The hypothesized implication and importance of liver volume reduction in the pathophysiology of severe malnutrition were examined and confirmed from multiple perspectives. To the best of our knowledge, this is the first study to examine liver volume in severe malnutrition. Focusing on the temporal relationship between hypoglycemia and the date of CT, we found that reduced liver volume was almost always associated with severe hypoglycemia (Figs. 3, 4, 5). In addition, hypoglycemia on admission was always accompanied by elevated ALT (Fig. 6b). Furthermore, the most unfavorable blood test outcomes over a 60-day period for all liver function-related parameters (Alb, AST, ALT, ChE, PT% [p < 0.001] and total bilirubin and platelet counts [p < 0.01]) were significantly poorer in the hypoglycemic group than in the non-hypoglycemic group (Table 4). These findings indicate that reduced liver volume is not a simple issue of size but is associated with serious liver dysfunction, including hypoglycemia.

In the five patients who showed severe hypoglycemia 13–18 days after admission to the OGMC, the amount of nutrients administered during this period was very low, and in four patients, the liver volume was markedly reduced during this period (Fig. 5a and Table 2). In three of these patients, as well as in 15 patients admitted with hypoglycemia, markedly low levels of TGs were concurrently observed in the blood (Table 2 and Fig. 6a). This implies that severe hypoglycemia manifests subsequent to the body's depletion of energy sources, encompassing TGs in the bloodstream, along with a shortage of external nutrients, and suggests that the precipitous reduction in liver volume immediately preceding hypoglycemia may indicate that the liver is utilized as an energy source. Figure 4, which illustrates the correlation between liver volume and BMI, reveals that the steep inclination of the regression line for patients with AN could signify liver consumption at a BMI lower than 13 kg/m2.

Liver volume from the control group with a median BMI of 20.85 kg/m2 suggested a normal liver eLW/IBW of 2.0%–2.5% (median: 2.22%) (Fig. 3). The threshold value of 1.08%, at which hypoglycemia occurred at an increased rate (Fig. 3), was approximately half the size of the native liver. Importantly, glycogen depletion in the liver can reduce its volume. However, the amount of glycogen that can be stored in the liver is approximately 100 g [22]; hence, all the volume reduction of the liver observed in this study could not be explained.

Notably, starvation-induced liver autophagy inevitably occurs in severely malnourished situations [16]. Rautou et al., who documented liver autophagy, included 12 patients with a PT% less than 50% and a median BMI of 11.9 kg/m2 [16]. Among them, four patients in whom autophagy findings were confirmed via electron microscopy exhibited a median BMI of 10.7 kg/m2 and a median PT% of 35%. In contrast, in this study, the 32 patients in the hypoglycemia group presented a median BMI upon admission of 10.16 kg/m2 and a median PT% of 47.3%, indicating that at least half of the patients with hypoglycemia had a PT% below 50%, mirroring the condition observed in the study by Rautou et al.

The poorest values in the first 60 days of treatment in the group with hypoglycemia were significantly worse than those in the group without hypoglycemia in most of the blood investigation items; prealbumin and total lymphocyte count exhibited significant associations (p < 0.001), as did TGs and hemoglobin (p < 0.01), along with serum magnesium and white blood cell count (p < 0.05). Furthermore, all the liver function-related parameters previously discussed were also significantly associated. However, the lowest serum P and K levels did not differ significantly between the two groups (Table 4). Moreover, transfusion of blood products was performed more frequently for patients with hypoglycemia than for those without hypoglycemia (Table 5). Delayed decrease of albumin, hemoglobin, and platelet is considered to reflect their disturbed production due to the hypoglycemic crisis, as these items are constantly replaced, and their blood levels are maintained by continuous production. These findings indicate that a range of physical complications encountered during re-nutrition is more closely associated with the pathology of malnutrition itself represented by hypoglycemia than with the narrowly defined refeeding syndrome, the primary pathogenesis of which is hypophosphatemia. In some other instances, thrombocytopenia in patients with AN is reportedly caused by decreased thrombopoietin production in the liver due to severe liver dysfunction [23]. Moreover, takotsubo cardiomyopathy and cardiac arrest may be associated with hypoglycemia, as well as low blood TGs, concomitant with elevated catecholamine levels in the blood [24, 25].

Furthermore, liver failure and severe hypoglycemia are associated with poor prognosis. Rich et al. [26], drawing from their own experience and previous studies, reported that seven of 10 patients with hypoglycemia died. In our study, all six deaths during hospitalization occurred in the hypoglycemia group, and hypoglycemic encephalopathy occurred in two other cases (Table 1). The post-discharge prognosis based on the information obtained from family members, related institutions, and details upon readmission to OGMC or other hospitals revealed that eight additional deaths and two additional cases of hypoglycemic encephalopathy occurred within 2 years of discharge; all 10 of these cases occurred in the hypoglycemia group. Severe hypoglycemia, accompanied by hepatic volume reduction and associated complications, constitutes the pathophysiology of the terminal stage of severe malnutrition, which is a major cause of death in AN. Refeeding syndrome is a well-defined concept that can lead to life-threatening complications; however, severe malnutrition itself is considered to have another pathophysiology closely related to poor prognosis.

A very small quantity of nutrition should not be sustained for an extended duration, as it may instigate and advance liver consumption. Importantly, during the initial years of the research period, we started minimal doses of nutrition and subsequently escalated them too judiciously. The 2006 edition of the NICE guideline recommends: “Start feeding 0.0418 MJ (10 kcal)/kg/day. Slowly increase feeding over 4–7 days. If the patient is severely malnourished (for example, BMI < 14 kg/m2), or if the intake is negligible for > 2 weeks, start feeding at maximum of 0.0209 MJ (5 kcal)/kg/day.” [3] Nevertheless, the guideline does not specify the duration for which low-calorie nutrition can be sustained. Moreover, it was posited that the feeding rate should be slowed down if refeeding syndrome is detected and that essential electrolytes should be administered [3]. In five patients who exhibited hypoglycemia on days 13–18 post-admission, an extremely low nutrient intake, averaging 252–556 kcal/day (approximately 10–20 kcal/kg/day), was maintained. One patient among them experienced severe hypoglycemia shortly after a reduction in nutritional intake, consequent to a gradual elevation in liver enzymes.

