General Anesthesia and the Ketogenic Diet: Clinical Experience in Nine Patients


Address correspondence and reprint requests to Dr. E.A. Thiele at Pediatric Epilepsy Program, Massachusetts General Hospital, VBR 830, 55 Fruit Street, Boston, MA 02114, U.S.A. This work was presented in abstract form at the Annual meeting of the American Epilepsy Society, Los Angeles, California, and December 1, 2000.


Summary:  Purpose: To determine if children actively on the ketogenic diet (KD) can safely undergo general anesthesia (GA) for surgical procedures.

Methods: The records of children treated with the KD at Children's Hospital (Boston, Massachusetts) from 1995 to the present were reviewed. The charts of children who had received GA while on the diet were evaluated with regard to demographics, procedure information, anesthesia records, blood chemistries, and perioperative course. Of 71 children on the KD during the period of the study, nine (12.7%) had procedures requiring GA while on the diet.

Results: Nine children received GA for surgical procedures ranging from central line placement to hemispherectomy while on the KD. At the time of GA, the children ranged from age 1 to 6 years, and had been on the KD for 2–60 months. The patients received carbohydrate-free intravenous solutions perioperatively. Anesthesia duration ranged from 20 min to 11.5 h; for longer procedures, serum pH, glucose, and electrolyte levels were monitored. Serum glucose levels remained stable in all patients, but serum pH typically decreased; the largest reduction was to 7.16. In three procedures, patients received intravenous bicarbonate because of level of acidosis. There were no perioperative complications.

Conclusions: Children on the KD can safely undergo GA for surgical procedures. Although serum glucose levels appear to remain stable, serum pH or bicarbonate levels should be monitored because of the risk of metabolic acidosis.

The ketogenic diet (KD) has been a nonpharmacologic treatment option for patients with intractable epilepsy since the 1920s. The diet consists of high fat with low protein–carbohydrate content, which induces an increase in the ketone bodies and produces a ketotic state (1,2). Most children started on the diet have medically intractable seizures, and many also have medical issues that may require procedures with general anesthesia (GA). Hinton et al. (3) described three children on the KD that underwent inhalation anesthesia for minor surgery without complications. The anesthesia time was <35 min in all the patients; the preoperative fasting period was 10 h. McNeely (4) reported one adolescent on the KD who had a procedure requiring anesthesia for 5 h and 20 min. However, the literature on the safety of undergoing GA while on the KD is scarce, and there is no consensus on how to proceed when children are on the diet and require GA procedures. Some centers taper or discontinue the KD 1 week or even the day before the surgery, whereas others continue the diet without a specific protocol (personal communications).

This retrospective study was undertaken to define the safety and the possible risks associated with surgical procedures and GA in children on the KD.


The records of children treated with the KD through the Comprehensive Pediatric Epilepsy Program at Children's Hospital, Boston, from 1995 through November 2000 were reviewed retrospectively. Of a total of 71 children, nine (12.7%) had procedures requiring GA while on the KD. A total of 24 procedures was identified, as many of the children had GA for more than one procedure.

The individual records were then analyzed, and data obtained from inpatient hospital records as well as the computerized results reporting system. Each patient's clinical history was reviewed, including demographic and procedure information, anesthesia records, blood chemistries, and perioperative course. Patient information included the age at the time of surgery, seizure type, length of time on the KD, parameters of the diet at time of GA, and blood chemistries including serum β-hydroxybutyrate levels. Procedural information included type of GA, type of surgery, length of surgery and anesthesia times, intraoperative blood chemistry monitoring and intervention, and perioperative course.


Patient information

Nine children had surgical procedures requiring GA while on the KD during the period of this study. Five of the children were boys, four were girls, and their ages ranged from 1 to 6 years at the time of GA. All children were on the KD because of medically intractable epilepsy and had been taking an average of five (range, three to nine) anticonvulsant medications (ACDs) before initiating the diet. Seizure onset was at younger than 1 year in eight of the children; one of the children had seizure onset at age 4 years. Four of the children had infantile spasms; two, generalized tonic–clonic seizures; one, atonic seizures; one, simple and complex partial seizures; and one had a mixed seizure disorder including complex partial, myoclonic, atypical absence, and generalized tonic seizures.

