Carnitine Levels and the Ketogenic Diet


Revision accepted August 14, 2001. Address correspondence and reprint requests to Dr. E. Berry-Kravis at Rush-Presbyterian-St. Luke's Medical Center, 1725 West Harrison Street, Suite 718, Chicago, IL 60612, U.S.A. E-mail:


Summary:  Purpose: To determine the long-term effect of the ketogenic diet (KD) on carnitine levels and whether carnitine depletion is a significant cause of clinical complications during KD initiation or treatment.

Methods: Carnitine levels at 0, 1, 6, 12, and 24 months of diet treatment, carnitine antiepileptic drug (AED) history, lowest blood glucose and time to achieve ketosis during diet initiation, and diet complications were analyzed for 38 consecutive patients who initiated the KD from May 1997 to March 2000. Carnitine levels at follow-up were analyzed for eight patients started on the diet before to May 1997.

Results: Total carnitine (TC) at diet initiation correlated negatively with the number of AEDs at diet initiation but not with number of past AEDs, lowest blood glucose, or time to ketosis. TC decreased in the first months of diet treatment and then stabilized or increased slightly with long term treatment. Only 19% of patients were supplemented with carnitine for low TC. No patient showed clinical signs of carnitine deficiency.

Conclusions: Multiple AED exposure lowers TC, but actual TC deficiency in patients initiating the KD is not common, and TC levels do not appear to predict hypoglycemia or problems achieving ketosis. Mild carnitine depletion may occur early in KD treatment and occasionally TC decreases out of the normal range, without clinical symptoms. TC stabilizes or increases back toward baseline with long-term treatment, and most patients do not require carnitine supplementation.

The ketogenic diet (KD) is an extremely high-fat, low carbohydrate diet originally designed to mimic fasting, which produces seizure control in some individuals, probably related to mechanisms involving effects of acidosis and ketosis on neuronal membrane function and metabolism (for recent reviews and large studies, see REFERENCES 1–6). As during catabolic stress or starvation, patients treated with the KD rely heavily on metabolism of fatty acids for energy. Carnitine is responsible for transport of long chain fatty acids into the mitochondria prior to oxidation and for sequestration of medium and short chain fatty acids which accumulate in the mitochondria during catabolic stress (7–10). Thus problems related to carnitine metabolism may be important during either KD initiation or prolonged KD treatment.

The KD is typically initiated with a prolonged fast to induce ketosis rapidly. This is a metabolic stress and may bring out subtle impairments in fatty acid utilization or unmask asymptomatic carnitine deficiency. Patients initiating the KD may be at risk for low carnitine stores due to use of multiple anticonvulsants (11–13), often including valproic acid (VPA). It has been proposed that patients with carnitine insufficiency, reflected by low plasma total carnitine (TC) or increased acyl-to-free carnitine ratio (AFCR), may be at greater risk for symptomatic hypoglycemia during fasting for KD initiation (14). Lower TC might also promote difficulty getting into ketosis due to decreased processing of fatty acids to ketone bodies (15).

Prolonged KD treatment produces chronic ketosis with increased formation of acetylcarnitine and other acylcarnitines from free carnitine, leading to an expected increase in the plasma AFCR and a drop in plasma free carnitine. Although acetylcarnitine will be effectively reabsorbed in the renal tubules, longer-chain acylcarnitines may be poorly reabsorbed (8) potentially leading to increased net carnitine excretion and gradual carnitine depletion with long-term KD treatment. This, coupled with a tendency to poor carnitine intake on the KD and the importance of maintenance of carnitine stores for fatty acid utilization in KD-treated patients, would suggest that low carnitine levels or clinical carnitine deficiency may develop with long-term KD treatment.

Although there are reports of abnormal carnitine levels in patients beginning the diet and after carnitine supplementation in association with KD treatment (16,17), carnitine insufficiency in these studies was defined predominantly by abnormal AFCR, an expected biochemical effect of ketosis and therefore not a valid marker of carnitine deficiency in KD-treated patients. There is huge variability in clinical practice with regard to carnitine supplementation with the KD, as some neurologists supplement all patients on the KD, whereas others treat all patients with abnormal AFCR, some treat patients with low total carnitine (TC), and yet others treat no patients and do not monitor carnitine levels. No study has systematically tracked carnitine levels throughout diet treatment in unsupplemented patients, attempted to determine whether carnitine-depleted patients on the KD show clinical symptoms, or tracked the effects of carnitine supplementation on carnitine levels over time in KD-treated patients, despite a consensus that these studies are needed to guide clinical practice regarding carnitine supplementation in KD-treated patients (17).

