People with spinal cord injury (SCI) may show signs and symptoms of impaired thermoregulation due to damage of both afferent and efferent thermoregulatory pathways [1].

Core body temperature (Tcore) is regulated by a homeostatic process controlled primarily in the preoptic/anterior hypothalamic area [2]. Afferent information from thermoreceptors located throughout the body are conveyed mainly via the spinal cord to the temperature regulatory centers, which in turn activate mechanisms to produce or dissipate heat in order to maintain Tcore within normal physiological range [3]. The sympathetic nervous system is one of the main thermoregulation effectors, being responsible for heat conservation and metabolic thermogenesis through control of vasomotor tone and involuntary shivering and heat loss through control of the sudomotor system [4].

Tcore is also regulated by the circadian system, that is a periodic variability over the 24 h with maximal values in late afternoon and minimal in the early morning. This rhythm is paced by the suprachiasmatic nucleus of the hypothalamus, which also modulates other physiological functions including the sleep–wake cycle [5]. Indeed, thermoregulation and sleep are closely connected. On the one hand, humans usually initiate sleep on the downward slope of the circadian Tcore rhythm, when the rate of decrease of Tcore and peripheral heat loss via selective vasodilatation in distal skin regions are maximal and when endogenous levels of melatonin begin to rise. Such body heat loss promotes the ability to initiate and maintain sleep [6]. On the other hand, the different sleep (non-rapid eye movement (NREM) and rapid eye movement (REM)) and wake states themselves influence thermoregulation. In fact, Tcore is set at a lower level during NREM sleep than during wake, as an energy conservation function typical of sleep. During NREM sleep, thermoregulatory responses are operative, while REM sleep is characterized by a marked inhibition of thermoregulation, with changes in body temperature occurring passively in relation to the environmental heat load [7, 8].

Therefore, it could be hypothesized that the impaired thermoregulatory function in SCI individuals may also affect the sleep–wake cycle. As a matter of fact, the prevalence of sleep disorders seems higher in people with SCI compared with the general population, with regard to complaints of poor sleep quality as well as breathing and motor disorders during sleep [9].

So far, only a single study, performed at home with telemetry pill and accelerometer to record rest-activity cycle, has examined with formal chronobiological analyses the hypothesis that impaired circadian variation of Tcore is present in SCI [10]. In this paper, people with tetraplegia demonstrated a pronounced disturbance of the circadian rhythm of Tcore, whereas those with paraplegia and able-bodied controls were comparable. A subsequent study from the same group demonstrated that the increase in salivary melatonin during the evening hours was associated with decreases in core and skin temperature in controls and individuals with paraplegia, but not in those with tetraplegia, suggesting a role for melatonin in the Tcore circadian alteration in SCI [11].

The aims of our study were to analyze, for the first time to our knowledge, the circadian rhythm of Tcore in SCI individuals and able-bodied controls under controlled environmental conditions and to examine the impact of level of injury on the circadian modulation of Tcore by comparing participants with a high (cervical) and low (thoracic) spinal lesion. Furthermore, we aimed to assess the interaction between sleep–wake cycle and Tcore in people with SCI, comparing Tcore values between the different sleep stages and wake (state-dependent modulation).



This prospective observational study involved five male participants with cervical SCI (tetraplegia, level of injury C4–C7, cSCI) and seven male participants with thoracic SCI (paraplegia, level of injury T2–T12, tSCI) who were referred to the IRCCS Institute of the Neurological Sciences of Bologna (Italy) from the Spinal Unit of Careggi Hospital in Florence (Italy) for the assessment of autonomic and sleep functions. Inclusion criteria to enter the study were clinically stable condition (at least 6 months after the SCI) and the absence of symptoms or signs of diabetes, cardiorespiratory disease or other medical conditions that might have affected study results.

Seven age-matched, drug-free, and healthy able-bodied male participants were included for comparison.

The study was approved by the Ethics Committee of the Local Health Authority of Bologna (AUSL of Bologna, Italy) and performed in accordance with the Declaration of Helsinki. All participants gave their written informed consent to participate in the study.

Study protocol

All participants were admitted to our hospital and completed a 3-day protocol. They had previously been tested with sympathetic skin response (SSR) to assess neurogenic activation of sweat glands. Upon admittance, all participants underwent a general and neurological clinical examination. In the afternoon, they received an enema to prepare for rectal temperature monitoring and at night they underwent nocturnal cardiorespiratory monitoring for detection of sleep-related breathing disorders. All individuals with SCI had a permanent urinary catheter for the duration of the protocol. On the following morning, Tcore, arterial blood pressure, heart rate, and sleep–wake cycle were continuously monitored for 48 h under controlled environmental conditions and according to standardized procedures [12].