At the time of GA, the children had been on the KD for an average of 21 months (range, 2–60 months). All children were on a classic KD; five were on a 4:1 ratio (g fat/ gm carbohydrate + protein); two, on a 3.5:1 KD ratio; one, on 3:1 KD ratio; and one, on a 2.5:1 ratio. Some of the children were taking medications known to affect acid–base balance: three children were taking Bicitra (sodium citrate and citric acid), and five were taking topiramate (TPM); one child was taking both. (A complete list of the medications that each child was taking at the time of GA is found in Table 1.) The preoperative serum β-hydroxybutyrate levels ranged from 3,370 to 9,418 μM (normal range, 23–211 μM), demonstrating that all were in ketosis. The KD had proved to be efficacious in controlling the seizure activity in the majority of this group of patients. Three patients had surgical procedures to improve seizure control; therefore they were not seizure free.

Table 1.  Anticonvulsant medications at the time of GA
  1. GA, general anesthesia.

1Clonazepam, vigabatrin
2Clonazepam, phenobarbital
3Lorazepam, phenytoin, topiramate, vigabatrin
5Phenobarbital, topiramate, valproate
6Diazepam, lamotrigine, lorazepam, topiramate
8Phenytoin, topiramate
9Clobazam, clonazepam, phenobarbital, topiramate, valproate

All children continued on the KD until made NPO (nothing per oral) for the procedure. All children were restarted on the diet in the postoperative period. The diet was continued even after hemispherectomy and other neurosurgical procedures, because it was thought that an abrupt change to a regular diet would oppose a greater metabolic risk in these patients. In the patient who had the hemispherectomy, the diet was modified, increasing the amount of calories and protein with the thought that it would promote healing and recovery.

The time at which the diet was restarted varied in each procedure, based on a case-by-case evaluation. Patients were initially given clear carbohydrate-free fluids, and then advanced to ketogenic eggnog or formula. If any undesirable reaction (e.g., significant nausea) occurred, the diet was tried at a later time. The diet was typically started with one third to one half–strength diet and slowly advanced to appropriate full diet.

General anesthesia procedures

Table 2 lists the surgical procedures, anesthesia times, and the number of children that had each procedure. In nine of the 24 procedures, the anesthesia time was ≥3 h. The maximal anesthesia time was 11.5 h for a left frontal cortical resection. Thirteen procedures lasted between 1 and 3 h; two procedures were <1 h. In addition to listed anesthesia times, most of the patients were without food (NPO) for ≥8 h before the procedure. Therefore the children were without caloric intake including carbohydrates for ∼9–20 h. For procedures lasting >3 h, blood chemistries including pH, glucose, and acid–base balance were monitored (Figs. 1 and 2).

Table 2.  Types of surgery and anesthesia times
Type of surgeryNumber of proceduresAnesthesia time (h:min)
  1. Number of procedures represent how many children in the group had the procedure.

MRT (magnetic resonance therapy–open magnet) guided resection of left frontal lesion111:28
Right hemispherectomy110:05
Right temporal lobectomy and corpus callosotomy18:00
Distal femoral varus osteotomy17:40
Placement of girds and strips for invasive epilepsy monitoring17:00
Nissen fundoplication16:10
Left frontal cranial bone flap revision13:55
Vagal nerve stimulator and pulse-generator placement13:00
Ventriculoperitoneal shunt insertion13:00
Right occipital ventricular catheter12:00
Placement of subclavian double-lumen central venous line21:10–2:20
Esophagogastroduodenoscopy with biopsy31:15–2:16
Closed reduction femur fracture21:42–2:12
Right hydrocelectomy11:20
Percutaneous endoscopy gastrostomy placement30:55–1:50
Bilateral myringotomy and tympanotomy tube placement30:20–2:35
Total procedures24 
Average anesthesia time (h:min) 3:30
Median anesthesia time (h:min) 2:14
Figure 1.

Perioperative glucose variation. Numbers at the bottom represent individual patients, and letters represent individual procedures. Some of the values were obtained after anesthesia during the postoperative period. Shown are the values for procedures >3 h. No data were available for procedures <3 h.

Figure 2.

Perioperative pH variation. Numbers at the bottom represent individual patients, and letters represent individual procedures. Bicarbonate was administered intravenously in three procedures (arrows) based on pH values. Some of the values were obtained after anesthesia during the postoperative period. Shown are the values for procedures >3 h. No data were available for procedures <3 h.

