Abstract

Patients and Methods

Results

Patient Flow and Baseline Characteristics

Details of patient flow, including all study exclusions and withdrawals, are shown in Figure ​Figure1.1. We randomized 47 patients, of which 44 commenced the diets and 38 completed the study (86% completion rate for patients commencing the diets); 6 patients withdrew for reasons unrelated to the diets and 3 patients withdrew as a result of diet‐related difficulties. Randomized patient baseline characteristics are shown in Table ​Table1.1. There were no significant between‐group differences in any baseline characteristics.

Open in a separate window

Figure 1

Patient flow.

Table 1

Randomized patient baseline characteristics

	Low‐fat group (n=23)	Ketogenic group (n=24)
Age (years)	61.48±7.12	64.29±6.69
Sex (male)	14 (61%)	17 (71%)
Ethnicity
European	22 (96%)	22 (92%)
Maori	0	2 (8%)
Asian	1 (4%)	0
Functional ability
Spouse at home	20 (87%)	21 (88%)
Screening MoCA	25.4±2.0	25.3±2.5
H & Y	1.78±0.82 (range, 1‐4)	2.13±0.76 (range, 1‐4)
MDS‐UPDRS
Part 1 (nonmotor daily living experiences)	8.96±4.34	11.15±4.15
Part 2 (motor daily living experiences)	11.13±5.59	12.75±5.30
Part 3 (motor examination)	34.93±13.40	36.48±13.29
Part 4 (motor complications)	4.33±4.44	4.90±3.95
PD meds
None	4 (17%)	1 (4%)
Monoamine oxidase B inhibitor	5 (22%)	4 (17%)
Dopamine agonist	5 (22%)	10 (42%)
l‐dopa	18 (78%)	22 (92%)
Catechol‐O‐methyl transferase inhibitor	2 (9%)	4 (17%)
Amantadine	4 (17%)	3 (13%)
Apomorphine	0	1 (4%)
Comorbidities
Type 1 diabetes	0	1 (4%)
Type 2 diabetes	1 (4%)	0
Cholecystectomy	1 (4%)	1 (4%)
Past renal stones	1 (4%)	3 (13%)
Past gout	2 (9%)	1 (4%)
Physical profile
Weight (kg)	78.13±19.45	83.71±19.38
BMI	26.96±6.35	27.77±5.29
Recommended calorie intake (kcal per day)	2,236±430	2,242±417
Blood profile
HbA1C (mmol/mol)	34.67 (5.32%)±5.22	36.15 (5.46%)±8.60
Triglycerides (mmol/L)	1.71±0.88	1.81±1.18
HDL (mmol/L)	1.51±0.47	1.50±0.44
LDL (mmol/L)	2.74±0.82	2.73±1.06
Total cholesterol (mmol/L)	5.03±1.00	5.03±1.05
Urate (mmol/L)	0.30±0.08	0.30±0.08
CRP (mmol/L)	2.95±2.99	1.95±1.59

Open in a separate window

Except for % variables, values are presented as mean±standard deviation.

Baseline values were obtained by averaging the two baseline clinical visits; 1 patient withdrew post‐randomization, several hours before their scheduled second baseline visit, so the first visit was used as the baseline value.

Blood Glucose and Ketone Levels

Patient‐monitored blood glucose and ketone levels are shown in Figure ​Figure2.2. Over the 8‐week diet intervention, the two diet groups significantly differed in mean weekly bedtime blood glucose (low‐fat group: 6.28±0.73 mmol/L vs. ketogenic group: 5.70±1.20 mmol/L; P=0.001) and ketone (0.16±0.05 mmol/L vs. 1.15±0.59 mmol/L; P<0.001) levels.

Open in a separate window

Figure 2

Mean weekly bedtime blood (A) glucose and (B) ketone (beta‐hydroxybutyrate) levels (n=23 for the low‐fat diet group, n=24 for the ketogenic group). Data were missing for 16.3% of the recordings (1.8% of the 38 completer recordings, 77% of the 9 withdrawal recordings). Days missing data were left blank, with the weekly mean calculated using the remaining days of the week. Error bars indicate standard error.

Changes in MDS‐UPDRS Parts 1 to 4 and Metabolic Parameters

We confirmed that patients in both diet groups took their l‐dopa at the same time before each visit by measuring the time interval from the last clinically relevant (taken within the previous 8 hours) l‐dopa dose to MDS‐UPDRS assessment, which did not significantly differ between any two clinical visits within either group (low‐fat group: 128±56 minutes at baseline, 126±58 minutes in week 4, 131±79 minutes in week 8, P>0.05; ketogenic group: 93±50 minutes at baseline, 92±61 minutes in week 4, 88±51 minutes in week 8, P>0.05).

