; PNI, prognostic nutritional index.

Relationship between clinical laboratory characteristics and PNI

According to previous literatures (16,21), ROC curve analysis was performed to evaluate the ability of PNI to predict postoperative complications (see Supplementary Figure 2, Supplementary Digital Content, https://links.lww.com/CTG/A905). The optimal cutoff values for PNI and area under the ROC curve were determined as 33.9 and 0.673, respectively, corresponding to a sensitivity and specificity of 57.1% and 78.7%, respectively. On the basis of this cutoff value, 85 patients (68.5%) were categorized into the high-PNI (PNI >34) group and 39 patients (31.5%) were classified in the low-PNI (PNI ≤34) group. Age, sex, BMI, disease duration, disease location, and previous medication had no differences between 2 groups (Table 2). The serum albumin levels (23.4 ± 4.8 vs 35.8 ± 5.2 g/L, P < 0.001), hemoglobin levels (98.0 ± 24.1 vs 115.8 ± 22.2 g/L, P < 0.001), and WBC counts (8.3 ± 5.4 × 109 vs 6.3 ± 3.0 × 109, P = 0.009) of the patients in the low-PNI group were significantly lower than those in the high-PNI group.

Table 2. - Difference of clinical features in low-PNI and high-PNI groups Variable Patients, mean ± SD or n (%) P value PNI ≤34 (N1 = 39) PNI >34 (N2 = 85) Age, yr 36.7 ± 14.3 37.22 ± 12.7 0.85 Sex 0.496  Male 25 (64.1) 49 (57.6)  Female 14 (35.9) 36 (42.4) BMI, kg/m2 18.1 ± 3.0 19.0 ± 3.2 0.148 HBI score 5.7 ± 2.1 3.8 ± 2.9 <0.001 WBC, ×109/L 8.3 ± 5.4 6.3 ± 3.0 0.009 Hemoglobin, g/L 98.0 ± 24.1 115.8 ± 22.2 <0.001 Platelet, ×109/L 261.5 ± 122.2 287.6 ± 119.0 0.798 Albumin, g/L 23.44 ± 4.8 35.8 ± 5.2 <0.001 Disease duration, yr 4.1 ± 4.4 4.9 ± 4.3 0.382 Location 0.951  L1 (ileal) 18 (46.2) 37 (43.5)  L2 (colonic) 2 (5.1) 4 (4.7)  L3 (ileocolonic) 19 (48.7) 44 (51.8) Behavior 0.037  Nonstricturing, nonpenetrating (B1) 3 (7.7) 19 (22.4)  Stricturing (B2) 20 (51.3) 47 (55.3)  Penetrating (B3) 16 (41.0) 19 (22.4) Perianal disease 6 (15.4) 28 (32.9) 0.052 Preoperative medication 0.509  No therapy 10 (25.6) 12 (14.1)  Steroid 7 (17.9) 13 (15.3)  5-ASA 8 (20.5) 16 (18.8)  AZA 10 (25.6) 26 (30.6)  MTX 3 (7.7) 15 (17.6)  IFX 1 (2.6) 3 (3.5) Overall postoperative complications 20 (51.3) 15 (17.6) <0.001 Specific complications 0.244  Skin or soft-tissue infection 10 (25.6) 8 (9.4)  Major intra-abdominal leak 4 (10.3) 4 (4.7)  Postoperative sepsis 4 (10.3) 0 (0.0)  Postoperative thrombosis 0 (0.0) 2 (2.4)  Organ space infection 2 (5.1) 1 (1.2) Length of hospital admission (after operation) 23.46 ± 23.85 13.13 ± 13.00 0.002

5-ASA, 5-aminosalicylate; AZA, azathioprine; BMI, body mass index; HBI, Harvey-Bradshaw index; IFX, infliximab; MTX, methotrexate; PNI, prognostic nutritional index; WBC, white blood cell.

