Anorexia nervosa (AN) is clearly a disorder of under-eating, but the question of whether appetite is impaired remains controversial and poorly researched. Studies employing subjective assessment have consistently reported reduced hunger and desire to eat and enhanced satiety and sensation of fullness in people with AN (e.g. Halmi & Sunday, 1991; Robinson, 1989). Furthermore, the subjective reward value of food is reduced (Drewnowski et al., 1987; Sunday & Halmi, 1990), and the rate of eating is slow (Halmi & Sunday, 1991).
Some authors argue that these findings reflect tight cognitive control of normal appetite (Palmer, 2000). However, relative to healthy comparison women, those with AN show reduced salivation (LeGoff et al., 1988) and a heightened autonomic response to food (Leonard et al., 1998). Images of food elicit fear and disgust (Ellison et al., 1998). These objective data suggest that appetite may indeed be impaired in AN (Pinel et al., 2000), although some capacity to respond to hunger and satiety cues clearly remains (Cugini et al., 1998; Rolls et al., 1992).
A predisposition to leanness may be a risk factor for AN (Hebebrand & Remschmidt, 1995), supporting the notion that heritable risk for the disorder may be exerted through the biological systems regulating appetite and weight.
Peripheral Signals in AN
In acute AN, basal and fasting insulin and IGF-1 concentrations are reduced (e.g. Alderdice et al., 1985; Argente et al., 1997; de Rosa et al., 1983; Fukuda et al., 1999; Stoving et al., 1999a) while growth hormone (GH) (de Rosa et al., 1983; Stoving et al., 1999b) and cortisol concentrations are elevated (Stoving et al., 1999c for review). These findings are consistent with the need to promote gluconeogenesis in the starving state and resolve with weight gain (Casper et al., 1988; Casper, 1996; Golden et al., 1994), suggesting that they arise as an adaptive response to prolonged negative energy balance. Similarly, altered thyroid function in acute AN does not differ from that associated with malnutrition (de Rosa et al., 1983) and normalises with weight restoration (e.g. Komaki et al., 1992). The endocrine response to a meal may be abnormal in AN. For example, the insulin response to a meal is delayed and reduced (Alderdice et al., 1985; Blickle et al., 1984).
This may be attributable to delayed absorption and, in chronic cases, impaired pancreatic beta cell function (Nozaki et al., 1994). Insulin sensitivity appears to vary greatly between individuals with AN (Kiriike et al., 1990) and there may be a tendency towards increased glucose utilisation rather than storage (Franssila-Kallunki et al., 1991). Blunting of the glucose and insulin response to a meal persists even after full recovery from AN, when there is also evidence of an abnormal cortisol response to the meal (Ward et al., 1998). It remains unclear whether these findings reflect vulnerability factors for the disorder, or a scar of the illness. Interestingly, the biology of restricting and binge–purge subtypes of AN may differ significantly. For example, relative to the binge–purge subtype, the restricting subtype is associated with higher basal and growth hormone-releasing hormone stimulated GH concentrations, lower basal IGF-1 (Brambilla et al., 2001) and elevated cortisol release (Connan et al., 2003a).
Enhanced satiety resulting from elevated basal and post-prandial CCK release (e.g. Brambilla et al., 1995a; Fujimoto et al., 1997) might contribute to the aetiology of AN. However, this finding has not been consistently replicated (e.g. Pirke et al., 1994) and, when present, resolves with partial weight gain (e.g. Tamai et al., 1993). Furthermore, the satiating effect of CCK may be reduced by the impairment of 5HT function associated with AN (Stallone et al., 1989). Pancreatic polypeptide release is elevated in response to a meal during the acute illness (Alderdice et al., 1985; Fujimoto et al., 1997) although the significance of this finding is unclear. Although a wealth of studies have examined a possible role for leptin in the aetiology of AN, little evidence of functional abnormality has been unearthed. During the acute illness, leptin concentrations are low (Casanueva et al., 1997; Herpertz et al., 2000; Monteleone et al., 2000a) but proportional to BMI and fat mass (Ferron et al., 1997; Grinspoon et al., 1996; Monteleone et al., 2000a), except at extremely low weight (Casanueva et al., 1997; Pauly et al., 2000). The diurnal rhythm of leptin and its relationship with insulin and IGF-1 release are preserved, but the temporal relationship with cortisol is disrupted until weight is restored (Eckert et al., 1998; Herpertz et al., 1998, 2000).Interestingly, there is a dissociation between BMI and leptin levels during weight gain (Hebebrand et al., 1997; Mantzoros et al., 1997) which could contribute to difficulty achieving full weight restoration with rapid refeeding. Similarly, leptin levels are elevated above that expected for BMI in those with the purging subtype of AN (Mehler et al., 1999). Elevation of the CSF : plasma leptin ratio in acute AN (Mantzoros et al., 1997) suggests that central levels may be relatively elevated, perhaps contributing to impairment of appetite. However, normalisation following full recovery from AN (Mantzoros et al., 1997) indicates that altered leptin dynamics are a state-related phenomena. In keeping with this hypothesis, leptin levels are predictive of amenorrhoea in eating-disordered patients (Kopp et al., 1997) and restoration of normal levels is necessary, but not sufficient, for restoration of menses (Audi et al., 1998).
