Our analyses of patient characteristics between those with and without postoperative infections showed that age, ASA, total operation time, blood loss, surgical procedure, CRP levels (POD1), preoperative ADN levels, and ADN ratio were significantly different between groups (). In the logistic regression analysis, ADN ratio was still significant even when both CRP levels and ADN ratio were included in Model 4, indicating that ADN ratio was an independent risk factor of postoperative infection following gastric cancer surgery. In addition, a history of T2DM was independently associated with postoperative infection in all models (), although a significant difference was not detected by univariate analysis (). CRP level was not an independent risk factor of postoperative infection. From the ROC curves, ADN ratio was a better predictor of postoperative infection compared with blood loss, operating time, or CRP levels ().
The multivariate adjusted odds ratios for postoperative infection.
ADN is secreted by mature adipocytes; however, in contrast to leptin, lower ADN levels are associated with obesity, insulin resistance, diabetes, and disordered lipid metabolism 
and exhibits anti-inflammatory properties that inhibit monocyte adhesion and macrophage function 
. The acute perioperative reduction in ADN levels, as observed in patients who subsequently developed infections, might be a contributing factor mediating disordered postoperative energy metabolism and a vigorous inflammatory response, which may contribute to a greater propensity for infection.
Recently, aggressive insulin-based perioperative glucose control was shown to improve patient outcome, including substantial reductions in infection rates 
. It is likely that lowered ADN levels induce insulin resistance resulting in disordered glucose metabolism and subsequent infection. Hence, insulin therapy might counteract the dysfunction in energy regulation due to reduced ADN levels.
Although the mechanism for an acute reduction in ADN levels following surgery remains unknown, the plasma ADN levels in obese subjects were found to change only slightly despite significant diet-induced weight loss 
. Several substances released into the circulation following surgery have been reported to suppress ADN expression, including inflammatory cytokines, such as IL-6 and TNF-α, hypoxia, reactive oxygen species (ROS), and counter-regulatory hormones including catecholamines and corticosteroids 
. In the present study, the significant increase in postoperative IL-6 levels, a pro-inflammatory cytokine, in response to surgical stress, could account for the larger postoperative reduction in ADN ratios; this could be attributed to the fact that potent anti-inflammatory cytokines, including ADN, and pro-inflammatory cytokines have been proven to be antagonistic 
. Moreover, we performed logistic regression and ROC analyses for post-operative IL-6 and TNF-α levels; however, they were not important predictors of infection (data not shown).
An alternative explanation for decreased plasma ADN levels may be perioperative ADN protein binding, as lipoproteins have been shown to bind with lipopolysaccharides (LPS) during sepsis 
. The decrease in high density lipoprotein cholesterol (HDL) observed during sepsis is related to the effect of LPS on a wide range of apolipoproteins, plasma enzymes, lipid transfer factors, and receptors involved in HDL metabolism. All lipoprotein classes have been shown to bind LPS and can attenuate the biological response of LPS in vitro
and in rodents. We and other investigators recently observed that ADN binds to LPS to decrease ADN levels in rats with polymicrobial sepsis, with a reciprocal increase in TNF-α levels 
. LPS levels decrease in proportion to ADN levels. Although the underlying mechanisms are not entirely known, reduced ADN levels may result from LPS-ADN binding with subsequent sequestration. However, further studies are needed to better understand the mechanism behind the postoperative reduction in ADN levels.
Regarding the mechanism behind acute reduction in ADN levels following surgery, its significant association with blood loss disappeared when the ADN ratio was included in the logistic regression analysis in models 3 and 4 (). This might be attributable to a strong correlation between ADN ratio and blood loss (), as it is most likely that plasma ADN levels decreased with blood dilution, thereby suggesting that blood loss is a stronger independent risk factor for postoperative infection.
We concluded that a postoperative decrease in plasma ADN levels was not due to blood loss or blood dilution due to infusion because the ADN levels were unchanged after the removal of 400 ml of blood for autotransfusion (median blood loss among the 150 patients, 395 ml) (data not shown).
Factors associated with wound complications following elective gastrointestinal surgery included smoking, male gender, perioperative blood loss, and duration of surgical comorbidities 
. In the present study, we found that the incidence of postoperative infection was greater in patients who sustained a substantial reduction in postoperative ADN levels and this correlation remained after statistical corrections for the effect of blood loss and perioperative inflammation.
As a limitation to this study, we could not validate the significance of the predictors in a validation cohort because the total number of patients was too small.
If ADN is the underlying cause of disordered postoperative energy metabolism resulting in more serious infections, treatments with an ADN secretagogue are justified. The administration of ADN secretagogues increases ADN levels and improves insulin resistance 
. Hyperglycemia and insulin resistance are common in critically ill patients, even if they did not have diabetes before. Intensive insulin therapy maintains blood glucose levels and reduces morbidity and mortality rates among critically ill patients in surgical intensive care units 
and minimizes the effects of insulin resistance, thereby substantially improving postoperative outcomes 
The CDC guidelines for the prevention SSIs recommend the administration of surgical antibiotic prophylaxis for 24 h in clean-contaminated surgeries, such as gastrectomy 
. In contrast, the duration of surgical antibiotic prophylaxis is 3–4 days in Japan 
. A multicenter examination is currently being performed by the Japanese Society for Surgical Infection: this examination evaluates prophylactic antibacterial properties of surgical antibiotic administration over 24 h and for 4 days after several types of major surgeries, including total gastrectomy. Because the ADN ratio can be assessed within 1–2 days after surgery, it can be used to predict the necessity of continual administration of antibacterial agents to control and/or prevent infections.