Changes in MDS‐UPDRS Parts 1 to 4 are shown in Table ​Table2.2. Both diet groups significantly decreased in Part 1 (low‐fat group: P=0.030 vs. ketogenic group: P<0.001), Part 2 (P=0.011 vs. P<0.001), and Part 3 (P<0.001 vs. P<0.001); the low‐fat group showed no change in Part 4 (P=0.13) whereas the ketogenic group decreased in Part 4 (P=0.005). The ketogenic group showed a larger magnitude of decrease in Part 1 (P<0.001), with the largest between‐group decreases observed for urinary problems (–0.70 points difference), pain and other sensations (–0.64 points difference), fatigue (–0.50 points difference), daytime sleepiness (–0.45 points difference), and cognitive impairment (–0.27 points difference). There were no between‐group differences in the magnitude of decrease for Part 2 (P=0.11), Part 3 (P=0.055), or Part 4 (P=0.32). The within‐ and between‐group comparisons of changes in Parts 1 to 4 remained similar, and the statistical conclusions unchanged, in all four sensitivity analyses (for details, see Table S1 in the Supplementary Appendix).

Table 2

Changes in MDS‐UPDRS Parts 1 to 4 and metabolic parameters.

	Low‐fat group (n=23)			Ketogenic group (n=24)
	Baseline	Week 8	Change	Baseline	Week 8	Change	P Value (between groups)
Part 1 (nonmotor daily living experiences)	8.96 ±4.34	7.96 ±6.56	−0.99±3.63	11.15 ±4.15	6.57 ±4.09	−4.58±2.17	<0.001
Part 2 (motor daily living experiences)	11.13 ±5.59	9.80 ±6.81	−1.33±3.28	12.75 ±5.30	9.62 ±5.64	−3.13±4.01	0.11
Part 3 (motor examination)	34.93 ±13.40	26.36 ±13.58	−8.58±5.50	36.48 ±13.29	30.20 ±12.88	−6.27±4.07	0.055
Part 4 (motor complications)	4.33 ±4.44	3.54 ±4.86	−0.79±2.71	4.90 ±3.95	3.33 ±3.02	−1.56±2.45	0.32
Weight (kg)	78.13 ±19.45	73.26 ±17.99	−4.87±2.47	83.71 ±19.38	79.34 ±17.13	−4.37±3.00	0.55
BMI	26.96 ±6.35	25.28 ±5.95	−1.67±0.79	27.77 ±5.29	26.31 ±4.46	−1.46±1.01	0.30
HbA1C (mmol/mol)	34.67±5.22	34.18±3.74	−0.49±2.72	36.15±8.60	34.72±5.68	−1.42±4.12	0.095
Triglycerides (mmol/L)	1.71 ±0.88	1.45 ±0.55	−0.26±0.74	1.81 ±1.18	1.71 ±0.81	−0.10±0.63	0.46
HDL (mmol/L)	1.51 ±0.47	1.40 ±0.38	−0.11±0.22	1.50 ±0.44	1.89 ±0.75	+0.40±0.59	<0.001
LDL (mmol/L)	2.74 ±0.82	2.35 ±0.77	−0.40±0.30	2.73 ±1.06	3.42 ±1.37	+0.70±0.67	<0.001
Total cholesterol (mmol/L)	5.03 ±1.00	4.40 ±0.93	−0.63±0.51	5.03 ±1.05	5.98 ±1.39	+0.94±0.80	<0.001
Urate (mmol/L)	0.30 ±0.08	0.28 ±0.08	−0.02±0.03	0.30 ±0.08	0.33 ±0.08	+0.03±0.05	<0.001
CRP (mmol/L)	2.95 ±2.99	1.95 ±1.79	−1.00±3.57	1.95 ±1.59	2.24 ±1.26	+0.29±1.93	0.10

Open in a separate window

Values are presented as mean ± standard deviation.

Bolded values highlight changes in scores and parameters; due to round‐off, some change values differ by 0.01 from the absolute value differences.

P values refer to changes in scores and parameters between (not within) groups.

