Fever and anapyrexia are the most studied thermoregulatory responses. They are defined as a body temperature (Tb) increase and decrease, respectively, occurring because of a shift in the set point (SP) and characterized by active defense of the new Tb. Although models of Tb control with a single SP (whether obvious or hidden) have been criticized, the SP-based definitions have remained unchallenged. In this article, the SP-based definitions of fever and anapyrexia were subjected to two tests. In test 1, they were compared with experimental data on changes in thresholds for activation of different thermoeffectors. Changes in thresholds were found compatible with an SP increase in some (but not all) cases of fever. In all cases of what is called anapyrexia, its mechanism (dissociation of thresholds of different effectors) was found incompatible with a decrease in a single SP. In test 2, experimental data on the dependence of Tb on ambient temperature (Ta) were analyzed. It was found that the febrile level of Tb is defended in some (but not all) cases. However, strong dependence on Ta was found in all cases of anapyrexia, which agrees with threshold dissociation but not with a decrease of the SP. It is concluded that fever (as defined) has only limited experimental support, whereas anapyrexia (as defined) does not exist. Two solutions are offered. A palliative is to accept that SP-based terms (anapyrexia, cryexia, regulated hypothermia, and such) are inadequate and should be abandoned. A radical solution is to transform all definitions based on comparing Tb with the SP into definitions based on balancing active and passive processes of Tb control.
- body temperature
- febrile response
- balance point
fever and anapyrexia are the two most studied thermoregulatory responses. Fever is caused by infectious, inflammatory, and other stimuli. In the laboratory, it is often studied by injecting animals with bacterial lipopolysaccharide (LPS) or mediators of its action: platelet-activating factor, pyrogenic cytokines (e.g., interleukin-1β and tumor necrosis factor-α), and prostaglandins of the E series (PGE). Fever is defined as an increase in deep body temperature (Tb) occurring because of an increase in the thermoregulatory set point (5, 6, 15). The inverse response, known as anapyrexia (5), cryexia (17), or regulated hypothermia (9), is commonly defined as a decrease in Tb resulting from a decrease in the set point. It is thought to occur in injury, trauma, hypoxia, shock (e.g., LPS induced), heatstroke, intoxications (e.g., with ethanol), anesthesia, starvation, and other conditions.
The current definitions of fever and anapyrexia are based on a model of Tb control requiring a single set point, either obvious (physiological) or hidden (mathematical). A physiological set point was used in many early models in which Tb was compared with an independent signal (for review, see Refs. 14 and 41). Some more recent models [the most famous is one by Mitchell et al. (19)] involve comparing Tb- or heat flow-dependent signals with each other; these models can be described as having a mathematical set point. More than 20 years ago, Werner (41) demonstrated that all set point concepts are built on unnecessary and unproven assumptions, and that all of them represent special cases of a more general concept. Such a general concept is based on the balance of active (controlling) and passive (controlled) processes and requires neither a physiological nor mathematical set point. Similar concepts have become standard in several areas of neuroscience dealing with complex functions (23, 24). To accept Werner's concept would require transforming the current definitions of thermoregulatory responses: they should be based not on comparing Tb with the set point but on determining at which value Tb would balance in a given response. In many cases, such a fundamental transformation can be performed in a surprisingly simple way, i.e., by substituting the term set point with balance (or equilibrium) point. However, such a transformation has not happened. The set point-based definitions of fever and anapyrexia are assumed to work just fine and remain unchallenged dogmas.
In the present work, the validity of these definitions is questioned, and the definitions are subjected to a twofold analysis. First, it is analyzed whether these definitions agree with qualitative and quantitative experimental data on changes of thermoeffector activity. Second, the current definitions of fever and anapyrexia are used to derive a corollary that both responses should be insensitive to ambient temperature (Ta); this corollary is then checked against experimental data. The analysis shows that one of the two most studied thermoregulatory responses (anapyrexia) does not exist, whereas the other (fever) finds only limited experimental support. Two solutions are then offered: one attempting to alleviate this problem and the other to eliminate it.