In this study, we sought to help define interactions between the KD and carnitine and clarify indications for carnitine supplementation with the KD. We have attempted to determine whether patients initiating the diet at our center are at risk for significantly low TC due to multiple AED use, and whether plasma TC was predictive of hypoglycemia or problems achieving ketosis during diet initiation. Even more important, we systematically tracked the effect of long-term KD treatment on TC, AFCR, and potential symptoms of carnitine deficiency, and determined the effects of carnitine supplementation for low TC in KD-treated patients.


Information was collected and recorded in a database for all patients (group 1) admitted to the RUSH-Presbyterian-St. Luke's Medical Center for classic ketogenic diet initiation from May 1997 to March 2000, as part of our standard clinical protocol, including total previous and current AEDs, plasma carnitine levels obtained on admission to the hospital, lowest blood glucose measured during the fast, and time to achieve ketosis with urine ketones ≥80 mg/dl. The KD was initiated and maintained according to a standard protocol, similar to that described in Freeman et al (1). Patients were fasted for 24–72 h until they showed high levels of ketosis in urine (80–160 mg/dl), and then the diet was titrated to full strength over 3 days. All patients age 2 and older were started and maintained on a 4:1 diet and subsequently weaned to a 3.5:1 or 3:1 diet if seizures were well controlled and remained equivalently controlled after weaning, while patients less than 2 were started with a 3:1 diet and moved later to a 4:1 diet if tolerated and if seizures were not controlled on the 3:1 diet. Calories were individualized for each patient, based on a detailed 72-h food record. Prediet calories were replaced 100% for normal weight-for-height patients or adjusted upward or downward by 5% for patients with low or high weight-for-height to achieve weight-for-height normalization during diet therapy. Calories/kg generally fell within the ranges recommended in the Freeman protocol(l) All patients received vitamin/mineral supplements and calcium.

All patients with static encephalopathy or developmental disorder of uncertain etiology (in addition to their seizure disorder) had additional negative metabolic testing including at least chemistries, liver function tests, serum lactate, serum amino acids, and urine organic acids. For patients who stayed on the KD, repeated plasma carnitine determinations were performed and recorded at 1, 6, 12, 24, and 36 months. Occasional patients missed a 1 or 6-month determination because of noncompliance with the protocol. Carnitine levels also, were obtained from all patients (group 2) started on the KD before May 1997 and still on the KD when they came in for follow up.

Patients found to have TC deficiency at KD initiation or at any time during follow-up received carnitine supplementation (50 mg/kg/day). Total carnitine deficiency was defined as plasma TC <31 μM for males and <25 μM for females, according to standard reference laboratory “normal” ranges (31–79 μM for males and 25–69 μM for females; Metabolic Analysis Laboratories, Madison, WI, U.S.A.). Patients with only high AFCR and/or low free carnitine levels were not supplemented because this was interpreted as an expected biochemical change related to ketosis, and normative values are not available for these parameters in ketotic subjects. AFCR was not thought to be a useful parameter for monitoring KD-treated patients for carnitine deficiency because the ratio is more sensitive to the patient's degree of ketosis than to the level of carnitine stores. At each follow-up visit, patients were screened by history and physical examination for changes in baseline level of symptoms that might be relevant to carnitine deficiency (7–8,15,18–19), including weakness, fatigue, cognitive deterioration, decrease in tone, change in motor skills, lethargy, and signs of cardiac or liver dysfunction, with an intent to treat any patients with persistent symptoms in these areas even with normal plasma TC. Liver monitoring by SGPT (serum alanine aminotransferase) was done concurrently with carnitine determinations.

