Energy expenditure in head and neck cancer: a systematic review and meta-analysis.

Introduction
As a group, head and neck cancers (HNC) are the seventh most prevalent cancer type globally, and their incidence is predicted to rise [1, 2]. While mortality rates are decreasing, the global healthcare burden of HNC remains significant [3]. HNC cancers originate in the oral cavity, pharynx, larynx, nasal cavity, or salivary glands [4], with the tumour and/or its treatment often affecting a patient’s ability to consume food and fluids as normal [5]. As a result, malnutrition is highly prevalent in this patient group, estimated to affect 57% of patients even prior to treatment [6].
Effective treatment of malnutrition is essential to avoid the known consequences of cancer-associated malnutrition, including shorter survival, poorer quality of life, increased hospital length of stay, post-operative complications, and healthcare costs [7–9]. Dietary fortification using recipe modification and oral nutrition supplementation to increase nutrient intake and pharmacological control of treatment-associated nutrition impact symptoms such as nausea are common strategies to prevent or treat malnutrition [10, 11]. Unlike many other cancer types, the use of enteral tube feeding to enable nutrient provision during treatment is also common practice, related to tumour obstruction and/or treatment side effects such as severe mucositis post radiotherapy [11, 12]. The need for accurate assessment of energy expenditure to provide appropriate nutrition support is therefore especially important in HNC [13], where enteral tube feeding regimens may account for some or all of a patients’ nutritional intake.
Total energy expenditure (TEE) is comprised of resting energy expenditure (REE) as well as energy expenditure associated with physical activity and diet-induced thermogenesis [14]. The reference method for determining TEE is doubly labelled water, where elimination rates of ingested tracers in the urine are measured over a 7 to 14 day period [15]. The high cost and complexity of this technique impede its use in research and are barriers to use in clinical practice [16–18]. As a result, indirect calorimetry is the reference method most often used in research and clinical practice to determine REE through measurement of pulmonary gas exchange [19]. However, this technique also has limitations including the cost of equipment, user training, measurement time, and patient burden in the use of a face mask or canopy hood [19]. In practice, predictive equations developed from healthy cohorts are widely used to estimate energy needs of people with cancer (pREE), despite known issues with accuracy. Previous scoping reviews on energy expenditure in upper gastrointestinal [20] and gynaecological cancers [21] have reported heterogeneity of available evidence regarding the accuracy of predictive equations, methods for determination of hypermetabolism, and the level of detail in reporting of important participant characteristics.
To better understand energy requirements in people with HNC for accurate provision of nutrition intervention, the overarching purpose of this systematic review and meta-analysis was to synthesise the existing evidence regarding measured resting energy expenditure (mREE) or measured total energy expenditure (mTEE) in people with HNC. The primary aim of the review was to answer the research question, “In patients with HNC, how does energy expenditure measured objectively using reference methods (indirect calorimetry or doubly labelled water) compare to (a) energy expenditure in non-cancer controls, (b) predicted energy expenditure using predictive equations, and (c) energy expenditure measured using a wearable device?” The secondary aim was to report on longitudinal changes in measured energy expenditure, e.g., from before to after treatment.

Methods
This systematic review and meta-analysis is reported according to the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) guidelines [22]. The protocol for this review was prospectively registered with the PROSPERO international database of systematic reviews on 23 May 2024 (CRD 2024533495).

Data sources and search strategy
A systematic search of four databases was conducted on 3 June 2024 and updated on 29 November 2024, following consultation with an experienced medical librarian to ensure a comprehensive search strategy. Databases searched included: Ovid MEDLINE and Epub Ahead of Print, In-Process, In-Data-Review and Other Non-Indexed Citations, Daily and Versions (1946 to date), Embase via Ovid (1947 to date), CINAHL plus (EBSCOhost) (1937 to date), and the Cochrane Central Register of Controlled Trials. The search strategy included a combination of Medical Subject Headings (MeSH) and keywords; details of search terms used for each database are available in Supplementary File 1.