Rectal temperature was monitored every 2 min by a Mini-loggerTM (Bend, Oregon, USA) portable device, which had a resolution of 0.04 °C and an accuracy of 0.1 °C. Sleep–wake cycle was monitored by an ambulatory polygraphic recorder (Albert Grass Heritage®, Colleague TM PSG Model PSG16P-1, Astro-Med, Inc, West Warwick, RI, USA or Neurofax Electroencephalograph EEG-1200, Nihon Kohden Corp., Tokyo, Japan). Polygraphic recording included electroencephalogram (EEG: F3–A2, C3–A2, CZ–A1, O2–A1), right and left electrooculogram, electrocardiogram, electromyogram of the mylohyoideus, left and right anterior tibialis muscles, thoraco-abdominal breathing (strain gauge), and video-synchronized recording. During the study, participants were allowed to sleep ad libitum, while in a temperature (24 ± 1 °C) and humidity (40–50%) controlled room, lying in bed except when eating and for toilet breaks, in a light–dark schedule (dark period: 11:00 p.m.–7:00 a.m.). The individuals wore their own pajamas, which allowed them to feel comfortable in the room. The bed sheets were the same type for all participants and no blanket was required. Furthermore, the recording was video-monitored to assess the individuals’ compliance to the procedure. The participants were placed on a 1800 kcal/day diet, divided into three meals (8:00 a.m., 12:00 a.m., 6:00 p.m.) and three snacks (10:00 a.m., 4:00 p.m., 11:00 p.m.). Participants were instructed to avoid alcohol and caffeinated beverages and to abstain from smoking, from midnight preceding the monitoring.

Core body temperature and sleep–wake cycle analysis

Day–night changes of Tcore were assessed by calculating mean values of daytime (7:00 a.m.–11:00 p.m.) and night-time (11:00 p.m.–7:00 a.m.) Tcore. The nocturnal decline of Tcore was determined by subtracting night-time from daytime values (∆Tcore).

Rhythmicity was analyzed by evaluating the time series for Tcore according to the single cosinor method, using a computerized procedure (Chronolab) [13]. This procedure elaborates the raw temperature curve obtained across the 24-h period using a specific mathematical curve-fitting method (cosine curve). The procedure then determines whether or not there is a rhythm with a 24-h period (p < 0.05) and evaluates the following parameters within their 95% confidence intervals: (1) Midline Estimating Statistic of Rhythm (MESOR), that is the mean value of the cosine function; (2) amplitude, defined as the difference between the maximum value measured at the acrophase and the MESOR of the cosine curve used to approximate the rhythm; and (3) acrophase, defined as the interval between midnight hour (reference time) and time of highest value of the cosine function (Fig. 1). The acrophase indicates the presence of circadian rhythmicity and the timing of the peak of activity (phase), hence exploring the function of the suprachiasmatic nucleus. Conversely, MESOR and amplitude are representative of the homeostatic thermoregulatory process regulated by the preoptic/anterior hypothalamic area. For each participant, we analyzed the 24-h rhythmicity of the last day of recording, starting at 11:00 a.m.

Fig. 1: Parameters of circadian rhythm analysis.
figure 1

Visual representation of a normal 24-h profile of Tcore and parameters of the circadian rhythm analysis. Shaded area indicates night-time. Tcore core body temperature.

The 24-h sleep–wake cycle was visually scored in 30-s epochs according to the American Academy of Sleep Medicine criteria [14] as NREM sleep stage 1 (N1), NREM sleep stage 2 (N2), NREM sleep stage 3 (N3), and REM sleep. Total sleep time (TST), sleep efficiency (SE: time spent asleep/time in bed × 100), duration, and percentage referred to TST of each NREM and REM sleep stage were calculated over the light-off and light-on period.

Finally, we determined the mean value of Tcore in wake and in each sleep stage (N1, N2, N3, and REM sleep) over the light-off period.

Statistical analysis

For descriptive analysis, data were expressed as mean (standard deviation).

Comparisons of demographics and sleep parameters among groups were obtained from an analysis of variance performed via a linear model including the group indicator as regressor. Comparisons of Tcore values between groups were derived from an analysis of covariance based on a linear model having as regressor terms the group indicator and the two confounders, i.e., age and body mass index (BMI).

State-dependent analysis of Tcore was performed by using smoothing spline models for the analysis of samples of curves. We adopted a hierarchical additive model specification with a linear predictor depending linearly on the main effects and the interaction of group and state (i.e., wake and sleep stages), on the confounders age and BMI and on unknown group-specific smooth functions of time.