Test 1: Is the Activity of Thermoeffectors During Fever and Anapyrexia Compatible With Set Point Changes?
During fever, Tb typically rises as the result of coordinated behavioral (e.g., seeking a warmer environment) and autonomic responses; the autonomic responses involved are aimed at decreasing heat loss (e.g., skin vasoconstriction) and increasing heat production (e.g., activation of nonshivering thermogenesis in the brown adipose tissue; see Ref. 15). Similarly, it has been found in many (for review, see Refs. 9 and 34), but not all (18, 31, 37), cases of anapyrexia that the fall in Tb is also achieved by coordinated behavioral (e.g., seeking a cooler environment) and autonomic responses; the autonomic responses involved are aimed at increasing heat loss (e.g., skin vasodilation) and decreasing heat production (e.g., inhibition of thermogenesis). In other words, both fever and anapyrexia can occur as the result of several effector responses all aimed at changing Tb in the same direction. Such a qualitative approach is often used as proof that fever and anapyrexia can be adequately described as changes in the set point (e.g., Refs. 9, 15, 34). However, a quantitative approach produces different results.
The activity of each thermoregulatory effector is a function of Tb (and other arguments). At a certain Tb, the effector is activated, and this threshold Tb can be viewed as a Tb value that this effector defends. As proposed by Satinoff (32), confirmed by solid experimental data (for review, see Ref. 22), and implied by analogy with other biological control systems (23, 24), effectors that form the thermoregulation system are largely independent. Therefore, when two thermoeffectors change their activity in what looks like a coordinated fashion (e.g., when skin vasodilation occurs simultaneously with inhibition of thermogenesis), they can still have different thresholds and defend different values of Tb. Hence, measurement of threshold Tb values for activating effector responses (and not just thermoeffector activity) becomes a tool to probe the definitions of fever and anapyrexia. Figure 1 shows threshold Tb values for two thermoeffectors, one representing cold-defense effectors (Tthr-cold) and the other representing heat-defense effectors (Tthr-heat). If fever and anapyrexia can be adequately described as an increase and a decrease (respectively) in a single thermoregulatory set point, as suggested by their definitions, experimental data should reveal parallel upward shifts of Tthr-cold and Tthr-heat in fever (Fig. 1B) and parallel downward shifts of Tthr-cold and Tthr-heat in anapyrexia (Fig. 1C).
The febrile response to LPS (and some other pyrogens) consists of at least three different phases (29), which are thought to be mediated differently (12) and have different thermoregulatory mechanisms (39, 40). When studying the second phase of LPS fever in rabbits, Vybíral et al. (40) found that Tthr-heat (ear skin vasodilation) increases by 1.0°C (reaches 39.9°C), whereas Tthr-cold (shivering) decreases by 1.5°C (reaches 37.4°C); such dissociated thresholds clearly contradict the set point definition of fever. However, the authors suggested (based on a literature analysis) that equal (or at least similar) shifts in Tthr-heat and Tthr-cold occur during the first phase of LPS fever. They later supported their suggestion by showing that Tthr-cold (cold thermogenesis) increases during the first phase of LPS fever in rabbits by 1.0°C and by estimating that Tthr-heat (vasodilation) shifts upward by a similar value (39). Szelényi et al. (38) obtained stronger support for the set point definition by measuring Tthr-heat (tail skin vasodilation) and Tthr-cold (thermogenesis) in rats and finding that both thresholds increase by the same value during PGE-induced fever.
