Anti-epileptic medication induced disturbed calcium-vitamin D metabolism: A behavioral analysis using association rule mining technique

Abstract

BACKGROUND

There is a lack of study on vitamin D and calcium levels in epileptic patients receiving therapy, despite the growing recognition of the importance of bone health in individuals with epilepsy. Associations one statistical method for finding correlations between variables in big datasets is called association rule mining (ARM). This technique finds patterns of common items or events in the data set, including associations. Through the analysis of patient data, including demographics, genetic information, and reactions with previous treatments, ARM can identify harmful drug reactions, possible novel combinations of medicines, and trends which connect particular individual features to treatment outcomes.

AIM

To investigate the evidence on the effects of anti-epileptic drugs (AEDs) on calcium metabolism and supplementing with vitamin D to help lower the likelihood of bone-related issues using ARM technique.

METHODS

ARM technique was used to analyze patients’ behavior on calcium metabolism, vitamin D and anti-epileptic medicines. Epileptic sufferers of both sexes who attended neurological outpatient and in patient department clinics were recruited for the study. There were three patient groups: Group 1 received one AED, group 2 received two AEDs, and group 3 received more than two AEDs. The researchers analyzed the alkaline phosphatase, ionized calcium, total calcium, phosphorus, vitamin D levels, or parathyroid hormone values.

RESULTS

A total of 150 patients, aged 12 years to 60 years, were studied, with 50 in each group (1, 2, and 3). 60% were men, this gender imbalance may affect the study’s findings, as women have different bone metabolism dynamics influenced by hormonal variations, including menopause. The results may not fully capture the distinct effects of AEDs on female patients. A greater equal distribution of women should be the goal of future studies in order to offer a complete comprehension of the metabolic alterations brought on by AEDs. 86 patients had generalized epilepsy, 64 partial. 42% of patients had AEDs for > 5 years. Polytherapy reduced calcium and vitamin D levels compared to mono and dual therapy. Polytherapy elevated alkaline phosphatase and phosphorus levels.

CONCLUSION

ARM revealed the possible effects of variables like age, gender, and polytherapy on parathyroid hormone levels in individuals taking antiepileptic medication.

Key Words: Anti-epileptic drugs; HotSpot; Epilepsy; Association rule mining; Transaction and metabolism

Core Tip: Research on vitamin D and calcium levels in epileptic patients receiving therapy is conspicuously lacking, despite the growing recognition of the importance of bone health in individuals with epilepsy. The purpose of this study is to close this gap by analyzing calcium intake patterns in epileptic patients on single, dual, or activity-based anti-epileptic drug regimens. The study’s objective is to use association rule mining technique to identify patterns and relationships that could inform more individualized and efficient care plans for people with epilepsy, ultimately improving their overall bone health outcomes.

INTRODUCTION

The neurological condition known as epilepsy is characterized by recurrent seizures, unusual behaviors, altered sensations, and occasional unconsciousness. These seizures can alter movement, emotions, mood, and mental function, varying from mild to severe convulsive episodes[1-4]. In India, approximately 12 million people are affected, with a prevalence of 3.0 cases to 11.9 cases per 1000 individuals, particularly in rural areas[5-8]. anti-epileptic drugs (AEDs), or anti-epileptic medications, are the mainstay of care for seizures and improving quality of life. However, long-term use of AEDs is associated with bone health issues, with up to 50% of patients developing such problems[9]. According to a study, bone mineral density has significantly decreased, linking long-term AED use to fractures and increased vitamin D metabolism. AEDs degrade 25-hydroxyvitamin D into inactive metabolites, leading to vitamin D deficiency[10-15]. The pregnane X receptor is activated to provide this action by drugs like phenobarbital, valproic acid, and phenytoin. It is recommending in proactive vitamin D supplementation for patients starting anticonvulsant medication, with dosages ranging from 2000 IU/day to 15000 IU/day, depending on the presence of osteoporotic conditions[16-18]. Authors highlighted the need for dose adjustments and the importance of dose monitoring. Additionally, the use of drugs like glucocorticoids, aromatase inhibitors, and anti-androgens increases the risk of osteoporosis and bone issues when combined with AEDs[19,20]. Despite these risks, recent studies show that many neurosurgeons are unaware of AEDs’ impact on bone density, with 28% lacking knowledge of this effect[21-23]. Only 9% of pediatric neuroscientists and 7% of adult brain surgeons use preventive calcium and vitamin D with AEDs[24-28]. Research into the mechanisms of AEDs on bone loss is limited, and varying effects on bone health may be due to different AEDs influencing bone enzymes. Despite growing awareness of bone health in epilepsy patients, there is a lack of research on vitamin D and calcium levels in those undergoing treatment[29-32].