Thirty eight group 1 patients were admitted for KD initiation over the indicated time period and had carnitine monitoring according to the above-defined protocol. Two patients were taking carnitine at entry, and 36 had no history of carnitine treatment. Group 1 consisted of 18 female and 20 male patients with an average age of 7.8 ± 6.3 years (range, 1–24). Carnitine determinations were obtained for eight of the 19 group 2 patients (11 patients were either off the diet or not in follow-up). This patient group had been on the KD for 1 month to 6 years before to the first carnitine determination and consisted of three female and five male patients with an average age of 9.8 years (range, 3–21). One male patient was eliminated from analysis because he had been empirically treated with carnitine throughout KD treatment. No patient had a previous diagnosis of carnitine deficiency or any recent changes in symptoms suggestive of carnitine deficiency at KD initiation.


Carnitine and diet initiation

A summary of data from the 38 group 1 patients at diet initiation is presented in Table 1. TC averaged well within the normal range. Only three of 38 (one male, two female) patients had TC deficiency (range, 11.9–21.4 μM) at diet initiation, and all were treated shortly after starting the KD when their test results became available. These patients were untreated during fasting for KD initiation, but had no complications. They did not have symptomatic hypoglycemia and had blood sugar nadirs of 38, 55, and 60 mg/dl (mean, 51 mg/dl), a range similar to that seen in patients with normal TC (mean, 51.4 mg/dl). Only two of 38 patients had symptomatic hypoglycemia (29–30 mg/dl) requiring treatment because of lethargy, but these patients had TC values in the normal range (26 and 65 μM). In addition, one group 2 patient manifested symptomatic hypoglycemia (blood sugar of 27 mg/dl) at KD initiation but had a high normal TC when carnitine testing was subsequently done after several months of KD treatment.

Table 1.  Patient data from KD initiation
  1. All values are based on data derived from 38 patients, except when indicated otherwise. Normal values for carnitine are defined as 31–79 μM for males and 25–69 μM for females.

  2. AED, antiepileptic drug; KD, ketogenic diet; TC, total carnitine; VPA, valproic acid.

No. AEDs failed4.83.11–13
No. AEDs current1.81.10–4
TC (μM)50.416.811.9–79.8
 TC (VPA treated, n = 9)46.919.113.2–67.9
 TC (VPA untreated, n = 29)51.516.211.9–79.8
 TC (past VPA treated, n = 26)52.115.513.2–76.2
 TC (never VPA treated, n = 12)46.919.511.9–79.8
Minimal glucose (mg/dl)51.411.829–72
 Minimal glucose (+VPA, n = 9)54.17.938–63
 Minimal glucose (−VPA, n = 29)50.612.829–72
Time to ketosis (h)42.718.818–84

Age correlated positively with minimal glucose during the fast (r = 0.59, p < 0.01) such that, as expected, younger patients did become more hypoglycemic. There was a weaker correlation of age with time to ketosis (r = 0.37, p < 0.05) such that younger patients tended to become ketotic more quickly. Age did not correlate with TC.

TC showed a weak negative correlation with the number of AEDs at KD initiation (p < 0.05), but not with total number of AEDs ever used, suggesting that patients on more anticonvulsants run lower TC (Table 2). The relation between TC and VPA treatment at KD initiation was analyzed specifically because VPA is the AED most implicated in carnitine depletion. (11,12) Removal of the patients treated with valproic acid at KD initiation (N = 9) did not alter the negative correlation of number of AEDs with TC. Patients treated with VPA at KD initiation did not demonstrate significantly different average TC than did patients not treated with VPA (Table 1, p = 0.24). Patients with any previous or present exposure to VPA did not show average TC significantly different from those never treated with VPA (Table 1, p = 0.19). These data suggest that specific treatment with VPA was not the sole reason for the lower TC in patients treated with more AEDs. Patients treated with VPA at KD initiation did not have a lower average blood glucose nadir than patients not treated with VPA (Table 1, p = 0.22).

Table 2.  Correlation of carnitine measures with blood glucose nadir, time to ketosis, and anticonvulsant data at KD initiation
Time to
  • All values represent correlation coefficient (r).

  • a

     p < 0.05.

  • TC, total carnitine; KD, ketogenic diet; AED, antiepileptic drug.