Study selection
Primary research studies reporting on energy expenditure (Outcome) measured using reference methods of indirect calorimetry or doubly labelled water (Intervention) in people with HNC (Population) were considered for inclusion. The eligible population was participants with any stage of HNC, undergoing any treatment modality, in any setting, with curative or palliative intent. Studies were eligible if they included a comparison (Comparator) to at least one of the following: energy expenditure of non-cancer controls measured using a reference method (indirect calorimetry or doubly labelled water), pREE derived from equations, or energy expenditure measured using a wearable device. Studies were also eligible if they included repeat measures of energy expenditure in patients with HNC at any timepoint, to assess longitudinal change. Studies investigating mixed cancer diagnoses were included only if energy expenditure data was reported separately for HNC. Letters, conference abstracts, systematic or narrative reviews, and primary research studies not able to be translated into English using artificial intelligence (Google Lens and Google Translate) were excluded.
References identified through database searching were exported to Endnote X9 [23], where duplicates were removed by the software. The remaining references were then uploaded to Covidence [24] for independent title and abstract screening by two reviewers. Full text review of potentially eligible studies was also conducted in duplicate. Conflicts were resolved through consensus discussion or a third reviewer before progressing through to each stage of screening.

Data extraction
Data extraction was conducted using a customised template which had been piloted with studies from a previous systematic review [20], and refined through discussion between reviewers. Details recorded included study characteristics (country, study design and setting, sample size), participant demographics and clinical information (age, sex, cancer type/s and stage, treatment status, weight, BMI, fat-free mass (FFM), energy expenditure assessment method, energy expenditure results (unadjusted energy expenditure per day, and/or adjusted for body weight and/or FFM), and comparative data (energy expenditure predicted using an equation or ratio method, energy expenditure measured in a non-cancer cohort, or energy expenditure measured using a wearable device). Data were extracted for every time point reported if the included study repeated assessments over time. Data extraction was completed in duplicate by two independent researchers; any errors or inconsistencies in the presentation of data that were identified through this process were corrected through further review and consensus discussion. Further details of data extracted from included studies are outlined in the registered protocol on PROSPERO.

Data synthesis and statistical analysis
All energy expenditure data were reported in kilocalories (kcal) with conversion from kilojoules (kJ) made where necessary using a factor based on 4.184 kJ per kcal. Unadjusted energy expenditure (kcal/day) was rounded to whole numbers for reporting. Inverse variance meta-analysis was conducted using Cochrane RevMan Version 9.1.0 [25], using random effects modelling in recognition of study heterogeneity [26]. Meta-analyses were conducted to determine the summary statistic for daily unadjusted mREE, to compare pre- and post-treatment mREE with pREE, and to investigate change in mREE from pre-treatment to end of treatment and post-treatment. The mean difference summary statistic (mean ± standard deviation) was used, as indirect calorimetry was used for mREE data in all studies. For studies where variance was not reported, standard deviations were imputed from either the average of standard deviations reported in the other studies in the meta-analysis, or, in the case of longitudinal measurements, imputed from baseline or follow-up data within the same study. Statistical heterogeneity using the I2 value was interpreted as low (I2 25%), moderate (I2 50%), or high (I2 75%) [27]. A p-value of less than 0.05 was considered statistically significant, and an a priori threshold of 10% was set to determine clinical importance [28].

Quality assessment
The Academy of Nutrition and Dietetics Quality Checklist for Primary Research was used to assess study quality and risk of bias [29]. Independent assessments were conducted for each study by two reviewers, followed by a group discussion with a third reviewer to resolve conflicts and reach consensus. For studies in which the primary aim was different to the aims of this review, outcome definition, measurement reliability and statistical analysis was considered appropriate if they were able to address the study’s own aims (questions 7 and 8). Studies were classified as either positive, neutral or negative quality. Certainty of the synthesised evidence was determined using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) system [30]; for key outcomes the domains of risk of bias, inconsistency, indirection and imprecision were critically appraised to produce an assessment of the certainty of evidence as either ‘high’, ‘moderate’, ‘low’ or ‘very low’.