Changes in metabolic parameters are also shown in Table ​Table2.2. Both diet groups significantly decreased in weight and BMI (P<0.001 for all within‐group comparisons), but there were no between‐group differences in the magnitude of weight loss (P=0.55) or BMI decrease (P=0.30). The low‐fat group did not change in HbA1C (P=0.92) whereas the ketogenic group decreased in HbA1C (P=0.032), although with the exclusion of 1 outlier, a patient with type 1 diabetes whose HbA1C decreased by –19.5 mmol/mol, the magnitude of HbA1C decrease in the ketogenic group lessened considerably (from –1.42±4.12 mmol/mol to –0.64±1.49 mmol/mol). There were no between‐group differences in the magnitude of change in HbA1C (P=0.095). There were no within‐group changes in triglycerides (low‐fat group: P=0.19 vs. ketogenic group: P=0.55), nor were there any between‐group differences (P=0.46). The low‐fat group decreased in HDL (P=0.027), LDL (P<0.001), total cholesterol (P<0.001), and urate (P=0.004) whereas the ketogenic group increased in HDL (P<0.001), LDL (P<0.001), total cholesterol (P<0.001), and urate (P=0.004), with notable between‐group differences observed for all four parameters (P<0.001 for all between‐group comparisons). There were no within‐group changes in CRP (P=0.47 vs. P=0.12), nor were there any between‐group differences (P=0.10).

Discussion

We have shown, in a general neurology hospital clinic of PD patients, that it is plausible and safe to maintain a low‐fat or ketogenic diet for 8 weeks. Both diet groups significantly improved in motor and nonmotor symptoms, but the ketogenic group showed greater improvements in nonmotor symptoms. Adverse effects were generally mild and differed between the two groups.

Several study strengths should be highlighted. First, we utilized a one‐phase design, strictly enforced consistent timing, diet‐blinded neurologists, and the MDS‐UPDRS (a validated rating scale with high internal consistency in measuring motor and nonmotor symptoms)18 to minimize assessment bias. Second, our diet plans utilized recipe ingredients readily available at local supermarkets, with an emphasis on affordability and palatability—thus, we chose not to use regular supplements in either plan (in particular, we chose not to supplement the ketogenic diet with medium‐chain triglyceride supplements, which are not available at some supermarkets, are relatively costly, and can produce adverse gastrointestinal effects). Third, protein intake was approximately 1 g per kg of body weight per day within each diet group and equal between groups; this is critical, given that low‐protein diets containing 0.5 to 0.8 g of protein per kg of body weight per day enhance L‐dopa absorption compared to high‐protein diets, which may improve motor symptoms and exacerbate dyskinesias in some people with PD.24, 25, 26 Fourth, our protocol involved multimedia supports, such as weekly educational videos, designed to mitigate the patient losses traditionally associated with studies involving ketogenic diets in adults,23 which probably contributed to the low number of patient withdrawals attributed to diet‐related difficulties (3 patients).

Certain study weaknesses warrant mention. First, the study population and duration were small; a larger sample size or longer diet intervention would have increased the statistical power or provided additional time, either of which may have enabled the detection of further significant differences between the diet groups. Thus, our findings should be viewed as preliminary. Second, there were 9 patient withdrawals; imputation techniques were used to replace the missing data. To assess the robustness of our primary outcome findings, we additionally performed sensitivity analyses on all primary outcomes using mean imputation, baseline observation carried forward, median imputation, and complete case analyses.

Patient adherence to each diet plan was monitored directly and indirectly. Rather than use food diaries, which may be perceived as burdensome,27 we directly monitored diet adherence by simply ensuring that patients ticked the day‐to‐day completion of each set meal as they proceeded. Perhaps more important, we indirectly monitored diet adherence by training patients to self‐monitor their daily blood glucose and ketone levels throughout the study; we chose blood pinprick testing as it is easier, more specific, and more accurately reflects ketone levels compared to urine dipstick testing.28 Using blood monitors, the low‐fat diet group measured negligible ketone levels, whereas the ketogenic group demonstrated a mean weekly blood ketone level of 1.15±0.59 mmol/L consistent with a state of physiological ketosis, a coordinated metabolic response in which the liver provides ketones as an alternative, fat‐derived fuel source when body glucose reserves are in short supply.29

Both diet groups significantly improved their MDS‐UPDRS scores over the 8‐week diet intervention. It is tempting to speculate that the low‐fat diet increased brain dopamine levels and/or gut short‐chain fatty acid production and that the ketogenic diet enhanced central and peripheral neuron energy metabolism. However, despite being reminded to eat until satiated, the average patient in both groups lost 4 to 5kg. It is well documented that overweight patients on unrestricted low‐fat and low‐carbohydrate diets can still experience significant weight loss, which may relate to altered levels of energy‐regulating hormones such as insulin.30, 31 Although weight loss is often associated with a negative impact on PD severity,32 a recent study has shown that patients with metabolic syndrome experience greater PD progression over time, mainly as a result of increased motor scores, compared to patients without metabolic syndrome.33 Given that the average patient in both groups lost 4 to 5kg, yet remained overweight at week 8, the observed weight loss may have contributed to the improved motor scores in both groups. In addition, the placebo effect may significantly impact motor symptoms in PD patients.34, 35 Since our protocol included multimedia supports designed to mitigate patient withdrawals, it is also possible that a placebo effect contributed to the improved motor scores in both groups. The potential effects of both weight loss and a placebo effect must be kept in mind when interpreting the within‐group comparisons.

