Stroke outcome in the ketogenic state – a systematic review of the animal data


Address correspondence and reprint requests to Sean Murphy, Department of Neurological Surgery, Box 356470, University of Washington School of Medicine, Seattle, WA 98195, USA.


As a predictor of potential clinical outcome, we performed a systematic review of controlled studies that assessed experimental stroke outcome in rodents maintained on special diets (calorie restriction and ketogenic diet) or following the direct administration of ketone bodies. Pre-clinical studies were identified by searching web databases and the reference lists of relevant original articles and reviews. Sixteen published studies (a total of 733 experimental animals) met specific criteria and were analyzed using Cochrane Review Manager software. This resulted in objective evidence to suggest beneficial effects of the ketogenic pathway on pathological and functional outcomes following experimental stroke.

For decades, we have known that caloric restriction and intermittent fasting extend the lifespan of rodent species (and probably primates) by reducing free radical production, inflammation, and/or increasing resistance to stress. Such strategies would, therefore, be predicted to benefit both cardio- and cerebro-vasculature (Mattson and Wan 2005; Yamada 2008). Other dietary regimens can also be beneficial to neurological function. For instance, the ketogenic diet, in which fat replaces carbohydrate, is clinically useful in the treatment of epilepsy, and there is growing interest in extending use of this diet to other neurological disorders (, including acute injury such as stroke and trauma (Gasior et al. 2006). To provide a prediction of potential clinical outcome, we performed a systematic review and meta-analysis of controlled rodent studies that assessed stroke outcome where the use of ketone bodies as an energy source had been promoted. These include ketogenic diets, as well as calorie restriction, which promote the use of triacylglycerols and fatty acids. In addition, ketone bodies can be directly administered. All three dietary interventions are evaluated here.


Study identification

Experimental animal studies assessing the effect of dietary interventions related to ketone body utilization on outcomes following experimental stroke were identified from Embase, PubMed, Web of Science, and Google Scholar by searching for all published articles up to the end of April 2012. The earliest study for analysis was that of Go et al. (1988). Search keywords included combinations of caloric restriction, cerebral ischemia, ketogenic diet, and rodent. Additional publications were identified from reference lists of all identified articles and narrative reviews. Pre-specified exclusion criteria were used to aid selection and prevent bias, and studies were only included if (i) experimental ischemia was induced, (ii) animals were exposed to a ketogenic state, (iii) there was a control group, and (iv) an appropriate outcome was measured following stroke.

Data extraction

The term ‘intervention’ refers to any intervention that induces a ketogenic state in animals such that the brain is deprived of its usual energy source, that is, glucose, and instead uses ketones as an alternative. A ketogenic state, experimentally, can be induced by (i) restriction of caloric intake, (ii) ketogenic diet, or (iii) exogenous administration of ketones. From the relevant studies, data were extracted on animal species, number, type of brain injury, type of intervention, timing of intervention relative to onset of experimental injury, and outcome post-stroke. Outcomes included lesion volume, brain water content, neuron counts, survival rate, and functional measures. Data for functional outcomes included the following: (i) open field activity (distance traveled), (ii) 8-arm radial maze (working memory errors), and (iii) novel object recognition task (discrimination index = time spent exploring novel object/total exploration time).

A comparison (C) was defined as the assessment of outcome in intervention and control groups, whereby intervention constituted either a dietary intervention (caloric restriction and ketogenic diet) or administration of a ketone body at a stated time point relative to the induction of ischemia. For each comparison, data were extracted for mean outcome, standard deviation, and the number of animals per group. Data were not extracted if mean values were not reported, that is, if only median and confidence intervals were given. If published studies (S) used multiple groups, for example to assess dose-response relationships, then data were individually extracted. If numerical data were not reported in the text, they were extracted from enlarged versions of the graphs. The methodological quality of each study was determined by minor modification to the ten-point scale of O'Collins et al. (2012).

