This study provides first evidence of significantly decreased complement C3 levels in patients with severe anorexia nervosa, compared to healthy control subjects. There was a strong correlation between index C3 levels and patients' and controls' BMI values (Spearman correlation coefficient = 0.71, P
<0.001), suggesting that serum C3 levels may represent a clinically relevant serum marker reflecting the severity of disease, and potentially serving as a guide for monitoring the refeeding process. That is, in patients with anorexia nervosa and severely low body weight, in whom basic laboratory tests are often normal, low serum C3 levels can confirm biochemical evidence of severe illness. It is reasonable that serum C3 levels did not correlate with BMI in the anorexia group alone, reflective of the fact that a 'threshold' of severe illness from anorexia has been crossed at these profoundly low body weights, which occasioned the low C3 level. The fact that serum C3 levels did not statistically correlate with weight restoration over the course of treatment may have a complex explanation, and thus restoration of non-edematous weight remains the best marker of physiologic recovery in anorexia nervosa. In contrast to the findings on C3 concentrations, serum levels of complement activation fragments (C3a, C5a, C5b-9) and the extent of complement hemolytic activity (CH50) did not significantly correlate with the patients' BMIs. There are several potential explanations for this negative finding. First, a recent study described direct cleavage of C5 via thrombin, thereby bypassing the traditional activation cascade using C3 convertases or C5 convertases [13
]. As a result, C5a may be generated via thrombin-mediated coagulation abnormalities that have been documented in anorexia nervosa patients [17
]. In addition, phacocytic cells are able to directly cleave C5 and locally generate C5a [19
]. Maj and colleagues revealed that peripheral mononuclear cells (PBMCs) isolated from anorexia nervosa patients exhibited significantly elevated levels of activated intracellular G proteins, indicating increased PBMC activity in these patients [20
]. Thus, activated PBMCs and neutrophils may further contribute to alterations of C3a and C5a levels bypassing the traditional complement activation cascade.
There are few reports on complement activation in anorexia nervosa available in the peer-reviewed literature, dating back to the 1970s and 1980s [21
]. Wyatt et al
. published a series of five anorexia nervosa patients and observed significantly decreased serum levels of C1q, C2, C3, factor B, β leutenizing hormone (β-LH), C3b inactivator, properdin, and C4 binding protein [22
]. After initiation of alimentation, β-LH, C3b inactivator, C3, and factor B rapidly returned to the normal range in response to therapy [22
]. In line with these findings, Sigal and colleagues found low serum levels of complement proteins in anorexia nervosa patients [23
]. A more recent report evaluated several components of the complement cascade and analyzed the activities of the alternative complement activation pathways [24
]. Serum levels of C3, Factor B and D, hemolytic activity of the alternative pathway, and the inhibitors H and I were found to be low in anorexia patients and normalized with weight gain [24
]. In our current study, we determined low C3 levels in anorexic patients, which is in line with those previous findings. However, while our findings suggest complement consumption secondary to increased activation in anorexic patients, Pomeroy and colleagues concluded that low serum complement levels were attributable to hypoproduction as opposed to increased consumption, and that percentage of ideal body weight, changes in body weight, and serum transferrin were each highly correlated with serum levels of complement proteins [24
]. These differences to our findings may be due to the fact that Pomeroy and colleagues assessed functional capacity of the alternative complement activation pathway exclusively, while our study focused on complement activation via the classical pathway (CH50) and complement activity further downstream (C3a, C5a, MAC). Moreover, Pomeroy et al
. failed to adjust their samples to total protein levels, which may have resulted in variable protein concentrations. In the present study, anorexia nervosa patients had significantly lower serum total protein levels than healthy controls on admission (61.9 ± 2.2 mg/ml vs 73.7 ± 2.3 mg/ml; P
= 0.003) [24
]. Nova and colleagues evaluated several biochemical markers in 14 anorexia nervosa patients and compared them to a healthy control cohort of (n = 15) [25
]. The authors reported significantly increased concentrations of C3 (and C4) upon admission in anorexia nervosa patients [25
]. At the 1-year follow-up, C3 and C4 levels had returned to levels comparable to the healthy control cohort. Nova et al
. also failed to adjust their measurements to total protein levels in their samples, which may account for the differences observed.
Our study has several limitations. First, the low patient numbers limit the power of our statistical analysis and make our data vulnerable to a statistical type II error. Therefore, our data do not allow for advocating complement serum levels as a new biomarker until definitively proven in future large-scale prospective studies. Moreover, follow-up studies will have to determine during which time frame complement levels return to healthy control levels after initiation of refeeding protocols, and whether complement serum levels may represent a valuable tool to monitor therapy success or failure in anorexia patients. Nevertheless, to our knowledge, our study is the first to describe a full complement screening in severely ill anorexia nervosa patients upon admission, and to correlate complement levels with gain of body weight as a function of time.