TC (n = 38)0.270.04−0.33a−0.19
Change in TC at 1 mo
 (n = 24)

Total carnitine did not correlate with minimal blood glucose during the fast nor did TC correlate with time to ketosis (Table 2). It was noted from follow-up data (Fig. 1) that TC tends to decrease somewhat in the first month of KD treatment. It seemed possible that the magnitude of the decrease in carnitine level over the first month on the KD might be more sensitive than TC at KD initiation, as an indicator of baseline problems with carnitine stores. Ketosis might be more likely to give large decrements in plasma TC in patients with borderline stores at KD initiation. The change in TC over the first month of KD treatment (−11.0 ± 14.0 μM for patients, n = 24, who were not supplemented with carnitine, completed one month of treatment, and were compliant with obtaining the 1-month carnitine levels), however, did not correlate significantly with minimal glucose or time to ketosis, AEDs at diet initiation, or total AEDs ever used (Table 2).

Figure 1.

A: Averaged total carnitine (TC) over long-term ketogenic diet (KD) treatment in unsupplemented patients (○) and patients supplemented at 1 mo with supplementation discontinued at 12 mo (□). *p < 0.05, **p < 0.01, ***p < 0.0001 for difference from baseline TC at KD initiation (zero time point on graph). Error bars represent standard error of the mean. B: Plots of TC for individual unsupplemented patients treated for >1 year with the KD.

Carnitine and long-term diet treatment

Of the 38 group 1 patients, seven did not continue the KD past 1 month because they derived no control or their seizures worsened (including one of the patients with carnitine deficiency discovered at KD initiation), two did not continue the KD because they could not tolerate the diet (including one of the patients already taking carnitine at KD initiation), one was eliminated from analysis because he was already taking carnitine when he started the KD, and two were eliminated from the analysis because they were given carnitine when their KD initiation laboratory values showed low TC. Therefore serial carnitine determinations were analyzed for 26 initially untreated group 1 patients (13 male, 13 female) who were continued on the KD for >1 month. Of these patients, five (19%) ultimately received carnitine supplementation for low TC, three (all male, TC range, 24.6–26.9 μM) after 1 month of the KD, and two (one male, 1 female, TC range, 19–21 μM) after 6 months of KD treatment. One of the patients receiving carnitine supplementation after 6 months had renal tubule dysfunction (which preexisted KD treatment) and appeared to lose excess carnitine via renal excretion. All patients who reached 6 months of KD treatment with normal TC continued to show TC in the normal range throughout the rest of the course of KD treatment. In the group of 7 untreated group 2 patients, one male (17%) had a low TC of 21 μM when tested (after 1 year on KD) and was started on carnitine. Thus for all initially unsupplemented patients analyzed, the frequency of low TC leading to carnitine supplementation was 6 (18%) of 33. No patients, treated or untreated, had changes in clinical signs suggestive of carnitine deficiency (as described in Methods). No patients had a significant increase in SGPT or clinical signs of heart or liver dysfunction. All patients were in consistent ketosis throughout the time of carnitine determinations.

Averaged TC values in group 1 patients decreased in the first 1–6 months of KD treatment and then stabilized (Fig. 1A) The decrease in TC over first month on the KD in the group of 26 initially unsupplemented patients with serial carnitine determinations including a determination at 1 month (−11.0 ± 14.0 μM, n = 24) was similar (p = 0.92) in males (−9.9 ± 15.2 μM) and females (−12.5 ± 10.7 μM), although all three patients requiring supplementation at 1 month were male. Initial TC also did not differ (p = 0.5) between the males (52.6 ± 12.1 μM) and females (47.1 ± 13.0 μM). Males were supplemented more often because the lower limit of “normal” for males was higher than that for females. Decrease in TC was not significantly influenced by diet ratio (−15 μM in the four unsupplemented very young patients on the 3:1 diet).

Averaged TC for patients never supplemented with carnitine (n = 21) was significantly lower at 1 month than at baseline (40.5 ± 12.4 μM vs. 49.9 ± 10.9 μM, n = 21, p < 0.01), significantly lower at 6 months (33.2 ± 5.4 μM, n = 13, p < 0.05) than at 1 month, and significantly higher at 12 (42.4 ± 5.4 μM, n = 7, p < 0.05) and 24 (48.3 ± 7.8 μM, n = 4, p < 0.0001) months than at 6 months. Averaged TC at 24 months was not significantly different from that at baseline. To demonstrate that individual curves for TC over time on the KD generated the same pattern as the averaged data, individual curves for all seven patients completing KD treatment for 1–2 years without receiving carnitine supplementation are shown in Fig. 1B.