Results

Study selection
The PRISMA flow diagram of study selection is presented in Fig. 1. The systematic database searches yielded a total of 1,262 studies. Following automated removal of duplicates, 788 references underwent title and abstract screening; at this stage, inter-rater reliability between reviewers demonstrated moderate to substantial agreement (Cohen’s Κ = 0.51 for LH-TB; K = 0.73 for LH-KN) [31]. Full-text review of 41 studies was conducted, with almost perfect to perfect reviewer agreement (K = 0.83 for LH-TB; K = 1.0 for LH-KN); 31 studies were excluded at this stage (Fig. 1). A hand search of reference lists of the 10 eligible studies identified 18 additional studies for screening, which were completed in duplicate and did not yield further studies. One study was identified from pre-print to meet inclusion criteria; thus, a total of 11 studies were included in this review.

Study characteristics
A summary of the characteristics of the 11 included studies is presented in Table 1. All studies reported mREE using indirect calorimetry; no studies reported mREE using a wearable device or TEE measured using doubly labelled water. Studies were published or in press between 1998 and 2024, in Brazil (n = 2), USA (n = 2), and Sweden, Spain, the Netherlands, the UK, Hong Kong, Switzerland, and Poland (n = 1 each). One study included participants with a specific HNC site (nasopharynx) [32]; all other studies included multiple HNC sites.

A total of 439 participants with HNC were enrolled in the 11 studies, with sample sizes ranging from eight [33, 34] to 140 participants [35]. Most studies were before-after studies (n = 8 studies), where REE was measured prior to treatment and at least one time point afterwards [32, 36–42]. Two cross-sectional studies [35, 43] and one retrospective cohort study [34] were included. Participants were naïve to medical and surgical treatment at baseline REE assessment in five studies (n = 153) [32, 37, 39, 40, 42]. In two studies [36, 38], up to half of participants had previously undergone surgery and baseline REE was assessed pre- and post-chemoradiotherapy (n = 50). In one study, all participants had previously undergone chemotherapy (n = 17) [41], and in three studies, treatment status at the point of REE assessment was mixed (n = 219) [34, 35, 43]. Protocols for measuring REE using indirect calorimetry were reported to varying degrees; all but one study [39] reported that participants fasted beforehand for periods ranging from five hours to overnight. Three studies reported the degree of physical activity leading up to testing [32, 34, 35], four studies reported rest periods immediately prior to testing (ranging from 15 min to one hour) [36, 41–43]; four studies reported neither of these parameters but stated the duration of the indirect calorimetry measurement (30–40 min) [37–40].

Measures of energy expenditure
Measured energy expenditure data is presented in Table 2. mREE data was reported unadjusted in 10 of 11 studies (all except Dev et al. [34]); in a meta-analysis (n = 431), the pooled estimate of pre-treatment unadjusted REE was 1502 kcal/day (95% CI 1387, 1616, I2 96%) (Supplementary Fig.  1A). The study by Dev et al. [34] did not report mREE data but reported a comparison with a predictive equation. Five studies reported mREE adjusted for body weight (n = 275), with a pooled estimate of pre-treatment mREE 23.8 kcal/kg/day (95% CI 21.5, 26.1, I2 96%, Supplementary Fig. 1B) [32, 35, 38, 39, 42]. Six studies reported mREE adjusted for FFM [32, 35–37, 39, 41], assessed using bioelectrical impedance in four studies (n = 261) [35, 36, 38, 39] and dual-energy x-ray absorptiometry in two studies (n = 55) [32, 41]. The pooled estimate of pre-treatment FFM-adjusted mREE (n = 316) was 31.3 kcal/kg FFM/day (95% CI 27.9, 34.8, I2 98%, Supplementary Fig.  1C); there was no subgroup difference between studies using BIA or DEXA to measure FFM (p = 0.74). Seven studies (58%) (n = 324) reported mREE expressed in more than one way [32, 35, 36, 38, 39, 41, 42].

Quality assessment and certainty of evidence
A summary of the quality rating for the 11 included studies is presented in Fig. 2. All studies were considered to have used blinding as the outcome of interest was measured using an objective test (indirect calorimetry). All studies clearly stated the research question and described REE measurement procedures and outcomes in sufficient detail. For most studies, questions relating to the comparison of study groups and handling of withdrawals were not applicable due to the design of the included studies (i.e., single group, observational studies). Eight studies received a ‘positive’ rating and three a ‘neutral’ rating; common reasons for downgrading the quality rating were a lack of clarity in (or omission of) reporting of inclusion and/or exclusion criteria and inappropriate or insufficiently described statistical analyses.