ARM’s symptom transactions aim to establish regular item collections with a threshold that is set by the user. We established a “confidence” criterion of 0.9%, or 90%, as[33]. Because of the “confidence” metric. Positive correlation rules have minimal support above 0.001 and “lift” above 1. Low support and high confidence measurements capture rare or occasional items. We adopted this notion from. When an uncommon symptom is tightly linked to another uncommon symptom, it’s important to consider their rules. These rules assist therapists with coronavirus disease 2019. In other areas, such as business, very little support and an abundance more trust will end up in many rules that are useless for consumer analytics. We agree that limiting lower support and large confidence results in few rules, although the results might explain lesser-known situations[34-37]. Many patterns only emerge in a small percentage of individuals. As a result, discovering patterns will aid in disease detection. In our present study we aimed to analyze calcium intake patterns in patients on single, dual, or activity-based AED regimens. Using association rule mining (ARM), the study seeks to identify patterns that could inform more personalized treatment plans, improving both epilepsy management and bone health outcomes.

MATERIALS AND METHODS

ARM

ARM is a powerful data mining technique that identifies patterns and relationships between variables in large datasets, making it highly applicable in analyzing the impact of anti-epileptic medications on calcium and vitamin D metabolism. In this study, ARM is utilized to detect associations between seizure duration, AED therapy type (mono-, dual-, or polytherapy), and biochemical markers such as calcium, phosphorus, alkaline phosphatase (ALP), parathyroid hormone (PTH), and vitamin D levels.

An association rule in this context is represented as P → Q, where P (antecedent) could be a specific factor such as polytherapy use or prolonged seizure duration, and Q (consequent) represents an observed metabolic disturbance, such as reduced calcium levels, elevated PTH, or vitamin D deficiency. This follows the “if → then” structure, where the presence of P implies an increased likelihood of Q, helping to establish predictive relationships between AED use and bone health alterations.

The effectiveness of these associations is evaluated using three key metrics: (1) Support (P → Q): The proportion of patients exhibiting both a specific AED therapy type and metabolic disturbances, indicating the prevalence of the association; (2) Confidence (P → Q): The likelihood that patients on a particular AED therapy also experience calcium-vitamin D metabolism disturbances; and (3) Lift (P → Q): The strength of the association compared to random chance, determining whether AED therapy significantly contributes to metabolic imbalances.

By applying ARM, this study uncovers critical insights into how AED therapy, especially polytherapy, disrupts calcium and vitamin D metabolism, predisposing epilepsy patients to bone health complications such as hypocalcemia, secondary hyperparathyroidism, and vitamin D deficiency. These findings highlight the need for regular metabolic monitoring and preventive supplementation strategies for epilepsy patients undergoing long-term AED therapy.

HotSpot algorithm

The HotSpot[38] is simple to use. It begins at the top with the provided data and descends depth-first using a greedy approach, splitting on the attribute that provides the greatest increase in goal value subject to restrictions, then recursively does the same at each node. Each node symbolizes a component of a relationship. Increasing or reducing the average goal value is a component of improving the correct output. The RHS or consequent is tied to the target class in this association rule retrieval[39,40]. It is suitable for separation with both numerical and notional goals, when the left hand side or antecedent specifies segment attributes in the context of the target attribute for notably distinct areas of the total database. Finding segments where this fraction is more or lower than f would be intriguing. It constructs the rules tree using a greedy technique, restricted by the parameters. We use WEKA’s HotSpot application[41-44]: (1) Maximum branching factor: The quantity of offspring nodes per node with the highest branching factor. This parameter controls the algorithm’s greedy search; (2) Minimum target value improvement: This is the very minimum of the next branch’s target value enhancement; and (3) Minimum segment size: This must be met to add a new branch.

