Efficacy and tolerability of the Modified Atkins Diet in adults with pharmacoresistant epilepsy: A prospective observational study


Address correspondence to Martin del Campo, M.D., 399 Bathurst Street, Suite WW5-443, Toronto, ON, Canada M5T 2S8. E-mail: martin.delcampo@uhn.on.ca


Purpose: Evidence from the pediatric population exists for the efficacy of ketogenic diets in reducing seizure frequency in patients with intractable epilepsy. Recent evidence suggests that a Modified Atkins Diet may be a beneficial form of cotherapy for adult patients with pharmacoresistant epilepsy.

Methods: A prospective, open-label study was performed of adults >18 years of age with pharmacoresistant epilepsy. Carbohydrates were restricted to 20 g/day. Fluids and calories from protein and fat were allowed ad libitum.

Key Findings: Eighteen patients, ages 18–55 years, were initially enrolled. Using an intent-to-treat analysis, 12% had a >50% seizure reduction after 3 months; 28% after 6 months, and 21% after 12 months. Response at 3 months predicted response at 12 months in 79% of patients. The mean decrease in weight was 10.9 kg and the mean decrease in body mass index (BMI) was 3.8, p = 0.01. Fourteen of 18 patients (78%) completed 12 months of this diet. Patients experienced a decrease in triglycerides from (mean) 1.22 to 0.9 mm (p = 0.02).

Significance: The Modified Atkins Diet demonstrates modest efficacy as cotherapy for some adults with pharmacoresistant epilepsy and may be also helpful for weight loss. Financial and logistical barriers were significant factors for those who declined enrollment and for those who discontinued the study.

The role of the “classic” ketogenic diet in the management of childhood epilepsy is well established (Wilder, 1921; Henderson et al., 2006; Freeman et al., 2007; Neal et al., 2008). There is a growing body of evidence from the pediatric population suggesting that a diet that promotes ketosis by severely restricting carbohydrate intake may also be helpful in reducing the frequency of pharmacoresistant seizures (Kossoff & Dorward, 2008; Kossoff et al., 2008; Muzykewicz et al., 2009). The Atkins Diet has been popularized in the media and calls for severe carbohydrate restriction in its “initiation phase” (Atkins, 2002). A Modified Atkins Diet (MAD) has been developed as a potential cotherapy for patients with intractable epilepsy (Kossoff & Dorward, 2008). Unlike the Atkins Diet, the MAD prolongs the initial severe carbohydrate restriction phase indefinitely, and the primary goal of the diet is improved seizure control, not weight loss. The MAD also differs from the Atkins Diet in that liberal fat intake is allowed and encouraged if weight gain is desirable. This diet is less complex and more palatable than a traditional ketogenic diet and, therefore, may be associated with greater compliance in adolescent and adult patient populations (Kossoff et al., 2008). There is still very little known about the efficacy of the MAD in the treatment of medically refractory epilepsy in the adult patient.

This study purports to contribute to previous work in this field intended to determine (1) if the MAD is an effective and sustainable method of cotherapy in intractable epilepsy in an adult population and (2) whether compliant patients show health gains beyond greater seizure control. Previous studies following patients for 6 months on the MAD report modest results, with approximately one third of patients experiencing >50% decrease in seizure frequency (Carrette et al., 2008; Kossoff et al., 2008). Our study adds to these reports by following patients over a longer period on the MAD, and tracking changes in body mass index (BMI) and weight during this time.


This trial is registered in ClinicalTrials.gov with number NCT00188500. The study protocol, consent forms, and all diet-related documents were developed and approved according to guidelines from the University Health Network Research Ethics Board.

A prospective, open-label study design was employed. Outpatients in the adult epilepsy clinic were invited to participate in the study. Participants were at least 18 years of age, had medically refractory epilepsy, and were unlikely to undergo epilepsy surgery or vagus nerve stimulator implantation during the period of follow-up. Patients who were deemed unable to understand or follow the principles of the diet; who had nonepileptic seizures, unstable medical conditions (e.g., diabetes mellitus, hypertension), or terminal conditions (e.g., cancer, dementia); and/or patients with a history of renal stones were excluded.