Primary outcomes

Measured energy expenditure compared with non-cancer controls
Measured energy expenditure was compared to that of non-cancer controls in one study [38]. Langius et al. [39] reported there was no significant difference in mREE between treatment-naive patients with HNC and age-, sex-, and FFM-matched non-cancer controls (n = 40 per group), that was unadjusted (1592 ± 304 vs. 1619 ± 244 kcal/day, p = 0.29), adjusted for body weight (21.9 ± 3.4 vs. 21.5 ± 3.3 kg/kg/day, p = 0.42), or adjusted for FFM (31.1 ± 5.0 vs. 30.7 ± 4.5, p = 0.38).

Measured energy expenditure compared with predictive equations
Eight studies (total n = 349) reported comparisons between mREE and pREE [34, 35, 38–43], presented in Table 3. Most studies conducted this comparison at baseline (pre-treatment) with or without an additional comparison at a specified follow-up time point; in three studies, the timing of the comparative assessment in relation to treatment was not reported [34, 35, 43]. Seven studies used the Harris-Benedict equation with or without additional equations; Zurlo et al. used the Fleisch equation. Unadjusted pre-treatment mREE was compared with pREE in five studies (n = 130) [38–42], with four studies reporting that mREE was higher than pREE prior to treatment (n = 113) [38–40, 42]: in one study by García-Peris et al. [38], group pre-treatment mREE was reported to be significantly higher than pREE by around 148 kcal (p = 0.035), and in three studies, the elevation in mREE was either non-significant [39, 40] or was less than 110% of pREE [42]. One study reported that mREE was slightly lower than pREE by around 2% (mREE/pREE 98 ± 11%) [41].

Meta-analysis of six studies (total n = 333) comparing pre-treatment mREE with pREE determined using the Harris Benedict equation is presented in Fig. 3. The pooled summary statistic indicates that pre-treatment mREE was significantly higher (mean difference 131 kcal/day) than pREE, with low heterogeneity (I2 15%). Only the study by García-Peris et al. [38] included a comparison of mREE adjusted for body weight or FFM, with pREE using the Harris Benedict equation; these adjustments did not change the result at any time point. The overall GRADE certainty of evidence for the comparison between mREE and pREE pre-treatment was low (downgraded due to serious risk of bias and serious imprecision).
Three studies (n = 51) compared mREE with pREE at multiple timepoints [38, 40, 41]; one study found a similar comparative result at the end of radiotherapy treatment (week 6) that was observed at baseline (mREE 1401 vs. pREE 1465 kcal/day, p > 0.05) [40], one study reported that the group mREE/pREE ratio increased to > 110% (i.e., hypermetabolic) at one month post-treatment where it had previously been 98% at baseline [41], and one study reported that mREE was significantly higher at pre-treatment, at the end of treatment, and 2 weeks post-treatment, but similar during treatment at week 2 and week 4 [38]. A meta-analysis of three studies (n = 69) with mREE vs. pREE comparisons conducted at the end or after treatment is presented in Fig. 4, demonstrating that mREE was significantly higher than pREE at follow-up (mean difference 108 kcal/day), with moderate heterogeneity (I2 49%). The overall GRADE certainty of evidence for the comparison between mREE and pREE at the end of treatment or post-treatment follow-up was very low (downgraded due to serious risk of bias and very serious imprecision).