When patients with low TC at 1 month were treated, TC levels increased dramatically (Fig. 1A) although AFCR did not normalize (Figure 2). These patients were all eventually taken off the carnitine supplements, and TC remained in the normal range after discontinuance of supplementation in all cases. TC increased less for the patient with renal tubulopathy treated at 6 months, but supplementation was eventually discontinued at 1 year with TC stable at just above the lower limit of normal. The other patient treated at 6 months was taken off carnitine by the family after less than a month of treatment but had a stable mildly low level (19.2 μM at 6 months, 20 μM at 12 months) 6 months later.

Figure 2.

A: Averaged acyl/free carnitine ratio (AFCR) over long-term ketogenic diet (KD) treatment in unsupplemented patients (●) and patients supplemented at 1 mo with supplementation discontinued at 12 mo (▪). No values were significantly different from baseline AFCR at 1 mo of KD treatment (baseline AFCR for ketosis). Error bars represent standard error of the mean.

As expected, AFCR values were “abnormal” (>0.4) for all carnitine determinations obtained at any time from patients on the KD (range, 0.44–3.25), regardless of carnitine treatment. Although abnormally high because of ketosis, AFCR values were stable-to-decreasing over time from 1 month of KD treatment (Fig. 2) and were not seen to increase progressively, as might be expected with progressive carnitine depletion.

If the KD depletes carnitine over time, then TC would be expected to correlate negatively with the length of time on the KD when TC is measured. For all patients who were treated with the KD for ≥6 months and were never treated with carnitine (13 group 1 patients and six group 2 patients), there was no correlation between TC at longest KD exposure time (time range, 6 months to 6 years) and KD exposure time (Fig. 3). Rather than tending to decrease in patients treated with the KD for longer times, TC levels drifted up (r = 0.47; p < 0.05) at time periods > 6 months, and AFCR trended down (r = 0.42; p < 0.1).

Figure 3.

Total carnitine (TC) (A) and acyl/free carnitine ratio (AFCR) (B) values at maximal time of ketogenic diet (KD) treatment for patients treated with the KD for ≥6 mo.


Carnitine and ketogenic diet initiation

Total carnitine deficiency by plasma screening was not common in our population, despite exposure to multiple AEDs in this patient population. TC was lower for patients taking multiple AEDs, and this appeared to be a complicated metabolic effect of polypharmacy. Analysis was performed with the total number of all AEDs ever used (past “AED load”) in addition to to analyzing concurrent AEDs, because it is not clear whether depression of carnitine values seen with past AED treatment is reversible, and even if so, the time frame for normalization is not known. In this study, it appeared that the number of AEDs in use concurrent to KD initiation had a stronger effect on TC than did past “AED load”. Neither numbers of AEDs at KD initiation nor past “AED load” correlated with minimal glucose during the fast or time to ketosis, suggesting that despite the tendency to lower TC with polypharmacy, treatment with larger numbers of AEDs was not itself a risk factor for hypoglycemia during KD initiation.

Plasma TC in our patient group seemed unrelated to levels of hypoglycemia during fasting for diet initiation, in that neither TC nor the decrease in TC over the early months on the diet predicted lower blood glucose nadirs during the fast nor longer time to attain ketosis. Plasma carnitine levels may be an imperfect screen for carnitine depletion as muscle and liver concentrations are ∼70 and 40 times higher than plasma, respectively (9,15), and occasional patients with tissue deficiencies but normal plasma carnitine levels have been described (8,15). Muscle or liver levels might provide more accurate data on carnitine stores, but plasma TC is low in most patients with carnitine deficiency, and although there might be discrepancies for individual patients, over a large series of patients, one would expect plasma TC to correlate at least roughly with carnitine tissue stores. Plasma TC values also are the only approximation of carnitine stores reasonably obtainable for our patient population. This study did not produce any suggestion of correlation between plasma TC and problems with KD initiation on which to base a recommendation for studies with tissue levels.