Data were analyzed using the software R (version 3.5.1) [15]. Statistical significance was set at p ≤ 0.05.


Participants’ characteristics

Clinical characteristics and medical regimens of each participant with SCI are listed in the Supplementary materials. There was no significant difference among groups in age (cSCI = 41 (18) years, tSCI = 40 (15), controls = 46 (11), p = 0.75) and time since injury (cSCI = 10.8 (9.9) years, tSCI = 1.7 (1.3), p = 0.07). BMI was significantly lower (p = 0.04) in cSCI compared with controls (cSCI = 23.8 (2.5) kg/m2, tSCI = 24.8 (5.0), controls = 28.9 (4.0)). All participants with SCI were classified as having a complete spinal lesion, being grade A in the American Spinal Injury Association Impairment Scale [16]. All SCI individuals but one were on stable medications, including benzodiazepines, antidepressants, opioids, and muscle relaxants, which we could not discontinue for ethical reasons. All participants with cSCI had absent SSR in all limbs. Thoracic SCI participants #7 to #12 had absent SSR in lower limbs and normal SSR in upper limbs; participant #6 had inconsistent SSR in upper limbs and absent SSR in lower limbs, whereas participant #13 had normal SSR in all limbs.

Day- and night-time values of core body temperature

Day and night-time values of Tcore in the three groups are shown in Fig. 2. The tSCI group had a similar pattern to that of controls, characterized by a gradual increase of Tcore values to reach a peak in the late afternoon, followed by a decrease during the night. Conversely, people with cSCI showed initial lower Tcore values compared with controls followed by a steeper continual increase until 10:00 p.m. and did not demonstrate the physiological nocturnal decrease of Tcore. At the end of the recording, Tcore in cSCI remained 1 °C higher than initial values.

Fig. 2: Day- and night-time Tcore in people with SCI and controls.
figure 2

Tcore core body temperature, SCI spinal cord injury, cSCI cervical SCI, tSCI thoracic SCI.

Day- and night-time Tcore values are reported in Table 1. Night-time Tcore values were significantly higher in cSCI compared with controls (p = 0.0003). ∆Tcore was significantly reduced in cSCI compared with controls (p = 0.0002). We found no statistical differences in any parameter between tSCI and controls, except for ∆Tcore (p = 0.04).

Table 1 Tcore values of participants with SCI and controls.

Circadian rhythm of core body temperature

Data obtained from the circadian rhythm analysis are reported in Table 1. A circadian rhythmicity of Tcore was detected in all participants (SCI and controls). Tcore MESOR was significantly higher in cSCI compared with controls (p = 0.002), whereas no statistical difference was found between tSCI and controls. Amplitude was comparable among the three groups (p = 0.78). Acrophase was significantly earlier in cSCI compared with controls (p = 0.05).

State-dependent analysis

The mean values of Tcore during wake, NREM, and REM sleep stages (Table 1) were significantly higher in cSCI compared with controls (p < 0.05 in all comparisons).

Tcore values distributed according to state (wake and sleep stage) and group (SCI, controls) are shown in Fig. 3. Tcore across states differed between SCI and controls: controls presented the expected Tcore reduction during NREM and REM sleep compared with wake and N1. This pattern of Tcore reduction was not observed in SCI, which displayed an impaired state-dependent modulation.

Fig. 3: State-dependent modulation of Tcore in people with SCI and controls.
figure 3

Tcore pattern over the 24 h was significantly influenced by each group (cSCI, tSCI, and controls) and state (W, N1, N2, N3, and REM). Tcore core body temperature, SCI spinal cord injury, cSCI cervical SCI, tSCI thoracic SCI, W wake, N1 NREM sleep stage 1, N2 NREM sleep stage 2, N3 NREM sleep stage 3, REM REM sleep.

Our analysis showed a statistically significant effect (p < 0.0001) for all predictors (age, BMI, group and state) on Tcore, as well as significant state–group interaction for N2 in both cSCI and tSCI and for N3 in cSCI. Tcore change over time was significantly different (p < 0.0001) between the three groups, highlighting a different dynamic of temperature in each group (cSCI, tSCI, and controls) across the states.

Sleep parameters

Sleep variables over the night-time (from 11 p.m. to 7 a.m.) are shown in Table 2. There was no significant difference in TST and SE among the three groups. Both cSCI and tSCI presented a significantly lower amount of N1 and higher amount of N2 compared with controls (p < 0.05 in all comparisons). All SCI showed a significantly reduced percentage of N3 compared with controls (p < 0.05). The percentage of REM sleep appeared to be reduced in cSCI and tSCI compared with controls, however it did not reach statistical significance (p = 0.12). All SCI and control participants manifested the physiological muscle atonia during REM sleep.