Secondary outcome

Change in energy expenditure over time
Six studies (n = 192) reported on mREE at more than one time point in addition to baseline (pre-treatment) (Table 2) [32, 36, 38–41]. Four studies reported unadjusted mREE at baseline and the end of treatment, with a meta-analysis demonstrating a significant decrease in unadjusted mREE over this time (Fig. 5a, mean difference −136 kcal/day (95% CI −195.42, −75.73), p < 0.001, n = 143) [32, 38–40]. Three of these studies also reported repeated mREE adjusted for both body weight and FFM [32, 38, 39]; a meta-analysis showed no significant change from baseline to end of treatment when mREE was adjusted for body weight (Fig. 5b, mean difference −0.25 kcal/kg/day (95% CI −1.02, 0.52), p = 0.52, n = 127) or FFM (Fig. 5c, mean difference –0.57 kcal/kg FFM/day (95% CI −1.56, 0.43), p = 0.27, n = 127). Heterogeneity was low for all three meta-analyses (I2 0%). The overall GRADE certainty of evidence for the change in unadjusted and weight- or FFM-adjusted mREE from baseline to end of treatment was low (downgraded due to serious risk of bias and serious imprecision).
Five studies (n = 176) reported unadjusted mREE at baseline and within three months after treatment was completed [32, 36, 38, 39, 41]. A meta-analysis demonstrated a statistically significant decrease in unadjusted mREE over this time (Fig. 6a, mean difference −140 kcal/day (95% CI −190, −90), p < 0.00001, n = 176, I2 0% low heterogeneity); however, when adjusted for FFM in these same studies, the decrease in mREE from baseline to the post-treatment period was not significant (Fig. 6c, mean difference −0.47 kcal/kg FFM/day (95% CI −2.36, 1.42), p = 0.62, n = 176, I2 75% high heterogeneity). A similar result was observed in a meta-analysis of results from two studies (n = 109) where body weight-adjusted mREE within three months post treatment was not significantly different to baseline (Fig. 6b, mean difference −0.31 kcal/kg/day (95% CI −1.15, 0.53), p = 0.47, n = 109, I2 0% low heterogeneity). The overall GRADE certainty of evidence for the change in mREE from baseline to post-treatment follow up was low for unadjusted mREE (downgraded due to serious risk of bias and serious imprecision), moderate for weight-adjusted mREE (downgraded due to serious imprecision), and very low for FFM-adjusted mREE (downgraded due to serious risk of bias, very serious inconsistency, and serious imprecision).