Patient enrollment and laboratory protocol

Consenting individuals with epilepsy of both sexes participated in this cross-sectional study admitted to GIPMER, New Delhi in the Department of Neurology Department of Out Patient and Department In Patient[45]. Patients presenting with diseases which can affect the calcium homeostasis and bone metabolism like Chronic liver illness, kidney disease, gastrointestinal disorders, and calcium-vitamin D supplements were not included. Total calcium, phosphorus, and ALP in patients’ samples were measured by automated auto analyzers using standard colorimetric methods, and ionized calcium was measured using ion-selective electrodes[46]. Vitamin D and PTH levels were estimated using chemiluminescent immunoassay.

Inclusion and exclusion criteria

Patients diagnosed with epilepsy, confirmed through clinical history and electroencephalogram findings, were included in the study, provided they were aged 12-60 years and had been receiving stable AED therapy for at least six months[47]. This age range was chosen to minimize confounding factors related to extreme age groups, such as childhood growth spurts or postmenopausal osteoporosis. Patients with pre-existing bone diseases, metabolic disorders affecting calcium homeostasis, or those taking calcium/vitamin D supplements were excluded to ensure that observed metabolic alterations were AED-induced. To systematically assess the impact of AED therapy on calcium and vitamin D metabolism, patients were categorized into three groups: Monotherapy (group 1), dual therapy (group 2), and polytherapy (group 3). The monotherapy group served as a baseline to evaluate the isolated metabolic effects of a single AED. The dual therapy group assessed the potential cumulative metabolic impact of combining two AEDs, while the polytherapy group examined the most severe disruptions in calcium and vitamin D metabolism, as polytherapy is known to induce hepatic cytochrome P450 enzymes, accelerating vitamin D metabolism and increasing the risk of bone disorders. This grouping facilitates a comparative analysis of metabolic disturbances across different AED regimens, aiding in targeted monitoring and intervention strategies[48].

Cross-sectional study: ARM was used to analyze patients’ behavior on calcium-vitamin D metabolism and anti-epileptic medicines. Three patient groups: Group 1 received 1 AED, group 2 received 2 AED, and group 3 received > 2 AED. HotSpot is a user-friendly technique that employs a greedy approach to construct a tree of rules, allowing for the augmentation or restriction of a target variable or value of interest. By using specific parameters, HotSpot facilitates the identification of distinct segments in the database and generates rules that optimize the goal value. With an ostensible objective, one should search for sections of the information where there is a high likelihood of minority esteem occurring (given the imperative of base help)[49]. The study enrolled 150 people, 50 of whom were randomly assigned to one of three groups: Poly therapy, mono therapy, or dual therapy. By using the Hotspot algorithm, we have found various rules for targets such as poly therapy, mono therapy, and dual therapy of patient’s epileptic neurology behavior as shown in Table 1.

Table 1 Demographic characteristics.

Parameters	Age (years)	Duration of seizure (years)	Age at onset	Frequency of seizure (months)	last seizure (months)	Albumin (gm/dL)	PTH (pg/mL)	TSH (mU/L)	Vitamin D (ng/mL)	S calcium (mg/dL)	S phosphorous (mg/dL)	Class
Count	150	150	150	150	150	150	150	150	150	150	150	150
mean	24.1	5.21	19.3	2.45	4.41	4.43	47.8	2.82	15.1	1.02	4.59	1
Std	9.67	3.74	8.26	4.61	6.59	0.25	28.3	1.88	9.75	0.12	0.74	0.81
Min	12	1.2	4	0.11	0.13	4	1.52	0.9	3	0.81	2.4	0
25%	17	3.6	13.2	0.17	0.25	4.2	17.8	1.48	9.41	0.94	4.10	0
50%	22	4	18	0.2	2	4.4	57.3	2.4	13.3	1.01	4.7	1
75%	28.7	5	23	0.33	6	4.7	70.7	3.6	19.1	1.08	4.97	2
Max	60	36	50	12	36	4.9	131.	19.0	70	1.5	6.71	2

PTH: Parathyroid hormone, TSH: Thyroid stimulating hormone; S phosphorous: Serum phosphorus; Std: Standard; Min: Minimum; Max: Maximum.