Whereas it is plausible that equal upward shifts in Tthr-heat and Tthr-cold can take place during the first phase of LPS fever or during the response to PGE, equal (or even similar) downward shifts in Tthr-heat and Tthr-cold have not been found in any model of anapyrexia. In a rat model of LPS-induced shock (28) and a guinea pig model of heat disorder (25), Tthr-cold (thermogenesis) drops by ∼2°C, whereas Tthr-heat (skin vasodilation) hardly changes. In a rat model of starvation, Tthr-cold (thermogenesis) drops by almost 1°C, whereas Tthr-heat (skin vasodilation) does not change (31). In a rat model of injury (limb ischemia), Tthr-cold (shivering) decreases by a dramatic 4–5°C, whereas Tthr-heat (tail vasodilation) does not decrease; on the contrary, it increases by ∼1°C (37). Many studies in anesthetized humans show dissociation between Tthr-heat and Tthr-cold (for review, see Ref. 33). Indeed, opioids, intravenous anesthetics (e.g., propofol), and volatile anesthetics (e.g., isoflurane and desflurane) all decrease Tthr-cold by several degrees Celsius, but they typically do not change Tthr-heat and may even increase it. In other words, a large fall in Tthr-cold is common in what is called anapyrexia, whereas a similar fall in Tthr-heat does not occur. Even when a decrease in Tthr-heat is relatively large [as in the studies by Borona and Gautier (4) and Greif et al. (10)], it is still much smaller than the characteristic, several-degree decrease in Tthr-cold. The decrease in Tthr-heat (thermal polypnea) seen by Borona and Gautier (4) in hypoxic cats was only 0.4°C, and it would be even several times smaller if the authors were to use a more conservative definition of Tthr-heat. In human volunteers treated with the opioid agonist nalbuphine (10), Tthr-heat (sweating) decreased by 0.5°C, whereas the magnitude of the simultaneous decrease in Tthr-cold (shivering) was more than two times larger.
Test 2: Are Fever and Anapyrexia Independent of Ta?
For both fever and anapyrexia, their set point-based definitions emphasize active defense of the new (increased in fever and decreased in anapyrexia) level of Tb (6). Such a proposed defense should render these responses relatively insensitive to Ta. In the case of fever, data on the sensitivity of the response to Ta are a mixed bag. On one hand, the same dose of LPS (13, 27, 30, 35) or platelet-activating factor (11) can increase Tb at a near-neutral Ta but cause at least a transient decrease in Tb at a subneutral Ta. The ability to both increase and decrease Tb has also been shown for “pyrogenic” cytokines interleukin-1β (20) and tumor necrosis factor-α (2) and is thought to reflect the dependence of the Tb response on Ta (20). On the other hand, there are data showing that the febrile level of Tb during PGE fever is independent of Ta (7, 36). In the case of anapyrexia, stimuli and conditions that are thought to cause this response (e.g., hypoxia, shock-inducing doses of LPS, severe heat, anesthesia, starvation) always produce responses characterized by a strong dependence on Ta: when Ta is low, Tb falls deeply; when Ta is high, Tb decreases slightly or not at all (3, 8, 28, 33, 42).
In summary, both test 1 and test 2 have shown that at least some (but definitely not all) fevers satisfy the set point definition, whereas every anapyretic response studied is incompatible with such a definition.
Solution 1: A Palliative
An easy solution would be to cease using the term anapyrexia and similar terms referring to a nonexistent phenomenon, a decrease in the set point. One can notice that the strategy shown in Fig. 1, D–I, represents exactly what happens with thermoeffector thresholds in the so-called anapyretic states (18, 25, 28, 31, 33, 37) and during the later phases of fever (40; also see test 1 above). This strategy is characterized by a wide interthreshold zone formed by a drastically decreased Tthr-cold at the low end and a slightly decreased (D and G), normal (E and H), or even increased (F and I) Tthr-heat at the high end and represents the poikilothermic type of Tb regulation (6). Within this wide interthreshold zone, Tb is the result of passive heat transfer between the animal and the environment. It is no surprise, therefore, that this strategy is characterized by a strong dependence of Tb on Ta. Such dependence has been repeatedly seen in “anapyretic” states (3, 28, 33, 42; also see test 2). When the environment is subthermoneutral (for definitions, see Ref. 26), the cooling pressure pushes the balance point (and Tb) down, toward Tthr-cold (Fig. 1, D–F). A decrease in Tb also occurs (and is especially efficient) when the poikilothermic type of thermoregulation is coupled with cold-seeking behavior. Activation of cold-seeking behavior and suppression of warmth-seeking behavior have been found in hypoxia, hemorrhagic shock, hyperosmolarity, LPS shock, and other anapyretic states (1, 9, 16, 28, 34). When the environment is suprathermoneutral, the warming pressure pushes the balance point (and Tb) up, toward Tthr-heat (Fig. 1, G–I). If Tthr-heat is elevated (as during LPS fever; see Ref. 40), Tb increases (Fig. 1I). Such an increase is especially efficient when coupled with warmth-seeking behavior, which is typical during fever (15).