Discussion
The purpose of this systematic review was to synthesize existing evidence regarding energy expenditure in patients with HNC, to support effective clinical nutrition practice and intervention in this nutritionally vulnerable group. To the best of our knowledge, this is the first study to summarise energy expenditure data measured using reference methods in HNC, with comparison to non-cancer cohorts or predicted energy expenditure derived from equations. All 11 included studies measured REE using indirect calorimetry; no studies measured TEE using doubly labelled water or compared with wearable devices. Most studies reported unadjusted daily REE; around half of the included studies additionally reported REE adjusted for body weight and/or FFM. Unadjusted mREE was higher than pREE before and after HNC treatment; the overall certainty of evidence according to GRADE was low for pre-treatment and very low for post-treatment comparisons. Only one study compared mREE in patients with HNC to non-cancer controls, reporting no significant difference. A decrease in unadjusted mREE from pre- to post-treatment was observed; however, when mREE was adjusted for body weight or FFM, the change over time was not significant.
The use of predictive equations to estimate energy requirements for people with cancer is standard practice in clinical settings for practical reasons, despite known issues regarding accuracy [44–46]. The most commonly used predictive equation (Harris Benedict) was developed from healthy populations [47] and has been demonstrated to be unreliable for estimation of energy needs in patients with cancer due to changes in REE associated with treatment and body composition [14]. Published scoping reviews examining comparisons between mREE and pREE in upper gastrointestinal [20] and gynaecological cancers [21] and a comparative study in breast cancer [48] have noted the heterogeneity of study findings regarding hyper-, normo-, and hypometabolism of mREE compared with predicted. In the present review, we also observed this heterogeneity, although mREE in patients with HNC was predominantly within 10% of predicted (normometabolic) or was more than 10% higher than predicted (hypermetabolic). To our knowledge, this is the first study to produce a meta-analysis for this mREE vs. pREE in cancer, where the pooled result demonstrated unadjusted mREE was higher than predicted in HNC by 131 kcal/day prior to treatment. While statistically significant, this was less than 10% of the summary value for daily unadjusted mREE; therefore, it is not considered to be clinically important (i.e. not evidence of hypermetabolism) at the group level. The heterogeneity in predictive equation accuracy observed in our study and others indicates that predictive equations should be used with caution at the individual level.
As FFM is a major determinant of REE, predictive equations incorporating FFM have been developed to improve accuracy [49–51]. Only one study included in this review compared predictive equations that incorporated body composition (such as FFM) with measured REE [43]; these equations all underestimated REE, and for two equations, the difference was clinically important (Table 3). Similar results have been reported in recent studies of people with breast cancer [48] and in older adults [50, 52], indicating little or no improvement in accuracy when predictive equations incorporating body composition are used. Porter et al. [50] validated new predictive equations for use in older adults over 65 years using FFM or lean body weight (derived from sex, height, and body weight [53]), developed using gold standard doubly labelled water-derived energy expenditure data in 1238 participants. In that study, the addition of FFM was found to add minimal value to the equation’s validity, and the authors hypothesized that FFM may already be accounted for by body weight. Further investigation into the discrepancy between measured and predicted FFM is needed to explain this finding [54]. Many factors aside from FFM, which are not incorporated into predictive equations, can also influence REE, such as fat mass and fat composition, type of fat, ethnicity, and amount of sleep [55, 56]. The use of FFM to improve the accuracy of predictive equations in cancer cohorts has yet to be determined, and further research is needed.
Evidence for elevation in energy expenditure in people with cancer compared with non-cancer cohorts has been inconsistent over several decades; while there is some evidence of a significantly higher REE in cancer compared with non-cancer controls [57], there is also a large body of evidence suggesting no difference or even lower REE in cancer [20]. Only one study included in this review compared mREE in HNC with non-cancer controls, reporting no difference [39]. This aligns with international consensus guidelines recommending that energy needs for people with cancer can be considered similar to healthy controls (generally 25–30 kcal/kg/day), as although there is increased REE in many patients as a result of inflammation or the increased metabolic demand of tumour glucose turnover and inflammation [56], overall TEE may be reduced due to a reduction in daily physical activity [10, 58]. This finding is also supported by a recent systematic review of studies assessing TEE using doubly labelled water reporting that mean weighted TEE in people living with cancer was lower than that of most other major chronic diseases [59]. However, there is some evidence that in HNC, energy intake at or above the top of the recommended range (i.e., 30 kcal/kg) is necessary to achieve relative stability in weight and muscle mass. This was demonstrated in a 2020 study of 1756 HNC patients which reported that in those consuming approximately 30 kcal/kg, mean weight loss was only 2.5%, suggesting that energy intake at the top end of this range would be consistent with relative energy balance [6]. This is further supported by a small 2019 longitudinal study, where energy intake of > 30 kcal/day during treatment was associated with significantly lower skeletal muscle loss compared to < 30 kcal/day (n = 41) [60]. Einarsson et al. [37] noted that treatment centres in Sweden often estimate energy needs at 30–35 kcal/kg/day for patients with HNC, which aligned with the TEE of their participants (n = 20) measured with a SenseWear armband before and after treatment. Overall, there is limited evidence comparing energy needs in HNC with non-cancer cohorts, and more research is needed to improve understanding in this area to inform practice and guidelines.