Patients were evaluated at baseline and at 3, 6, and 12 months after starting the diet. Patients or caregivers were asked to keep detailed food intake and seizure diaries throughout. Self-reported seizure frequency and detailed food intake records were requested at each visit. Weight, BMI, and urinary ketosis were evaluated at each visit and lipid profile, electrolytes, and complete blood count (CBC) were done every 6 months.

All patients consulted with a dietician, where they were shown a basket of acceptable foods (see Appendix), and where costs and principles of the diet were explained. All patients received a document that had been assembled by the dietician, including instructions on how to follow the diet and acceptable recipes. Patients were shown which types of food were ad libitum (e.g., all fish, fowl, meat, and shellfish) and which were to be limited (e.g., only 2–3 servings of vegetables per day and limited fruit intake).

Patients were instructed to limit their carbohydrate intake to 20 g/day. Fluids and calories in the form of fat and protein were ad libitum. Patients purchased dipsticks to test for urinary ketones and a standard multivitamin tablet to supplement with daily. Medical visits and pharmacotherapy were modified in accordance with best medical practice at the discretion of the treating epileptologist.

Means were compared using a paired two-sample t-test. The significance level for all tests was p < 0.05.



One hundred thirty patients were approached and invited to participate in the study. Only 18 patients consented and were subsequently enrolled and treated between the summer of 2005 and the fall of 2008. Fourteen patients completed the study. Financial and logistical difficulties associated with adherence to the diet were cited as the reason for dropout in all four cases. These patients were of lower socioeconomic status and reported that financing a diet that was high in meat and low in processed carbohydrates was not possible for them. These patients also reported insurmountable difficulties in adhering to the rigorous and indefinite carbohydrate restriction required in the MAD. Eleven patients (61%) were female. The median age was 29 years (range 18–55 years). The median self-reported seizure frequency at baseline was 12 per month (range 2–600 seizures per month). The median number of antiseizure medications at the initial visit was three (range one to four medications); one patient had a vagus nerve stimulator and three (17%) had previous epilepsy surgery (two had frontal lobe resection, one had temporal, and one had occipital). Eleven patients (61%) had partial and secondarily generalized seizures, four had complex partial seizures, two had myoclonic seizures, and one had simple partial seizures.


At 3 months, 2 (12%) of 17 patients reported >50% reduction in seizure frequency, 4 (28%) of 14 at 6 months, and 3 (21%) of 14 at 12 months.

Of the 18 patients who enrolled in the study, 4 (22%) dropped out between baseline and 12 months. Seventeen (94%) presented to clinic at 3 months, and 14 patients (78%) presented at 6 and 12 months. Only three patients experienced >50% seizure reduction and none became seizure-free. Response to the diet at 3 months predicted response at 12 months in 11 patients (79%).

The influence of patient demographics on 12-month outcomes is listed in Table 1. All three patients who experienced a >50% seizure reduction were female, compared to only six (55%) of nonresponders (p < 0.01). Changes made to antiepileptic drug (AED) regimens after 12 months of the diet are summarized in Table 2. Two patients decreased their AED regimens despite reporting an increase in seizure frequency while on the diet. One patient was able to decrease her total AED regimen after starting a new medication (felbamate), and one did so against the recommendations of the treating epileptologist due to financial constraints. Two patients reported that their postictal recovery time had improved significantly while on the diet.

Table 1.   Prediet demographics and their influence on diet efficacy at 12 months
Patient demographicResponse to diet (n = 3)No response to diet (n = 11)p- Value
  1. Values are expressed as median (range) or number (%).