The surgical details of patients with CD in the 2 groups are listed in Supplementary Table 1 (see Supplementary Digital Content, https://links.lww.com/CTG/A905). Sixteen patients had emergency surgery. The main indication for surgery of patients with PNI ≤34 was perforation (48.7%) while 48.2% of patients in the high-PNI group were operated on because of complete or incomplete bowel obstruction (P = 0.003). Laparoscopic operations were performed in 23.1% of low-PNI patients while colostomy was performed in 30.8%, with no significant difference between the low-PNI group and the high-PNI group. Of the total patients, 35 cases (28.2%) of the overall postoperative complications were found in the total cohort. Such complications were observed in 20 (51.3%) of 39 patients in the low-PNI group compared with those in 15 (17.6%) of 85 patients in the high-PNI group (P < 0.001). The postoperative hospital admission of the patients in the low-PNI group was significantly longer than that of the patients in the high-PNI group (23.46 ± 23.85 vs 13.13 ± 13.00, P = 0.002) (Table 2).

Feature selection

The LASSO logistic regression algorithm was used to select the most significant predictors in the training data set. Then, these candidate predictors were used to construct the prognostic nomogram. In this study, patient demographics and clinical characteristics divided by postoperative complications are listed in Supplementary Table 2 (see Supplementary Digital Content, https://links.lww.com/CTG/A905). A total of 19 clinical features were used in the LASSO logistic regression, and 7 features with nonzero coefficients, namely sex, BMI, myopenia, age, indication, PNI, and HBI scores, were subsequently selected (Figure 1a and b). In our study, the age, BMI, and HBI scores were subdivided into several groups to enhance analysis.

Figure 1.:

Demographic and clinical feature selection using the least absolute shrinkage and selection operator (LASSO) regression model. (a) LASSO variable screening process. (b) LASSO model coefficients.

Development of an individualized prediction model

Based on the results of LASSO logistic regression analyses, a nomogram was generated to predict the postoperative complications in CD patients with initial bowel surgery (Figure 2). The C-index of this cohort was 0.824 (95% confidence interval 0.750–0.898), which was confirmed to be 0.701 through bootstrapping validation, indicating that this nomogram had good accuracy (Figure 3a). The area under the ROC curve for the prediction model was 0.824 (Figure 3b).

Figure 2.:

Nomogram for predicting postoperative complications in patients with Crohn's disease. BMI, body mass index; HBI, Harvey Bradshaw Index; PNI, prognostic nutritional index.

Figure 3.:

Assessment of the nomogram in the training set. (a) Calibration curve of the nomogram in the training set. The x-axis is the nomogram-predicted probability of postoperative complications, and the y-axis is the actual rate of postoperative complications. (b) Receiver-operating characteristic (ROC) curves of the nomogram in the training data set. AUC, area under the ROC curve. (c) The decision curve showed that the threshold probability is between >2% and <74% for our nomogram.

Clinical utility and validity of the nomogram

The clinical practicability of this nomogram was examined through decision curve analysis (Figure 3c). Our analysis indicated that this nomogram could be used to predict the postoperative complication risk if the threshold probability was between 2% and <74% predicted by this nomogram, contributing more benefits than the other schemes.

Performance of the prognostic nomogram in stratifying risk

In the training data set, the total prognostic scores calculated by the nomogram were categorized into 2 risk groups to predict the postoperative complications: high risk (score >273) and low risk (score ≤273) based on the cutoff value calculated through ROC analysis. A significant difference was observed in the length of postoperative stay between patients with high-risk and those with low-risk postoperative complications (17.07 ± 24.73 vs 10.36 ± 4.51, P = 0.02; see Supplementary Figure 3, Supplementary Digital Content, https://links.lww.com/CTG/A905). It showed from the side that this constructed nomogram model could well predict the occurrence of postoperative complications.

DISCUSSION

Various clinical parameters have been evaluated as a prognostic factor of CD surgery (19,22,23). However, the prognostic value of PNI has not been fully elucidated and analyzed in patients with perioperative CD. In this study, the results clearly showed a preoperative low PNI to be a predictor of postoperative complications in patients with CD who underwent bowel resection. However, the optimal cutoff value of PNI (33.9) was lower than reported. Mori et al reported a PNI of 45 for colorectal cancer in their study, which was calculated by the ROC curve. Okita et al reported a PNI of 47 for patients with ulcerative colitis undergoing proctectomy with ileal pouch-anal anastomosis (16). Possibly because patients with CD underwent generally poor nutritional status, approximately 75% of hospitalized patients with CD suffer from malnutrition and 33% have BMI <20 kg/m2 (22). Although the mechanism for the association between PNI and postoperative outcomes in patients with CD is not clear, this study showed that PNI was correlated well with disease activity, behavior, WBC, and other nutritional parameters. These findings were consistent with those of Zhou et al. (24).