Repeat assessment of nutritional requirements throughout a patient’s cancer treatment is a cornerstone of evidence-based clinical nutrition practice, to facilitate change in interventions based on patient adherence and nutritional or treatment-related circumstances [11, 61]. A reduction in weight and/or muscle mass due to poor oral intake as a result of treatment side effects may lower energy needs; conversely, there has been some suggestion that the metabolic effect of chemotherapy treatment may increase REE [38, 62]. To our knowledge, this is the first study to meta-analyse change in energy expenditure data from pre- to post-treatment in patients with cancer. Unadjusted mREE in patients with HNC decreased significantly from baseline to the end of treatment and up to 3 months post-treatment; however, the decrease was not clinically important. The GRADE certainty of evidence for change in unadjusted mREE over time was low, related to study quality (risk of bias) and imprecision (sample size), indicating a need for larger, well-designed studies. Loss of weight and muscle mass during chemo- and/or radiotherapy for HNC is known to occur [63, 64]; over half of the studies included in the present review reported weight before and after treatment with group weight loss between 6–12% [32, 36–38, 40, 41], and three studies reported FFM loss of 5–11% from pre- to post-treatment [32, 36, 41]. It is likely that this observed decrease in mREE over time was contributed to by change in body composition, with subsequent meta-analyses using body weight- and FFM-adjusted mREE showing no significant change between pre- and post-treatment time points. This aligns with previous observational studies showing no meaningful change in REE over time in patients with breast cancer receiving chemotherapy [65, 66], and in patients with mixed cancer types receiving palliative care [67]. Taken together, it appears that timing in relation to medical cancer therapies may not be relevant in the assessment of energy expenditure.
Einarsson et al. [37] was the only study to measure TEE assessed using the SenseWear device, but this was not compared to indirect calorimetry or doubly labelled water as a reference method. There were no studies included in this review that reported on TEE in HNC using doubly labelled water. A 2024 systematic review [59] exploring the use of doubly labelled water to measure TEE in major chronic diseases found that only five out of 50 studies were conducted in people living with cancer (pancreatic, gastric, breast, colorectal, mixed cancers) or cancer survivors (acute lymphoblastic leukemia). The cost, complexity, and applicability to clinical care are commonly reported barriers to the use of doubly labelled water in acute care [17], and likely also affect the use of TEE assessment in cancer populations. Wearable devices have potential to improve the practicality of data collection for clinical and research purposes; however, their accuracy for energy expenditure assessment is poor in non-cancer populations, such as stroke, COPD, and healthy cohorts, compared predominantly with indirect calorimetry as the reference method [68]. In one small 2007 study (n = 10) [69], the SenseWear armband accurately estimated REE measured with indirect calorimetry in patients with acute myelogenous leukemia undergoing chemotherapy; however, to our knowledge, no other studies have since validated this or any other wearable device for energy expenditure measurement in people with cancer. Further research to validate the accuracy of wearable devices against reference methods in cancer cohorts could identify a more cost-effective and practical method of REE or TEE assessment, although evidence in other groups suggests this may not be the case.
Although adjustment of REE for body weight (five studies) and/or FFM (six studies) was conducted, there are known limitations with this method of analysis. The relationship between energy expenditure and either body weight or FFM is non-linear, with different tissue and organ types having differing metabolic needs. To divide REE by body weight or FFM is an oversimplification of this relationship and may hide or incorrectly indicate differences between groups [70]. The use of regression or analysis of covariance modelling where body weight or FFM are included as confounding variables in analytical models has been previously proposed as a solution to this issue, and should be incorporated into future studies examining energy expenditure in people with cancer [20, 56, 70, 71].
Strengths of this systematic review were the broad search strategy, including studies published at any time, in any language to capture all relevant information, the prospective registration of the review protocol, reporting of study methodology according to PRISMA guidelines, and the rigorous GRADE assessment of the certainty of evidence. The meta-analyses completed are an additional strength of this study, allowing pooling of data from studies with often small sample sizes. A limitation of this study was the exclusion of 11 publications of mixed cancers where REE data had not been reported separately by cancer type. The included studies were heterogeneous in terms of treatment modalities, timing of energy expenditure assessment, and pre-assessment fasting and activity protocols, which limits interpretation of pooled results [19, 72]. The influence of HNC stage on REE also cannot be determined from the current evidence base, but it is important to explore in future studies due to the potential for elevation in REE with increasing tumour burden [56, 73]. Overall, the certainty of evidence ranged from moderate (change in weight-adjusted mREE from pre- to post-treatment) to very low (mREE vs. pREE at end of or after treatment, and change in FFM-adjusted mREE from pre- to post-treatment); most GRADE assessments showed a serious risk of bias (relating to participant selection and/or statistical analysis), and either serious or very serious imprecision. There is a need for larger, well-designed studies using robust methodology to increase the certainty of evidence regarding energy expenditure in HNC.
This systematic review has synthesised the existing evidence regarding energy expenditure in people with HNC, with REE measured in all studies using indirect calorimetry. The Harris Benedict predictive equation significantly underestimated unadjusted mREE, but the difference was not clinically important at the group level. There was no meaningful difference observed in mREE from before to after treatment. Only one study compared mREE in HNC with non-cancer controls; no significant difference was observed, but the lack of evidence means the impact of a HNC diagnosis on REE remains uncertain. Heterogeneity in cancer stage, treatment, and REE assessment timelines between studies underscores the need for cautious interpretation and highlights the importance of individualised approaches in clinical assessment through monitoring and review. Use of indirect calorimetry in HNC research is needed to strengthen the evidence base, and where feasible, its use in clinical practice would ensure accuracy at the individual level.

Supplementary Information
Below is the link to the electronic supplementary material.