Table 2 Sleep parameters of participants with SCI and controls.

One cSCI (participant #2) and two tSCI (participants #10 and #11) did not reach N3 during the night-time. One cSCI (participant #2) and one tSCI (participant #8) did not reach REM stage during the night-time.


In this study, we analyzed the circadian rhythm and the state-dependent modulation of Tcore in people with SCI and healthy controls under controlled environmental conditions.

Our results showed that individuals with cSCI had an impaired physiological night-time decrease of Tcore and, compared with controls, presented significantly higher Tcore during night-time, whereas tSCI had a 24 h Tcore profile similar to controls. The circadian rhythm analysis of Tcore showed a higher MESOR in cSCI compared with tSCI and controls, while amplitude was comparable in the three groups. Acrophase was earlier in individuals with tetraplegia compared to those with paraplegia and controls. Overall our data suggest an impairment of both the circadian rhythmicity (i.e., advanced acrophase) and homeostatic control (i.e., increased MESOR) of thermoregulation in people with tetraplegia. In our study, performed in laboratory controlled conditions, the amplitude was comparable in the three groups. We cannot exclude that in a real-life setting, with more variable ambient conditions, amplitude might be different in individuals with SCI.

Considering the 24 h profile of Tcore, at the beginning of the recording Tcore in individuals with cSCI was lower than in controls, possibly due to lower metabolic state, altered body composition and interruption of vasomotor control caused by the SCI. After a few hours, Tcore in people with cSCI presented a steep increase and significantly deviated from the other two groups from 5:00 p.m. throughout the rest of the recording. This could be due to an impairment of the mechanisms of heat dissipation that usually become active in the evening in order to promote a decrease in Tcore and favor sleep initiation and maintenance.

Interestingly, we observed that the mean Tcore in cSCI individuals reached 37.9 °C at around 10:00 p.m., and at the end of the recording did not return to initial values, remaining ~1 °C higher. Such Tcore values may be unexpected considering the thermoneutral state of the investigation room and that the participants were in a resting state [17]. In our study, the only variables which were not under our control were the participants’ garments and preferences regarding the use of bed sheets to cover/uncover themselves during the study. We cannot exclude that persons with cSCI may have worn warmer garments compared with the other groups in order to feel comfortable. Furthermore, having tetraplegia they were unable to cover/uncover themselves at their will. Considering the impaired ability to dissipate heat, this may have lead Tcore to reach the high values observed. Moreover, people with cSCI may present an altered adaptation to the environmental temperature. Intercurrent infections could also induce a Tcore increase, however this hypothesis seems an unlikely explanation for our results considering the absence of signs/symptoms and the similar Tcore profile displayed by all individuals with cSCI. Finally, our data are not dissimilar from the only available previous study on prolonged Tcore monitoring in individuals with SCI [10], which reported values up to 37.5 ± 0.3 °C in non-controlled conditions.

Overall these observations further support an impaired circadian and homeostatic control of thermoregulation in people with tetraplegia.

Regarding the Tcore circadian rhythmicity, Thijssen et al. reported a higher nocturnal Tcore and an advanced acrophase in cSCI compared with tSCI and controls, but did not observe a higher MESOR in cSCI as we did [10]. However, this study [10] assessed participants during their standard routine activities at their home, and, even if they were instructed to remain indoors, stable environmental conditions could not be guaranteed. Our study, instead, was performed under controlled ambient conditions. Controlled ambient conditions allowed the exclusion of the well documented influence exerted on Tcore rhythm by external factors such as light–dark cycle, food intake, sleep, room temperature, humidity, and body position (the so called photic and non-photic entrainment, zeitgebers, of the human circadian system) [5]. Moreover, people with SCI were catheterized for the whole duration of the protocol, thus avoiding bladder stimuli that could have caused autonomic dysreflexia [18]. The presence/absence of these confounders may explain the discrepancies between their findings and our results.

Similarly to Thijssen et al. we did not find any difference in the 24-h variation in Tcore of tSCI compared with controls. Therefore, even if persons with tSCI have a reduced surface area available for thermoregulation, this may be sufficient to regulate Tcore.

Since the neural pathway for the endogenous production of melatonin passes through the cervical spinal cord [19] and people with tetraplegia show no nocturnal release of melatonin [11], Thijssen et al. hypothesized that these changes in endocrine function may contribute to the observed differences in Tcore circadian rhythm in people with tetraplegia compared with those with paraplegia. However, no interventional studies regarding the effect of melatonin supplementation on Tcore in SCI are available. Moreover, our data suggest a 24 h (daytime and night-time) elevation of Tcore in cSCI compared with tSCI and controls (even though it reached statistical significance only during night-time), which may be difficult to relate to the circadian secretion of melatonin [11].