The LASSO regression model revealed that sex, BMI, myopenia, age, surgical indication, HBI, and PNI were the risk factors of postoperative complication. Although, the PNI is calculated from the serum ALB levels and lymphocyte counts, the postoperative outcomes of this study were not associated with each parameter individually. It seemed that PNI was a better prognostic factor to predict postoperative complications of patients with CD. We also found that myopenia was an independent predictor of postoperative complications, which was consistent with our previous study (25).

Although numerous studies have focused on identifying the risk factors of postoperative complications, a scientific model that can integrate different parameters to build a prognostic model is lacking. A nomogram is a more comprehensive and integrative tool in discovering complication risks. It allows us to assess the risk of postoperative complications in a more scientific and objective manner. In our study, a novel nomogram was then developed and validated internally for predicting postoperative complications in patients with CD. The proposed nomogram was based on basically collected preoperative data and perioperative parameters to optimize its clinical applicability and ensure that it was feasible and simple to use. The nomogram showed that the score of PNI was higher than that of traditional nutritional parameters such as BMI and myopenia in this prognostic model. In our later analysis, the risk score of the nomogram had good correlation with the length of postoperative hospital stay.

Our study has several limitations. First, this study was retrospective and had a limited number of patients. As such, it might bring some bias in the statistics and the nomogram model. Second, a few patients received albumin supplement, and some patients preoperatively used steroids or immunosuppressants. They could affect the PNI. Finally, the nomogram was validated internally by bootstrap resampling; however, future studies are needed to validate the proposed nomogram externally using an independent data set. Thus, this nomogram could be modified and optimized.

In conclusion, PNI is closely associated with postoperative complications. Moreover, it is more promising for its convenience and more assignment than others. Its inclusion in a prognostic nomogram provides a highly efficient and convenient way to distinguish high-risk patients who may develop postoperative complications.

CONFLICTS OF INTEREST

Guarantor of the article: Zhengting Wang, PhD, and Jie Zhong, MD.

Specific author contributions: Z.W. and J.Z. contributed to the conception of the study. C.Z. and T.Z. performed the data analyses and wrote the manuscript. Z.S. helped perform the analysis with constructive discussions. All authors provided critical feedback and helped shape the research, analysis, and manuscript.

Financial support: This study was sponsored by the National Natural Science Foundation of China (No. 81802906) and Shanghai Sailing Program (No. 20YF1428200).

Potential competing interests: None to report.

Ethics approval and consent for publication: The study received approval from the institutional review board of Ruijin Hospital, Shanghai Jiaotong University School of Medicine. Written informed consent for publication was obtained from all participants.

Availability of data and materials: The data sets used or analyzed during this study are available from the corresponding author on reasonable request.