Age (years)36 (19–45)34 (18–55)0.8
Seizure frequency (per month)407 (60–600)16 (2–67)0.15
Antiseizure drugs at diet onset3 (2–4)3 (1–4)0.88
Female gender3 (100%)6 (55%)0.01
Prior resective epilepsy surgery1 (33%)3 (27%)0.88
Table 2.   Changes in AED regimen at 12 months
Change in AEDSeizure reduction at 12 monthsSeizure increase at 12 monthsp-Value
Increased n (%)2 (28)4 (58)0.17
Decreased n (%)1 (14)2 (28)0.36
None n (%)4 (58)1 (14)0.08


Urinary ketosis could not be evaluated satisfactorily. Patients did not keep sufficiently accurate records to be useful. Data from patients recording at least one measurement since their last visit are as follows (if more than one measurement, the last recorded value was used): at 3 months, 11 (85%) of 13 were ketotic (seven small, four trace). Of the two patients reporting a >50% reduction in monthly seizure frequency, one was ketotic (trace) and the other did not have ketosis data at this point. At 6 months, 9 (82%) of 11 were ketotic (four moderate, two small, three trace). Of the four patients reporting a >50% monthly seizure reduction, two (50%) were ketotic (one moderate, one trace). At the final 12 month visit, 8 (80%) of 10 were ketotic (one moderate, four small, three trace). Again, only two (50%) of the three patients reporting a >50% reduction in seizure frequency were ketotic (both trace).

Effect on weight

Of the 14 patients who completed the 12-month study, 12 (86%) were overweight at baseline and were encouraged to lose weight. These patients lost weight from a baseline mean (standard deviation, SD) of 88.9 kg (21.1 kg) to 78 kg (19.6 kg) at their final clinic visit. BMI was similarly reduced from a mean (SD) of 31.6 (5) to 27.8 (4.6), p < 0.01. The median BMI change was −3.1 (range 1.1–7.6). At study onset, nine (75%) were clinically obese (BMI > 30) versus four (33%) at their final clinic visit, p < 0.02. Percent reduction in weight was not correlated with percent change in self-reported monthly seizure frequency at 12 months (p < 0.4). Likewise, percent reduction in BMI at 12 months was not correlated with percent change in self-reported monthly seizure frequency (p < 0.46).

Adverse effects

No patient discontinued the diet due to abnormal laboratory results or increased seizure frequency. Constipation, renal stones, and irregular menses were not reported. The range for total cholesterol at 12 months was 3.82–6.61 mm (normal <5.2 mm) with a median of 5.24 mm, considered within the average cardiovascular risk ranges (Foster et al., 2003; Samaha et al., 2003). Triglycerides decreased from 1.22 to 0.9 mm (p = 0.02). Fasting blood glucose (4.98 mm, normal 4.0–7.0 mm) was also within the normal range. One patient developed a new type of seizure (left arm jerks) after 3 months on the diet.


The results of this small, prospective observational study corroborate previous findings (Carrette et al., 2008; Kossoff et al., 2008), which suggest that some benefit exists for the use of the MAD as a form of cotherapy in the management of intractable epilepsy in an adult population. We were successful in following 14 of 18 patients for a total of 12 months on the MAD. Our study demonstrates a beneficial effect for some adult patients with pharmacoresistant epilepsy up to 1 year after beginning the diet. Response to the diet at 3 months predicted the response at 12 months in the majority of patients in this study. A 3-month trial of the MAD, therefore, may be sufficient to determine whether or not this is an efficacious and sustainable cotherapy in an adult patient with intractable epilepsy.

Importantly, as described by Kossoff et al. (2008), ketosis appeared to be unrelated to diet efficacy, and these findings may have relevance to theories about its mechanism in inducing seizure reduction. Indeed, there is a large body of evidence to support the claim that level of ketosis is not correlated with level of seizure inhibition (Appleton & DeVivo, 1974; Bough et al., 1999; Likhodii et al., 2000; Greene et al., 2003). Proponents of this theory argue that it is reduced levels of glucose, and not elevated ketones per se, that are responsible for the efficacy of dietary treatments for epilepsy such as the ketogenic diet, and that the efficacy of said treatments is influenced mainly by diet composition. Our study provides clinical support for this theory, as some patients were ketotic yet responded poorly to the MAD, whereas others were not ketotic and responded quite well.