Data analysis

The data were analyzed using Cochrane Review Manager (version 5;, as in a previous animal meta-analysis (White and Murphy 2010). The effect of ketogenic state, as compared with control, on post-stroke outcomes was assessed using the standardized mean difference, whereby the difference in effect between intervention and control treatment is divided by the total standard deviation. A standardized mean difference of zero represents a lack of intervention effect, whereas a positive value indicates a beneficial effect, and a negative value demonstrates a detrimental effect of the intervention. This allows comparisons to be made even though different methods of measurement and/or different animal models are used. Statistical heterogeneity was accounted for through the use of the DerSimonian and Laird (1986) pooling model of random effects. The data were grouped and stratified meta-analyses based on (i) type of intervention used to induce ketogenic state, (ii) quality score, (iii) type of outcome assessed following ischemia, (iv) quality score, and (iv) duration (chronic or acute) of intervention. For those tests in which an increase in nominal value represented an improvement in outcome (e.g., neuron counts following FluoroJade labeling) the inverse of the extracted data was used for data comparisons. Studies were weighted by sample size, and the results are expressed as standardized mean difference with 95% confidence intervals. The significance was set at α = 0.05 and p values of < 0.05 from meta-analyses were considered to be significant.


Design of studies

On the basis of the stated search criteria, we identified 19 studies that investigated the effect of a ketogenic state on outcome following cerebral ischemia. However, three of these (Combs and D'Alecy 1987; Marie et al. 1990; Chiba and Ezaki 2010) were excluded, as mean values and the distribution of data (standard error or standard deviation) were not reported. The main characteristics of the 16 included studies are reported in Table 1. All of these reported the effect of inducing a ketogenic state, versus control, on one or more outcomes following cerebral ischemia. The 16 included studies represent published data from 12 research groups; four research groups published two studies each with the remaining eight studies coming from separate research groups. Data from a total of 733 experimental subjects were included for analysis.

Table 1. Characteristics of the studies included in the review
StudyParameters assessedInterventionFirst dose timingaSpeciesModel of ischemia
  1. a

    First dose timing in relation to onset of experimental stroke.

  2. MCAO, middle cerebral artery occlusion; VO, vessel occlusion; SD, Sprague–Dawley rat; LE, Long Evans rat; NOR, novel object recognition.

Arumugam et al. (2010)Lesion volume, neurological scoreCalorie restriction −3 monthsC57 mouseMCAO
Bobyn et al. (2005)Open field, neuronal viabilityCalorie restriction−28 daysGerbil2VO
Go et al. (1988)Survival, edemaCalorie restriction−4 daysRatGlobal
Gueldry et al. (1990)EdemaKetone administration−0.5 hSD ratMicrospheres
Massieu et al. (2003)Lesion volumeKetone administration−14 days, −7 days, 0 hWistar ratIodoacetate
McEwen and Paterson (2010)Open field, neuronal countCalorie restriction+1 hGerbil2VO
Puchowicz et al. (2008)Lesion volume, edemaKetogenic diet−21 daysWistar ratMCAO
Roberge et al. (2008a)Radial maze, neuronal countCalorie restriction−3 monthsWistar rat4VO
Roberge et al. (2008a)Elevated maze, open field, neuronal count Calorie restriction−3 monthsWistar rat4VO
Robertson et al. (1992)Lesion volumeCalorie restriction, ketone administration−12 hLE rat2VO
Suzuki et al. (2001)Survival, edemaKetone administration−0.5 hddY mouseGlobal
Suzuki et al. (2002)Lesion volume, edemaKetone administration+0.5 hWistar ratMCAO
Tai et al. (2008)Neuronal countKetogenic diet−25 daysSD rat4VO
Xu et al. (2010)NORKetogenic diet−21 daysF344 ratHypoxia
Yoon et al. (2011)Lesion volumeCalorie restriction−3 monthsC57 mouseMCAO
Yu and Mattson (1999)Lesion volume, neurological scoreCalorie restriction−3 monthsSD ratMCAO