REFERENCES 1. Prideaux L, Kamm MA, De Cruz PP, et al. Inflammatory bowel disease in Asia: A systematic review. J Gastroenterol Hepat 2012;27(8):1266–80. 2. Ooi CJ, Makharia GK, Hilmi I, et al. Asia Pacific consensus statements on Crohn's disease. Part 1: Definition, diagnosis, and epidemiology: (Asia Pacific Crohn's disease consensus—Part 1). J Gastroenterol Hepat 2016;31(1):45–55. 3. Carter MJ, Lobo AJ, Travis SP. Guidelines for the management of inflammatory bowel disease in adults. Gut 2004;53(Suppl 5):V1–16. 4. Soop M, Larson DW, Malireddy K, et al. Safety, feasibility, and short-term outcomes of laparoscopically assisted primary ileocolic resection for Crohn's disease. Surg Endosc 2009;23(8):1876–81. 5. Iesalnieks I, Dederichs F, Kilger A, et al. Postoperative morbidity after bowel resections in patients with Crohn's disease: Risk, management strategies, prevention [in German]. Z Gastroenterol 2012;50(6):595–600. 6. Appau KA, Fazio VW, Shen B, et al. Use of infliximab within 3 months of ileocolonic resection is associated with adverse postoperative outcomes in Crohn's patients. J Gastrointest Surg 2008;12(10):1738–44. 7. Kasparek MS, Bruckmeier A, Beigel F, et al. Infliximab does not affect postoperative complication rates in Crohn's patients undergoing abdominal surgery. Inflamm Bowel Dis 2012;18(7):1207–13. 8. Tzivanakis A, Singh JC, Guy RJ, et al. Influence of risk factors on the safety of ileocolic anastomosis in Crohn's disease surgery. Dis Colon Rectum 2012;55(5):558–62. 9. Tokunaga R, Sakamoto Y, Nakagawa S, et al. Prognostic nutritional index predicts severe complications, recurrence, and poor prognosis in patients with colorectal cancer undergoing primary tumor resection. Dis Colon Rectum 2015;58(11):1048–57. 10. Cao X, Zhao G, Yu T, et al. Preoperative prognostic nutritional index correlates with severe complications and poor survival in patients with colorectal cancer undergoing curative laparoscopic surgery: A retrospective study in a single Chinese institution. Nutr Cancer 2017;69(3):454–63. 11. Kanda M, Fujii T, Kodera Y, et al. Nutritional predictors of postoperative outcome in pancreatic cancer. Br J Surg 2011;98(2):268–74. 12. Pinato DJ, North BV, Sharma R. A novel, externally validated inflammation-based prognostic algorithm in hepatocellular carcinoma: The prognostic nutritional index (PNI). Br J Cancer 2012;106(8):1439–45. 13. Mohri Y, Inoue Y, Tanaka K, et al. Prognostic nutritional index predicts postoperative outcome in colorectal cancer. World J Surg 2013;37(11):2688–92. 14. Chohno T, Uchino M, Sasaki H, et al. Associations between the prognostic nutritional index and morbidity/mortality during intestinal resection in patients with ulcerative colitis. World J Surg 2018;42(7):1949–59. 15. Ren H, Wu L, Hu W, et al. Prognostic value of the c-reactive protein/prognostic nutritional index ratio after hip fracture surgery in the elderly population. Oncotarget 2017;8(37):61365–72. 16. Okita Y, Araki T, Okugawa Y, et al. The prognostic nutritional index for postoperative infectious complication in patients with ulcerative colitis undergoing proctectomy with ileal pouch-anal anastomosis following subtotal colectomy. J Anus Rectum Colon 2019;3(2):91–7. 17. Silverberg MS, Satsangi J, Ahmad T, et al. Toward an integrated clinical, molecular and serological classification of inflammatory bowel disease: Report of a Working Party of the 2005 Montreal World Congress of Gastroenterology. Can J Gastroenterol 2005;19(Suppl A):5A–36A. 18. Harvey RF, Bradshaw JM. A simple index of Crohn's-disease activity. Lancet 1980;315(8167):514. 19. Ryan E, McNicholas D, Creavin B, et al. Sarcopenia and inflammatory bowel disease: A systematic review. Inflamm Bowel Dis 2019;25(1):67–73. 20. Martin L, Birdsell L, Macdonald N, et al. Cancer cachexia in the age of obesity: Skeletal muscle depletion is a powerful prognostic factor, independent of body mass index. J Clin Oncol 2013;31(12):1539–47. 21. Tominaga T, Nagasaki T, Akiyoshi T, et al. Prognostic nutritional index and postoperative outcomes in patients with colon cancer after laparoscopic surgery. Surg Today 2020;50(12):1633–43. 22. Kondrup J, Allison SP, Elia M, et al.; Educational and Clinical Practice Committee, European Society of Parenteral and Enteral Nutrition (ESPEN). ESPEN guidelines for nutrition screening 2002. Clin Nutr 2003;22(4):415–21. 23. Brennan GT, Ha I, Hogan C, et al. Does preoperative enteral or parenteral nutrition reduce postoperative complications in Crohn's disease patients: A meta-analysis. Eur J Gastroenterol Hepatol 2018;30(9):997–1002. 24. Zhou W, Cao Q, Qi W, et al. Prognostic nutritional index predicts short-term postoperative outcomes after bowel resection for Crohn's disease. Nutr Clin Pract 2017;32(1):92–7. 25. Zhang C, Yu D, Hong L, et al. Prevalence of sarcopenia and its effect on postoperative complications in patients with Crohn's Disease. Gastroenterol Res Pract 2021;2021:3267201.