The diet was well tolerated; the only change in biochemistry was a significant decrease in triglyceride levels after 1 year. There is no known consequence of lowered triglyceride levels. In the four cases of discontinuation, financial and logistical difficulties associated with adherence to the diet were cited. Strict adherence to the MAD requires that patients purchase and prepare high protein and high fat meals (e.g., poultry, meat, fish, eggs), which tend to be both more expensive and more difficult and time-consuming to prepare than processed carbohydrates (e.g., pasta, rice, pastries). Some caregivers related difficulties controlling carbohydrate intake in cognitively impaired subjects attending day programs or special schools where other individuals would bring and maybe offer candy and other “comfort foods.” Others felt “guilty” constantly denying favorite snacks and treats. Patients with epilepsy tend to be of lower socioeconomic status than the general population (Heaney et al., 2002; Elliott et al., 2009). Although the cost and diligence required for adherence to the MAD (or other ketogenic diets) may not be a significant concern for pediatric patients who are living with caregivers, it may prove to be a significant obstacle for adults living with intractable epilepsy in the community.

Weight loss coupled with a significant reduction in BMI occurred in all patients who were overweight at the onset of the study. Nine of the patients who began the study met diagnostic criteria for clinical obesity (BMI > 30). After 12 months on the diet, five of these patients no longer met this diagnostic criterion. As hypothesized, patients who are compliant on the MAD do show health gains (in the form of weight reduction) beyond achieving greater seizure control. Therefore, the MAD may be an especially beneficial cotherapy for pharmacoresistant epilepsy in adult patients with clinical obesity. It is important to note that not all patients on the MAD will lose weight, as the primary goal of the diet is seizure control, and not weight loss (Kossoff & Dorward, 2008). Indeed, Patient 112 (Table 3) was underweight at the onset of the study, and his diet was modified so that he was able to successfully gain weight throughout the course of the study. He reported a reduction in monthly seizure frequency at 3 months and again at 6 months despite a gain in weight and BMI. He did not present for follow-up at 1 year.

Table 3.   Change in monthly self-reported seizure frequency for all participants
Patient IDBaseline3 months6 months12 months
  1. D/C, dropped out of the study; N/A, did not present for follow-up.


Female gender predicted a >50% seizure reduction after 12 months on the MAD in this study. This may be due to the fact that women are more motivated than men to comply with dietary restrictions. Weight loss (in addition to seizure control) may be more desirable for women than men. Indeed, marketing studies prove that women are more concerned about weight loss and diet than about health indicators such as cardiovascular disease and cancer (Meredith Corp/NBC Universal, 2008). However, the number of patients is so few that it would be difficult to make a definitive statement in this regard.

This study has several limitations, mostly relating to the small sample size. It is possible that the results could represent a placebo effect or regression to the mean, although this is unlikely given the length of the study. In addition, although prospective, this was an open-label study and no control group was included (it was extremely difficult to recruit patients for the control group as there was no incentive to keep accurate food records). This might have been circumvented by a long observational period prior to commencement of the MAD during which patients would provide such information without knowing what type of diet they were to undertake. However, it was deemed unethical due to the financial implications of the MAD. Importantly, data relating to food intake and seizure frequency were all self-reported and, therefore, may in actuality have been construed.

It is also possible that limiting the carbohydrate restriction to 20 g/day may have limited the efficacy of the diet. We felt the traditional 10 g or the 15 g/day used by the Hopkins group may have proven too restrictive and less likely to be adhered to.

It was also evident that our patient population was unable to keep adequate food or ketosis records. It is, therefore, impossible to do a comparative analysis of dietary composition.

There is also a possible effect of the manipulation of AED dosages and/or type. This was left to the attending neurologist’s judgment. Although statistically this patient population is unlikely to attain seizure freedom with such manipulations (Kwan & Brodie, 2000), the effect of these manipulations cannot be considered irrelevant. Indeed, eight patients witnessed some alteration to their AED regimen through the course of this study. Finally, there is a possibility that weight reduction may have resulted in proportionally higher AED serum concentrations, although there was no significant correlation between reduction in seizure frequency and this variable.