The poikilothermic type of Tb regulation is widely spread in the animal kingdom (6). Phylogenetically, the vast majority of ectothermic animals are also bradymetabolic and poikilothermic; the transition to homeothermy (a feature of endothermic, tachymetabolic animals) is associated with the emergence of thermogenesis. There are also heterothermic animals that switch between poikilothermy and homeothermy; they do so by turning thermogenesis on and off. As a logical continuation, homeothermic animals use the same mechanism to decrease their Tb during hibernation, REM sleep, starvation, hypoxia, shock, and intoxications: they shut down thermogenesis (drastically decrease Tthr-cold) and become poikilothermic. As suggested by Myers (21) for the thermoregulatory effect of alcohol, the term poikilothermy (not the set point-based anapyrexia or alike) should be used to describe thermoregulatory responses characterized by widening of the interthreshold zone.
Solution 2: A Cure
A more radical solution would be to transform all terms for thermophysiological responses so that they are based on balancing active and passive processes of Tb control (41) rather than comparing Tb with the set point. This would make fever a response in which Tb balances above its normal value and anapyrexia a response in which Tb balances below its normal value. The balance point-based definitions work for all cases where the set point-based definitions work (e.g., typical fever; Fig. 1B). Such definitions also work for all cases where the set point-based definitions do not work. For example, the dissociated changes in Tthr-cold and Tthr-heat seen in poikilothermy (Fig. 1, D–I) do not agree with a change of the thermoregulatory set point but readily agree with a change in the balance point, which has a definite position uniquely determined by Ta and other environmental factors in a poikilothermic state. Four examples of positions at which Tb can balance in poikilothermy are shown in Fig. 1: substantially below (D–F), slightly below (G), at (H), or above (I) the level of Tb seen in normothermy (A). Figure 1 clearly shows how the same adjustment of the thermoregulatory system (the same set of Tthr-cold and Tthr-heat) can result in drastically different Tb values [compare, e.g., Fig. 1F and Fig. 1I corresponding to the later phases of the response to LPS (40)]. Such a difference is difficult to explain by using the set point-based definitions: how can the same shift in the set point represent both fever and anapyrexia at the same time, or how can the same shift in the set point result in either a decreased or an increased Tb?
For an experimenter, the balance point-based definitions are much more useful than set point-based definitions. Who has not seen a study in which a final conclusion is that the set point is increased or decreased under certain conditions? Such a conclusion is of no utility for understanding the regulatory mechanism (14, 41). It creates the illusion of understanding but does not offer any mechanistic insight into what is happening with Tb control: the single “command center” required to compare Tb with the set point and to send specific “orders” to different effectors probably does not exist and, as such, cannot be studied. By eliminating the single set point (with all the underlying machinery), the balance point-based definitions draw attention to thermoeffector loops and passive elements of the system, i.e., to physiological and anatomic entities that exist and can be studied in direct experiments. For example, if a certain stimulus decreases the balance point of rat Tb, this stimulus most likely affects, directly or indirectly, the loop controlling thermogenesis in the brown fat, the major cold-defense effector in the rat. Hence, accepting solution 2 not only gets rid of definitions that are experimentally unconfirmed (anapyrexia) or applicable only to specific cases (fever) but also dispels the illusion of understanding the thermoregulatory processes and opens them for exploration.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2004 the American Physiological Society