In terms of the type of intervention used to induce a ketogenic state, three studies utilized a ketogenic diet (Puchowicz et al. 2008; Tai et al. 2008; Xu et al. 2010), five studies exogenously administered ketone bodies (Gueldry et al. 1990; Robertson et al. 1992; Suzuki et al. 2001, 2002; Massieu et al. 2003), and the remaining eight studies along with the study by Robertson et al. 1992 used calorie restriction. Of the five studies that exogenously administered ketone bodies, two administered β-hydroxybutyrate (Suzuki et al. 2001, 2002), two administered 1,3 butanediol (Gueldry et al. 1990; Robertson et al. 1992), one administered acetoacetate (Massieu et al. 2003), and the study by Robertson et al. (1992) also administered triacetin. When applying an intervention to induce a ketogenic state, the majority of studies did so prior to the onset of cerebral ischemia. Ten of the included studies introduced the intervention at least 7 days prior to the onset of cerebral ischemia, whereas four studies (Go et al. 1988; Gueldry et al. 1990; Robertson et al. 1992; Suzuki et al. 2001) introduced the intervention less than 7 days prior to the onset of cerebral ischemia. The remaining two studies (Suzuki et al. 2002; McEwen and Paterson 2010), along with the study by Massieu et al. (2003), examined the effects of inducing a ketogenic state following the induction of ischemia.

To assess the effects of a ketogenic state on outcome following cerebral ischemia, the majority of studies utilized various models to induce cerebral ischemia; thirteen used vessel occlusion (filament occlusion of the middle cerebral artery, or clip occlusion of a number of vessels including the middle cerebral artery), one study used microspheres (Gueldry et al. 1990), one study used a model of glutamate excitotoxicity via application of iodoacetate (Massieu et al. 2003), and one study used a hypoxic chamber (Xu et al. 2010). Male rat models (F344, Long Evans, Sprague-Dawley, Wistar) were used in 12 of the 16 studies; the remaining four used either a mouse strain (Suzuki et al. 2001; Arumugam et al. 2010; Yoon et al. 2011) or gerbils (McEwen and Paterson 2010).

In terms of assessing outcome following cerebral ischemia, the majority of studies used lesion volume (see Table 1), five used % water content (Go et al. 1988; Gueldry et al. 1990; Suzuki et al. 2001, 2002; Puchowicz et al. 2008), three used neuron count (Roberge et al. 2008a, b; McEwen and Paterson 2010), two studies used survival rate (Go et al. 1988; Suzuki et al. 2001), and one study used neurological score (Arumugam et al. 2010). A number of studies also used measures of functional outcome, including the elevated maze (Roberge et al. 2008b), open field activity (Bobyn et al. 2005; McEwen and Paterson 2010), radial maze (Roberge et al. 2008a), and novel object recognition (Xu et al. 2010).

Overall efficacy of inducing a ketogenic state

We determined initially that there was an overall significant protective effect of a ketogenic state on outcome following cerebral ischemia (1.24, 1.55–0.93, p < 0.001), regardless of individual study characteristics. Our literature search indicated that studies tended to induce a ketogenic state either by intervention (calorie restriction or a ketogenic diet) or through the exogenous administration of ketones. Thus, we also determined if a beneficial effect was seen following either intervention (1.12, 1.55–0.69, p < 0.001) or exogenous ketone administration (1.42, 1.84–1.0, p < 0.001; Fig. 1).

Figure 1.

Standardized mean difference and 95% CI of all outcome measures by type of intervention, that is, all, dietary restriction only, and exogenous administration of ketones. A beneficial effect following any type of intervention was observed. N, number of animals; C, number of comparisons; S, number of studies.

Type of parameter used to assess outcome

The effect of intervention on cerebral ischemia outcome was analyzed according to type (Fig 2). First, data were grouped according to intervention, that is, all, intervention, or exogenous administration, and then analyzed to determine if they had a significant beneficial effect on either pathology or function (Fig. 2a). Pathological outcomes included lesion volume, brain water content, and neuronal counts, whereas functional outcomes included all measures of behavior. Regardless of whether a dietary intervention or administration had been used to induce a ketogenic state, a significant beneficial effect was seen on both outcomes (< 0.01). However, a variety of outcomes were used in the included studies, and we went on to analyze each separately (Fig. 2b). Induction of a ketogenic state was found to have a beneficial effect on functional tests (1.46, 2.19–0.73 < 0.001), lesion volume (1.29, 1.7–0.88, < 0.001), and brain water content (1.04, 1.65–0.43 < 0.001). However, a ketogenic state did not have a beneficial effect on neuron count, as assessed by FluoroJade labeling (−0.03, 0.39–−0.44, = 0.9), which included data from three studies, four comparisons, and 91 experimental subjects.

Figure 2.