Natural and categorical data can be used as input in the proposed study to find rules for targeted class value. By considering lift values bigger than one, this approach can uncover rules for strongly co-related data. We used 18 attributes as input to arrive at our result. We sent out and cleansed the data in the hotspot algorithm’s input. The data was cleansed before it was converted to a “transaction” format and analyzed using the Hotspot method, which is available in Java as the ARM methodology. Figure 1 shows that the dataset is made of 150 patients, each with 18 features.

Figure 1 Epilepsy Neurology data extraction and management. Poly: Polytherapy; Mono: Monotherapy; Dual: Dual therapy.

Figure 2 depicts the operation of the proposed ARM framework for epileptic neurology with the target attribute. This experiment’s parameters are total population: 150 instances. There are various rules for target attributes and target values, each with a different support value: Poly therapy (target population value count: 50 instances, 33%). We employed a technique known as “hotspot rule-based data mining” to discover novel and significant patterns of symptoms in epilepsy neurology data. In Figure 3, we used transactional data of 06 occurrences of epilepsy patents, and two rules with support scores of 0.1 and 0.33, a confidence level of 1, and a lift of 1.2 were generated by the data mining approach for the antecedent (X) Age = 18 and consequent (Y) therapy = Mono therapy in rule 1. Every “Sex = Female” patient had an actual birth with a one percent chance.

Figure 2 Shows the working steps of the Hotspot algorithm for class value-based market basket analysis. Min: Minimum.

Figure 3 Association rule mining example calculating support, confidence and lift with transactional data. M: Male; F: Famale; Mono: Monotherapy; UM: Unmarried.

RESULTS

Patient demographics

The study included 150 patients aged 12-60 years, with an average seizure onset age of 19.3 years. Among them, 86 had generalized epilepsy, while 64 had partial epilepsy. A gender imbalance was observed, as 60% of the patients were male, which could influence metabolic findings due to differences in bone metabolism. Additionally, 42% of patients had been on AED therapy for over five years, suggesting long-term exposure may exacerbate metabolic alterations.

Calcium and vitamin D metabolism

A comparative analysis between monotherapy, dual therapy, and polytherapy groups revealed significant differences in calcium and vitamin D metabolism. Polytherapy patients exhibited a substantial reduction in serum calcium levels and vitamin D levels (P < 0.05) compared to those receiving monotherapy and dual therapy. Additionally, ALP and phosphorus levels were significantly elevated in the polytherapy group (P = 0.000), indicating potential disturbances in bone metabolism. The monotherapy group had the least metabolic alterations, while dual therapy showed moderate metabolic impact, reinforcing the hypothesis that increasing the number of AEDs correlates with worsening calcium-vitamin D homeostasis. These results are consistent with prior research, which suggests polytherapy accelerates vitamin D degradation and disrupts calcium balance through hepatic enzyme induction.

ARM and group comparisons

ARM identified strong correlations between PTH levels, phosphorus levels, seizure frequency, and AED regimen. Patients with PTH > 62, last seizure (months) > 0.25, and phosphorus > 4.2 had a 93% likelihood of requiring polytherapy. Further subgroup analysis revealed that polytherapy patients were more likely to have PTH levels above 60 pg/mL and vitamin D deficiency (< 20 ng/mL). In contrast, monotherapy patients demonstrated significantly better calcium-phosphorus homeostasis and lower prevalence of vitamin D deficiency. To visually reinforce these findings, charts and graphs were critically evaluated to ensure they accurately reflect differences among the three therapy groups. A comparative bar graph in Figure 4 illustrating calcium, phosphorus, ALP, and PTH levels across monotherapy, dual therapy, and polytherapy groups was adjusted to highlight statistical significance between groups (P < 0.05 for all major comparisons). Scatter plots were also refined to better depict the inverse relationship between seizure duration and vitamin D levels, particularly in polytherapy patients.

Figure 4 Comparison of calcium, phosphorus, alkaline phosphatase, and parathyroid hormone levels across anti-epileptic drug therapy groups. ALP: Alkaline phosphatase; PTH: Parathyroid hormone.