Rigaud et al. [27] measured the resting metabolic rate of 41 patients with AN with a BMI of 12.1 ± 1.5 kg/m2 at admission and reported it to be 845 ± 51 kcal/day. Additionally, Gentile et al. [28] measured the resting metabolic rate of 33 patients with AN with a BMI of 11.2 ± 0.7 kg/m2 and reported it to be 776 ± 145 kcal/day. Although not an exact analogy, simply dividing these values by the patients’ average admission weight yields values of 24.9 ± 1.5 kcal/kg/day and 26.7 ± 5.0 kcal/kg/day based on the studies of Rigaud et al. [27] and Gentile et al. [28], respectively. Per Scalfi et al. [29], who measured the basal metabolic rate of 120 female patients with AN in their teens and 20s, the resting metabolic rate of patients weighing 25–35 kg can be estimated to be 23–24 kcal/kg/day on average. Therefore, 25 kcal/kg/day might be an approximate basal metabolic rate in severely malnourished patients. Notably, energy consumption can increase when inflammation, trauma, or other physical illnesses occur.

Recent recommendations advocate initiating refeeding with a high caloric intake (1400 kcal/day or more) and escalating it rapidly, as this approach did not increase the risk of complications; rather, it reduced the hospital stays [30]. However, such nutritional therapies are designed for patients with a BMI of 15–17 kg/m2 [31,32,33,34] whose pathophysiology differs from that of severe malnutrition. Moreover, Garber et al. [35], in a systematic review, suggested that there is insufficient evidence to assert that a high-calorie intake is superior in cases of severe malnutrition. Patients suffering from severe malnutrition with smaller livers should be more susceptible to nutritional load than those with normal-sized livers. In this study, two cases treated by other physicians indicated that an overly rapid increase in nutrients can cause fatty liver and hepatomegaly, alongside other complications.

We did not measure the resting metabolic rate; instead, we assumed that the minimum nutritional requirement for day-to-day survival was approximately 700–800 kcal/day. Moreover, we posited that an extremely low nutritional intake during the initial days could exacerbate malnutrition. Conversely, refeeding syndrome, caused by excessive carbohydrate intake, must be avoided. Thus, we commenced with an intake of approximately 500 kcal/day, progressively increasing it to 800 kcal/day in a week. Importantly, the patient's physical condition stabilized when the intake was augmented to 1200 kcal after > 1 additional week. An increase in nutrients, including fatty acids and proteins (amino acids), along with phosphorus and potassium supplementation, could prevent deaths from undernutrition. Notably, this refeeding schedule for patients with severe malnutrition was informed by our experience and personal communications with experts in this condition. The verification of the re-nutrition protocol for such patients is beyond the scope of this study and warrants future investigation.

The liver, having lost volume, loses its capacity to regulate BG by rapidly converting ingested carbohydrates into glycogen for storage and subsequently releasing them into the bloodstream as needed. Hence, dosing regimens that substitute for the BG-regulating ability of the liver are of vital importance. Continuous 24-h dosing is the best, and in the case of oral intake, small, frequent doses are preferable. We encountered a case with hypoglycemia where liver enzyme levels rose after three oral doses per day but declined after six smaller doses per day. Notably, short-term, high-dose carbohydrate administration significantly burdens the liver, leading to elevated liver enzyme levels, pronounced hyperglycemia and insulin secretion, and a risk of hypoglycemia and refeeding syndrome due to excess insulin.

Furthermore, liver failure with reduced liver volume cannot tolerate fasting, and severe hypoglycemia can occur quickly. Thus, consuming midnight snacks for at least the first 2 weeks may be beneficial when sustained nutrition is lacking. One of the five patients with severe hypoglycemia after hospitalization refused to eat after breakfast, resulting in a deep coma the next morning due to severe hypoglycemia.

Carbohydrates not only burden the liver but also diminish blood P levels, consequently elevating the risk of refeeding syndrome [36, 37]. Lipids are regarded as the primary energy source during starvation and, notably, do not reduce blood P levels through insulin secretion. Additionally, animal studies indicate that liver autophagy elevates blood amino acid concentrations, and gluconeogenesis is initiated from amino acids [38]. Therefore, even amid hepatic contraction and liver failure, the preservation of BG levels using lipids and proteins (amino acids) can be sustained. Several cases in this study supported the benefit of lipid and amino acids to stabilize BG levels (data not shown). Although hypoglycemic episodes necessitate immediate glucose administration, a reassessment of the nutrition protocol is imperative to avert hypoglycemia recurrence.

Our study had several limitations that need to be considered. First, it was a single-center study. The pathophysiology of malnutrition may have been more prominent due to the administration of very low nutritional doses at our institution, especially in the early study period. Second, due to the retrospective observational nature of the study, missing values were unavoidable. Third, this study covers a span of 14 years, during which the treatment members and methods have evolved. However, two of the authors have been actively involved with OGMC for nearly the entirety of the study period. Fourth, the evaluation of liver volume was performed preliminarily. The reference values for liver volume in healthy participants and other factors affecting liver volume warrant investigation in future studies. Finally, liver biopsy was not performed in this study; therefore, microscopical evidence of liver autophagy was not confirmed in our cohort.