In summary, this observational study supports previous work, suggesting that the MAD may constitute a beneficial cotherapy in the treatment of pharmacoresistant epilepsy in some adult patients. Independent from its effect on seizure frequency/severity, the MAD may be beneficial in patients with clinical obesity or in those patients desiring weight loss. Financial and logistical barriers associated with adherence to the diet are significant and should be strongly considered on a case-by-case basis before initiating treatment. It would appear that a multicenter study on the efficacy, mechanism, and associated health benefits of the MAD in the treatment of intractable epilepsy in adult populations is needed to establish effective clinical use guidelines. However, it could be argued that a single center study with diet subsidy, if positive, could stimulate a “grass roots” movement to demand insurance or government financial support for diet supplementation and/or dietary counseling.


None of the authors has any conflict of interest to report. We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.


Foods that may be eaten liberally:

All fish including: tuna, salmon, sole, trout, flounder, sardines, herring.

All poultry including: chicken, turkey, duck, goose, Cornish hen, quail, pheasant.

All shellfish including: oysters, mussels, lobster, clams, squid, shrimp, crabmeat.

All meat including: beef, pork, lamb, bacon, veal, ham, venison.

Other acceptable foods: eggs, cheese.

You may consume three to four ounces daily of the following full fat, firm soft and semisoft aged cheeses, including: cheddar; cow, sheep, and goat cheese; cream cheese, Gouda, mozzarella, Roquefort, blue cheese, Swiss.

Salad vegetables: you can have 2–3 cups per day.

Alfalfa sproutsDaikonParsley
Bok ChoyEscaroleRadicchio

These salad vegetables are high in nutrients and provide a good source of fiber.

Salad garnishes: crumbled crisp bacon, grated cheese, minced hard-boiled egg, sautéed mushrooms, sour cream.

Beverages: club soda, decaffeinated coffee or tea, diet soda made with sucralose (Splenda), herb tea without barley or and fruit sugar added, lemon juice or lime juice (note that each contains 2.8 g carbohydrate per ounce; limit to two to three tablespoons), mineral water, spring water, water, clear broth/bouillon (not all brands; read the label).

Spices: All spices to taste, but make sure none contain added sugar. Herbs.


Fats and oils: butter, mayonnaise (regular), cream (heavy or light); limit to two to three tablespoons a day; note carbohydrate content. For salad dressing use oil and vinegar (but not balsamic vinegar, as it contains sugar) or lemon juice and herbs and spices. Prepared salad dressing without added sugar and no more than two carbohydrates per tablespoon serving are also fine. Olive oil (virgin or extra-virgin). Canola, peanut, or grapeseed oil.

Nutritional supplements: A multivitamin is recommended and depending on your need a calcium supplement is also recommended.

Other vegetables: You can have one cup per day if salad does not exceed two cups. These vegetables are slightly higher in carbohydrate content than the salad vegetables. Do not eat liberally.

Artichoke heartsCauliflowerScallions
AsparagusChardSnow peas
Bamboo shootsCollard greensSpaghetti squash
Bean sproutsEggplantSpinach
Beet greensLeeksString or wax beans
BroccoliOkraSummer squash
Brussel sproutsOnionTomato
 RhubarbWater chestnuts

Sample daily menu of 20 g of carbohydrate

Breakfast: three-egg omelette with avocado, mozzarella cheese, and tomato; decaffeinated coffee with cream.

Lunch: beef round steak (8 oz), spinach, and mixed leaf salad with mushrooms, onions, celery, and parmesan cheese.

Dinner: broiled salmon (9 oz), Kale topped with garlic, lemon, and sesame seeds.

Follow this diet plan for 2 weeks, or 14 days. After 2 weeks, most people begin to feel better.

If a vegetable, such as spinach or tomato, cooks down significantly, it must be measured raw so as not to underestimate its carbohydrate amount.