Both diet groups improved in Part 1 (nonmotor daily living experiences), but the ketogenic group improved more; every single patient in the ketogenic group improved in Part 1, resulting in a substantial 41% reduction in baseline Part 1 scores (as opposed to 11% in the low‐fat group) over the 8‐week diet intervention. The robustness of this between‐group finding to various imputation conditions was confirmed with multiple sensitivity analyses, including the conservative baseline observation carried forward method, which assumed that none of the 6 patient withdrawals from the ketogenic group would have improved in Part 1 at all. This is a potentially important finding, given that nonmotor symptoms ultimately represent the most disabling aspect of PD—for example, depression alone may have over twice the impact of motor symptoms on health.4 Moreover, nonmotor symptoms, such as urinary problems, pain, fatigue, daytime sleepiness, and cognitive impairment, are among those least responsive to l‐dopa,2, 36 yet these are the nonmotor symptoms that improved the most in the ketogenic group compared to the low‐fat group in our study. Since the magnitude of weight loss and BMI decrease were similar between groups, and the protocol and assessments were applied equally to both groups excepting the diets, neither weight loss nor a placebo effect explains the between‐group differences in nonmotor symptom improvements.

Perhaps unsurprisingly, the low‐fat diet group showed decreases, whereas the ketogenic group showed increases, in HDL, LDL, total cholesterol, and urate. The health benefits of lower versus higher LDL and total cholesterol in PD are debatable, with some studies correlating higher LDL and total cholesterol levels with a lower risk of PD.37, 38 The association between urate levels and PD is less controversial, with higher serum urate levels predicting a slower rate of PD progression.39

Interestingly, the adverse effect profile differed between the two diet groups. The most common adverse effect in the low‐fat group was excessive hunger. In contrast, the most common adverse effect in the ketogenic group was an intermittent exacerbation of the PD tremor and/or rigidity, which resulted in 2 patient withdrawals from this group at the end of week 1. Although this adverse effect improved or resolved in many patients in weeks 5 to 8, suggesting that it was a largely transient phenomenon, it may still have impacted the Part 3 (motor examination) scores to some extent at week 8. We can only speculate as to the reason the ketogenic diet exacerbated the PD tremor and/or rigidity in some patients, but it is conceivable that the abrupt increase in fat intake temporarily augmented dopamine depletion and/or oxidative stress in the SN, followed by an adaptive response over time.40

In conclusion, this pilot randomized, controlled trial shows that modified diets based on readily available ingredients, with normal protein levels, are plausible and safe treatment approaches in PD, with the ketogenic diet leading to greater improvements in many of the more disabling, less l‐dopa‐responsive nonmotor symptoms. It is possible that a ketogenic diet could play a complementary role alongside l‐dopa in the treatment of PD, but due to the preliminary nature of our findings, larger and longer randomized, controlled studies are needed before this can be stated with confidence.

Author Roles

(1) Research Project: A. Conception, B. Organization, C. Execution; (2) Statistical Analysis: A. Design, B. Execution, C. Review and Critique; (3) Manuscript Preparation: A. Writing of the First Draft, B. Review and Critique.

M.C.L.P.: 1A, 1B, 1C, 2A, 2B, 3A, 3B

D.K.J.M.: 1A, 1C

L.J.G.: 1C, 3B

F.J.S.A.: 1C, 3B

C.D.P.L.: 1C, 3B

Financial Disclosures

D.K.J.M. runs a whole‐foods coaching business, however none of her recipes were used in this study. Both study diets were based on recipes from sources with no personal or financial affiliation to any of the authors.

Supporting information

Acknowledgments

Notes

Funding agencies: Waikato Hospital Neurology Research Fund.

Relevant conflicts of interest/financial disclosures: Nothing to report.

Full financial disclosures and author roles may be found in the online version of this article.

Registry: Australia New Zealand Clinical Trials ACTRN12617000027314.

References