The anesthetic agents used varied among the procedures. In 22 procedures, general endotracheal anesthesia was used; two used mask anesthesia without intubation. Medications used included fentanyl, halothane, isoflurane, sevoflurane, nitrous oxide, propofol, thiopental, and ketamine. Table 3 lists the medications, doses, and intravenous fluids used during the procedures.

Table 3.  Intraoperative medications, total doses, and intravenous fluids used
Patient/procedureMedication (dose)
  1. NO, nitrous oxide; ISO, isoflurane; SEVO, sevoflurane; TS, thiopental sodium; P, pancuronium bromide; N, neostigmine bromide; S, normal saline; RL, Ringer's lactate.

1ANO, ISO, halothane, fentanyl (16.6 μg/K), morphine (0.1 mg/K), P (0.25 mg/K), N (0.06 mg/K), S 1,509 cc/m2, albumin 226 cc/m2
1BNO, ISO, TS (6.25 mg/K), vecuronium (0.16 mg/K), N (0.05 mg/K), S 377 cc/m2
1CNO, ISO, TS (5 mg/K), fentanyl (25 μg/K), P (0.45 mg/K), S 566 cc/m2, albumin 283 cc/m2
1DNO, ISO, TS (10.4 mg/K), fentanyl (56.2 μg/K), midazolam (0.3 mg/K), P (1 mg/K), P (0.08 mg/K), N (1 mg/K), S 2,641 cc/m2
1ENO, ISO, TS (6.25 mg/K), P (0.08 mg/K), N (0.06 mg/K), S 377 cc/m2, RL 396 cc/m2
1FNO, ISO, TS (4.1 mg/K), fentanyl (3.75 μg/K), P (0.16 mg/K), N (0.1 mg/K), RL 566 cc/m2
2ANO, ISO, TS (6.25 mg/K), propofol (7.5 mg/K), fentanyl (53 μg/K), P (1.9 mg/K), S 3070 cc/m2, albumin 175 cc/m2
2BNO, ISO, SEVO, P (0.4 mg/K), RL 757 cc/m2
3ANO, ISO, SEVO, S 545 cc/m2
4ANO, ketamine (1.6 mg/K), S 181 cc/m2
4BNO, ISO, SEVO, fentanyl (12.4 μg/K), P (0.3 mg/K), S 2,377 cc/m2, RL 454 cc/m2
4CISO, SEVO, P (0.13 mg/K), RL 82 cc/m2
4DNO, ISO, SEVO, propofol (2.5 mg/K), morphine (0.05 mg/K), mivacurium (0.15 mg/K), RL 253 cc/m2
4ENO, ISO, SEVO, S 443 cc/m2
5ANO, ISO, morphine (0.07 mg/K), P (0.13 mg/K), N (0.06 mg/K), RL 434 cc/m2
5BNO, ISO, SEVO, S 71 cc/m2
5CNO, ISO, TS (6.7 mg/K), P (0.25 mg/K), N (0.04 mg/K), succinylcholine (1.6 mg/K), S 1700 cc/m2
6ANO, TS (5 mg/K), fentanyl (2 μg/K), S 217 cc/m2
6BISO, TS (7.5 mg/K), P (0.1 mg/K), N (0.07 ng/K), RL 217 cc/m2
6CNO, ISO, SEVO, TS (4 mg/K), S 108 cc/m2
6DNO, propofol (3.6 mg/K), fentanyl (1.8 μg/K), RL 345 cc/m2
7ANO, SEVO, S 56 cc/m2
8ANO, ISO, SEVO, P (0.1 mg/K), RL 151 cc/m2
9ANO, ISO, TS (8.1 mg/k), fentanyl (2.3 μg/K), P (0.14 mg/K), N (0.06 mg/K), S 588 cc/m2

Effect on blood chemistries

During the perioperative period (while NPO before the procedure if intravenous fluids administered plus anesthesia time and postoperative period until restarted on diet) only carbohydrate-free solutions were used in all cases, such as normal saline or Ringer's lactate. Normal saline was used in 13 procedures, Ringer's lactate in eight and both saline and lactate in three procedures. Blood products were administered in three cases; packed red blood cells in two procedures, and whole blood in one.