Standardized mean difference and 95% CI of all outcome measures by type of outcome following cerebral ischemia. A beneficial effect of any type of intervention was observed when assessing pathological or functional outcomes (a). When data were analyzed according to the different types of outcome (b), that is, function, edema, lesion volume, and cell viability, a beneficial effect of a ketogenic state was observed for all outcomes, apart from the last. N, number of animals; C, number of comparisons; S, number of studies.

Reported study quality

The quality scores of the included studies ranged from 1 (low) to 5 with a median score of 3. To determine whether the quality score of an individual study had any impact on whether the intervention had a beneficial effect following cerebral ischemia, included studies were analyzed according to quality score (Fig. 3). Only those with a score of 2, 3, or 5 demonstrated a significant (< 0.01) beneficial effect of the intervention. Studies that had the lowest quality score had no significant beneficial effect (= 0.07). In addition, three studies received a quality score of 4 and these showed no beneficial effect (= 0.08).

Figure 3.

Standardized mean difference and 95% CI for all outcome measures when data are grouped according to study quality. A beneficial effect was only observed in those studies that received a quality score of 2, 3, or 5. N, number of animals; C, number of comparisons; S, number of studies.

Duration of intervention

When comparing data across all studies, it was determined that there was a beneficial effect of intervention to induce a ketogenic state, regardless of whether this occurred pre- or post- experimental ischemia (Fig. 4). A beneficial effect was observed, when the intervention began more than 7 days prior to the onset of experimental ischemia (1.44, 1.88–1.0 < 0.001) or less than 7 days prior to the onset of experimental ischemia (1.42, 1.95–0.89 < 0.001). In addition, a significant beneficial effect was observed if the intervention began following the onset of experimental ischemia (0.82, 1.38–0.26 = 0.004).

Figure 4.

Standardized mean difference and 95% CI for all outcome measures when data are grouped according to onset of intervention. A beneficial effect was observed regardless of whether the ketogenic state was induced more than 7 days prior to onset, less than 7 days prior to onset, or at a time point following onset of experimental ischemia. N, number of animals; C, number of comparisons; S, number of studies.


To restate the major findings from our analyses, we found beneficial effects on pathological and functional outcomes of dietary intervention, or exogenous ketone administration, either prior to or following experimental stroke.

Neuropathologies, such as a stroke, cause a mismatch between energy demand and supply: blood flow goes awry, oxygen levels fall, and mitochondria malfunction. A period of fasting results in a short-term ketosis, and increased reliance on ketone bodies appears to be a form of cerebral metabolic adaptation (Manzanero et al. 2011). Ketone metabolism is enzymatically simpler and more efficient than glucose or pyruvate metabolism (Veech 2004) and is reported to increase global cerebral blood flow (Gasior et al. 2006; Prins 2008). Indeed, ketone bodies are the only circulating substrates in addition to glucose known to contribute significantly to cerebral metabolism. The precise mechanisms whereby caloric restriction, the ketogenic diet, and ketone bodies provide protection in ischemic stroke remain unclear, but there is notable improvement in mitochondrial function, a decrease in inflammation, and an increase in expression of neurotrophins such as BDNF and bFGF (Maalouf et al. 2009; Manzanero et al. 2011). Whether this relates to the higher level of potential energy available in the C–H bonds of beta hydroxybutyrate compared with pyruvate remains unclear but potentially important to recovering ATP levels during reperfusion (Veech 2004). More recently, the sirtuins have also been implicated. These histone deacetylases localize to different subcellular compartments and have a variety of substrates. Activity depends on nicotinamide adenine dinucleotide (NAD+), which links this family of enzymes to cellular energy levels. Indeed, some sirtuins (SIRT3) are located within mitochondria. Caloric restriction could activate and/or increase expression of sirtuins, which then modulate proteins involved in cell survival and apoptotic cell death (Maalouf et al. 2009; Morris et al. 2011).

Caloric restriction and the ketogenic diet appear to represent cost effective and efficient strategies through which stroke incidence and/or subsequent pathology could be reduced.

Conflicts of interest

The authors declare that they have no conflicts of interest. Support for this study is provided by American Diabetes Association and NIH-NIDDK (A.N.M.), American Heart Association and NIH-NINDS (S.P.M.).