Important new insights from our study were provided by the state-dependent analysis of Tcore. This analysis revealed that Tcore was significantly higher in cSCI during wake, NREM, and REM sleep stages compared with controls and tSCI. Furthermore, it demonstrated that both cSCI and tSCI did not display the physiological modulation of Tcore between wake, NREM, and REM sleep. Indeed, Tcore was modulated across the states differently in people with tetraplegia, paraplegia and controls.

In cSCI, the altered state-dependent modulation of Tcore could be partly explained by the impaired circadian and homeostatic control of Tcore that we found in this group. However, further mechanisms might be involved to explain the impaired state-dependent modulation that we observed also in tSCI individuals, who conversely presented a 24-h Tcore profile similar to controls. It would be possible that the disconnection between the spinal sympathetic nervous system and its central control caused by the spinal injury might hinder the modulation of the sympathetic activity during sleep. Another possibility would be that the altered sleep structure (higher representation of light sleep and reduced percentage of deep sleep) demonstrated by both SCI groups interfered with body temperature changes during sleep stages. Whether sleep alterations are secondary to Tcore impairment or vice versa is yet to be defined.

In our study, TST and SE were comparable between persons with SCI and controls; however, persons with SCI presented a higher proportion of light sleep (N2) and a lower representation of slow wave sleep (N3) compared with controls. Very few previous studies assessed sleep by means of polysomnographic study in persons with SCI, leading to conflicting results. Findings by Adey et al. support our results as they found a higher prevalence of light sleep (78 ± 9.6% of TST) in six persons with high cervical lesion; however, they did not find abnormalities in the one participant with a lower thoracic lesion [20]. A marked diminution of slow wave sleep was also reported in another study involving 22 participants (15 cervical and 7 thoracic SCI) [21]. Conversely, Scheer et al. [22], assessing three persons with cervical and two persons with thoracic SCI did not find any difference in the proportion of the different sleep stages in individuals with tetraplegia, paraplegia, and controls. Furthermore, they observed a significantly lower SE in cSCI compared with tSCI and controls. However, the number of participants included in this study is too small to draw any conclusion, and this may explain the discrepancies between our results and others.

Further studies will be necessary to assess the circadian and state-dependent relationship between sleep–wake cycle and the variation of Tcore observed in people with SCI. Notably, the study of both macrostructure and microstructure of sleep and the evaluation of the presence of specific sleep disorders in SCI might provide additional insights on the alterations observed in these individuals.

Finally, it is interesting that we found Tcore during REM sleep in cSCI to be different compared with tSCI and controls. Physiological REM sleep is characterized by an inhibition of thermoregulatory mechanisms, and the human body becomes poikilothermic. Yet, even in the absence of such physiological thermoregulation and with the impossibility to dissipate heat in a warm environment such as in our investigation room, REM Tcore of cSCI was dissimilar from the other groups.

We acknowledge the following limitations: we could not ethically discontinue medications of persons with SCI, which could have exerted an effect on sleep and Tcore. However, all participants were on stable dosages at the time of the study and well tolerated them. Furthermore, we could not present the data from the whole 24 h due to early termination of the protocol by some participants.


In the present study, we assessed, for the first time, the Tcore circadian rhythm and Tcore modulation according to wake and different sleep stages in individuals with SCI compared with controls under controlled environmental conditions. In summary, our results demonstrate that: (1) cSCI (but not tSCI) individuals presented an abnormal circadian rhythm of Tcore, and (2) both cSCI and tSCI individuals showed an impaired state-dependent modulation of Tcore and an altered representation of sleep stages. The pathogenic mechanisms responsible for the loss of state-dependent modulation of Tcore in people with SCI are yet to be defined. The damage to the sympathetic nervous system caused by SCI could affect thermoregulation including Tcore modulation during sleep. It is possible that the reduced representation of deep sleep in people with SCI also impairs such ability.

Further studies are necessary to investigate whether interventions intended to improve sleep could affect thermoregulation and vice versa. For instance, it is possible that a tailored treatment of sleep (including melatonin supplementation) could also lead to a better control of thermoregulation. Special attention should be paid to measures aimed at reducing trigger stimuli for autonomic dysreflexia, as we did in our study, so as to prevent episodes of uncontrolled and unstable activity of the sympathetic nervous system, which could worsen thermoregulation mechanisms and, consequently, the consolidation of sleep.