Glucose levels remained stable even in prolonged procedures (Fig. 1). In four of the procedures, glucose levels obtained over 9 h into anesthesia (and therefore >17 h after holding caloric intake) remained similar to baseline. During one of the prolonged procedures, hemispherectomy, the child received steroids and 4 units of packed red blood cells intraoperatively, possibly explaining the increase in glucose levels.

However, there was a tendency for the children to develop metabolic acidosis intraoperatively during the longer procedures (Fig. 2). Acid–base balance was typically monitored intraoperatively by blood gas and pH; during eight procedures, bicarbonate levels were also measured. Similar to pH, bicarbonate levels tended to decrease during the longer anesthesia times, although the effect appeared less dramatic. Because of the change in pH, bicarbonate was given intravenously (5–20 mEq per dose) during three of the longer procedures and was administered twice during the procedure in two of these patients. During the longest procedure (Fig. 2, 1D) the patient initially received extra fluids intraoperatively because of the decrease in pH; only after bicarbonate administration did pH normalize.

Postoperative course

All children were restarted on the KD in the postoperative period. No perioperative or postoperative complications occurred in any of the procedures. The children appeared to recover from anesthesia and the surgical procedures at a usual rate. Even though a formal count of seizures was not performed during this study, none of the patients was noted to have increased seizures in the perioperative report or medical record, and none of the patients required additional ACD therapy perioperatively.


This retrospective review shows that children on the KD can safely undergo GA and surgical procedures while remaining on the KD. We identified nine children who, as a group, had GA for 24 surgical procedures. Before and during GA, the children received carbohydrate-free solutions, and all were restarted on the KD in the postoperative period. None of the children experienced any complications from the GA or the procedures; none of the children experienced an exacerbation of the seizure activity perioperatively.

Two other reports in the literature have described children on the KD having GA for surgical procedures (3,4). However, the two reports include a total of four patients, and three of the procedures used short-inhalation anesthesia. As seen in our study, glucose levels remain stable during the procedures, and no complications were reported in the two studies.

Although one would predict a decrease in serum glucose during a several-hour period without caloric intake, all of the children maintained stable glucose levels. This suggests that glucose metabolism is differently regulated in children who are in a ketotic state. Although it is not well studied in humans, studies in animals have suggested that the ketotic state produced by the KD may offer protection from insulin-induced hypoglycemia (5).

However, children on the KD do appear to be at risk for developing metabolic acidosis during GA, particularly during prolonged surgical procedures. Seven of the children were taking medications known to modify the acid–base balance at the time of GA. Three children were taking Bicitra (sodium citrate and citric acid), and five were taking TPM (6); one child was taking both. Bicitra is an alkalinizing oral agent that can be used to control symptomatic acidosis in children on the KD. However, no trend could be found in the current study to suggest that patients taking Bicitra were more resistant to developing acidosis during GA. Although previous reports have observed a higher incidence of metabolic acidosis in children taking TPM and on the KD (7), the current data do not show that children also taking TPM were more likely to become acidotic during GA.

Based on this study, children on the KD can safely have GA for surgical procedures. If the procedures are long, serum pH or bicarbonate levels should be used to monitor for acidosis. Currently at our institution, we recommend that levels be checked preoperatively and every 2–3 h during procedures lasting >3 h. If the baseline bicarbonate level or pH is low normal, we recommend rechecking the level perioperatively even for the shorter procedures. Blood pH and/or bicarbonate levels should be monitored in the postoperative period until the child is restarted on the full KD. If the child does develop acidosis perioperatively, intravenous bicarbonate should be used as indicated clinically.

These results are important, as they suggest that children who are on the KD for management of epilepsy can safely remain on the diet during GA and surgical procedures with no increased risk of perioperative complication or increase in seizure frequency. Children on the KD often have multiple medical issues and may require GA for necessary surgeries, as 12.7% of our KD patient population did during the time frame of this study. Because most of these children have a history of medically intractable epilepsy and have responded to the diet, it is clearly advantageous to be able to continue the KD safely rather than to taper or abruptly discontinue the diet before GA.

Acknowledgment: Dr. I. Valencia was the recipient of the Michelle S. and Roger M. Marino Epilepsy Research Fellowship.

This work also was supported by a generous contribution from Albert J. and Rea Elias.