It appears that carnitine screening is not expected to be useful for prediction of risk for hypoglycemia during KD initiation, although carnitine determinations are indicated for any patient with symptomatic hypoglycemia during KD initiation, if not already done. Carnitine screening is probably not necessary or helpful for patients with a well-defined nonmetabolic etiology for their epilepsy, good nutritional status, no obvious symptoms of carnitine deficiency, and no obvious risk factors for carnitine deficiency other than use of multiple AEDs (including VPA) for seizure control. Certainly carnitine screening, organic acids, and possibly an acyl-carnitine profile should be done prior to KD initiation in epileptic patients with potential symptoms of carnitine deficiency, and for patients for whom the etiology of their seizure disorder is undefined, especially if neurological or developmental abnormalities also are present. Indeed, several KD-treated patients have been described with elevation of liver functions (20) or unexplained encephalopathy (5,21). It is not known whether these cases represent true KD-related events, perhaps on a background of a subtle undiagnosed metabolic problem, but they underscore the need for careful evaluations for metabolic conditions including those associated with carnitine deficiency before initiating treatment with a diet expected to constitute a significant metabolic stress. These cases also underscore the need for further studies to characterize fully metabolic changes that occur during KD initiation and treatment, to differentiate ominous and clinically significant findings that require treatment from normal biochemical adaptations to the KD.

Carnitine and long-term ketogenic diet treatment

TC decreases over the first months of KD treatment for most patients, and in some cases, 18% in our series, dips into the “carnitine deficient” range. After the first months of diet treatment, TC stabilizes or normalizes, with no evidence of progressive decline with increasing time on the KD. If TC does not decrease into the subnormal range in the first 1 to 6 months of KD treatment, it appears to remain in the normal range over long-term treatment with the KD and even may show a tendency to drift back gradually toward baseline levels with very prolonged KD treatment.

The patients who developed low TC which normalized with treatment were all KD responders (two 50–90% seizure reduction and three >90% reduction). These patients did not experience a worsening of seizures when TC was low suggesting seizure control was not dependent on normal TC. Even in patients who did not develop low TC, seizure control was not less at the TC nadir at one to six months. Further, TC at diet initiation was not significantly different in good KD (>90% reduction) responders and poor (<50% reduction) responders, at 49.1 and 62.1μM, respectively. One of the 2 patients who did not tolerate the diet did have low TC but the effects of supplementation on diet tolerability could not be assessed because of intractable dehydration, vomiting, and constipation in this multiply-impaired child. Our study suggests that seizure control on the KD is not strongly related to TC, but does not rule out the possibility that carnitine treatment may allow KD-intolerant patients to stay on the diet.

The AFCR is not expected to be a useful measure for assessing carnitine insufficiency at any point in the course of KD treatment because, as seen in the patients in this study, it will always be elevated even with carnitine supplementation, and norms do not exist for subjects at various levels of ketosis. Increased AFCR is a normal biochemical adaptation to the diet and the degree of elevation of AFCR is more likely to be dependent on the level of ketosis for a given patient at a given time than anything to do with carnitine stores, as elevations in AFCR decrease rapidly to normal when fasted subjects are refed (22–24). AFCR may decrease (normalize) somewhat with increasing time on the KD, possibly related to cellular adaptations to ketosis.

We chose to supplement patients developing low TC while on the KD because carnitine treatment is quite benign, but in fact, these patients did not have any change in their condition suggestive of clinical manifestations of carnitine deficiency. In all of these patients, TC levels dipped just below the lower limit of “normal TC.” Their TC increased rapidly and dramatically with supplementation and then stabilized. Even without treatment their TC levels might have been stable or even “self-corrected” back into the normal range without ever resulting in symptoms of carnitine deficiency. Thus, for patients on the KD without clinical symptoms of carnitine deficiency, perhaps our criteria for supplementing was too stringent. The lower limit of normal TC has been empirically established, and perhaps the 18–20 μM range would be a better lower limit. Thus carnitine supplementation may be necessary only very rarely in KD-treated patients. Indeed, clinical symptoms of carnitine deficiency on the KD appear to be very infrequent, were not seen in our series, and may arise almost exclusively in the setting of an undiagnosed preexisting metabolic condition.

It appears that the optimal time to check carnitine levels in KD-treated patients is at 1 to 6 months of treatment, especially for patients with low-normal TC at KD initiation. After this time frame, our study provides no evidence to suggest ongoing depletion of carnitine stores or clinically relevant carnitine deficiency will appear with longer-term KD treatment. Most patients do not develop low TC or appear to need supplementation. Clinically reasonable approaches to the subgroup developing asymptomatic low TC would be either to supplement several months until TC returns to the normal range or to follow levels and clinical status and treat for ongoing decline in TC or suggestive symptoms.