Central Regulation of Appetite and Weight in AN
Far less is known about the central mechanisms regulating appetite and weight in AN. Data examining the role of hypothalamic peptides in the eating disorders are limited to estimations of plasma and CSF levels because of the inherent difficulty in measuring regional brain peptide activity in vivo. In terms of orexigenic networks of the hypothalamus, NPY levels are elevated in the CSF of those with AN (Kaye et al., 1990), consistent with a reduced leptin signal from the periphery and an orexigenic response to starvation. Levels normalise with full recovery (Gendall et al., 1999; Kaye et al., 1990), at which time CSF polypeptide Y (PPY) levels are also reported normal (Gendall et al., 1999). The NPY Y5 receptor mediates the orexigenic effects of NPY (Schaffhauser et al., 1997) and there is no evidence for an association between receptor gene polymorphisms and AN (Rosenkranz et al., 1998). Galanin has been measured only in peripheral plasma, in which levels do not differ from normal (Baranowska et al., 1997). There are no data currently available regarding orexin, AgRP, MCH or ghrelin function in AN, but all are potential candidates for an aetiological role. Plasma β-endorphin (β-EP) levels are reduced in acute AN (Baranowska et al., 1997).
Circadian rhythmicity is lost and there is some evidence of disrupted regulation of POMCderived peptide secretion (Brambilla et al., 1991). CSF levels appear to be reduced, but only in the most severely underweight patients (Kaye et al., 1987a). Although β-EP stimulates feeding, there is some evidence that infusion of an opioid antagonist improves weight gain during refeeding in AN (Moore et al., 1981). This may reflect blockade of central motivation and reward systems, or more direct effects on the hypothalamic regulation of feeding.
Again, little is known about the functional activity of anorexigenic peptides, such as MSH and CART, in AN. One recent study found elevated plasma CART concentrations in those with the acute disorder, and normal concentrations in women who had fully recovered (Sarah Stanley, personal communication). This finding must be interpreted with caution, because recent data suggests that CART may have orexigenic properties in at least some areas of the hypothalamus (Abbott et al., 2001).
Studies ofHPAaxis activity inANsuggest that an impairment of feedback inhibition at the level of the hypothalamus gives rise to elevatedCRHactivity. This is turn overrides feedback inhibition at the level of the pituitary, giving rise to the HPA axis hyperactivity associated with AN (Kaye et al., 1987b; Licinio et al., 1996). Elevated concentrations of CRH in the CSF (Kaye et al. 1987b) and blunting of the ACTH and β-EP responses to exogenous CRH (Brambilla et al., 1996; Hotta et al., 1986) are consistent with this hypothesis. Although HPA axis AVP activity is up-regulated in the context of chronic stress in animals and depression in humans, there is no evidence of elevated AVP activity in acute AN (Connan et al., 2003b, 2003c). This is significant becauseAVP modulatesCRHrelease and sensitivity of the axis to feedback inhibition via altered glucocorticoid receptor activity (Felt et al., 1984; Plotsky et al., 1984). An abnormal HPA axis response to chronic stress may therefore contribute to persistently elevated CRH activity in AN. In addition to the inhibitory effect on appetite and feeding, elevated CRH activity has widespread behavioural and physiological effects, which include many of the features of AN, such as: increased locomotor activity; cardiovascular changes; reduced social and sexual behaviour; impaired sleep; and increased anxiety behaviours (Dunn & Berridge, 1990). There is some evidence to suggest that a heightened cortisol response to stress and a blunted cortisol response to a meal may persist even after full recovery (Connan 2003a; Ward et al., 1998). Subtle abnormalities of HPA axis regulation might therefore contribute to susceptibility to AN.