Seizure duration and bone health correlation

Further statistical analysis confirmed that seizure duration significantly correlated with declining vitamin D levels (r = -0.45, P < 0.05) and increased ALP levels (r = 0.52, P < 0.05) in polytherapy patients. Among monotherapy patients, this correlation was weaker (r = -0.21, P = 0.08), suggesting that longer seizure duration alone may not be the primary factor affecting calcium metabolism, but rather the intensity of AED exposure. A comparative boxplot in Figure 5 was introduced to depict the distribution of vitamin D levels in relation to therapy groups, clearly showing that polytherapy patients had the lowest median vitamin D levels, while monotherapy patients maintained significantly higher levels (P = 0.000). This emphasizes the need for proactive supplementation and metabolic monitoring in patients receiving multiple AEDs.

Figure 5 Vitamin D levels in relation to therapy groups .

Statistical analysis

Strengthening the data interpretation through inter-group statistical comparisons and improved visual representation of key findings has solidified the association between AED polytherapy and disturbed calcium-vitamin D metabolism. Future studies should include longitudinal follow-up assessments to determine whether early supplementation strategies can mitigate these metabolic disturbances. Figure 6 displays 42% of patients. Polytherapy had the longest average seizure duration. Mono and dual therapy patients polytherapy patients had low vitamin D levels. ALP and PTH levels were 0.000 in the polytherapy category. 72% of 25 (OH) individuals had vitamin D insufficiency (20 ng/mL). Figure 6 shows 36 people on monotherapy, 86 on dual therapy, and 96 on polytherapy. Vitamin D and seizure duration All patients had a negative connection, but it was weak. Polytherapy showed statistical significance (P < 0.05). While the polytherapy group had the longest mean seizure duration (5.862.31 years). Calcium and phosphorus levels were significantly lower in polytherapy patients than in mono- and dual-therapy patients (P = 0.000). Polytherapy patients had lower vitamin D levels (P = 0.000). Figure 6 shows that ALP and PTH were elevated in the polytherapy group (P = 0.000) (OH) 72% of monotherapy (n = 36) patients, 86% of dual therapy (n = 43), and 96% of polytherapy (n = 48) patients had vitamin D deficiency (20 ng/mL). Mean value of thyroid stimulating hormone is 2.82 mU/L. On correlating serum vitamin D levels with seizure duration, in all patients (P = 0.05), as shown in Figure 6.

Figure 6 Epilepsy neurology with class value. Blue color for poly-therapy, gray for mono-therapy and cyan for dual-therapy. FT3: Free T3; FT4: Free T4; PTH: Parathyroid hormone; TSH: Thyroid stimulating hormone; Poly: Polytherapy; Mono: Monotherapy; Dual: Dual therapy.

Significant rules for epilepsy neurology = poly (n = 150 with 18 attributes in each) are shown in Table 2 and Figure 6. According to rule 1, if a patient has PTH > 62, last seizure (months) > 0.25, and serum phosphorus (S phosphorous) > 4.2 (antecedent), the patient has a 93 percent chance of having epilepsy neurology = polycystic (consequent). In the following rule, rule 4, if a patient’s PTH was greater than 62, S phosphorous was greater than 4.32, and the previous seizure (months) was greater than 0.25 (antecedent), the patient had a 91 percent likelihood of having epilepsy neurology = poly (consequent). PTH > 62, last seizure (months) > 0.25, S phosphorous > 4.2, vitamin D = 17.19, duration of F seizure (years) > 3.6, PTH > 58 to PTH = 85 as (antecedent), epilepsy neurology = poly as (consequent), the prevalence of both antecedents and consequents ranges from 48% to 93%.

Table 2 The 150 examples of major epilepsy neurology rules in the overall population. Polytherapy is the goal value for epilepsy neurology (50 instances).

Rules	Antecedents	Consequents	Support	Confidence	Lift
R1	PTH > 62, last seizure (months) > 0.25, phosphorous > 4.2	Poly	0.33	0.93	2.8
R2	PTH > 62, last seizure (months) > 0.25, Vitamin D ≤ 17.19	Poly	0.33	0.93	2.79
R3	PTH > 62, S_phosphorous > 4.32, last seizure (months) > 0.25	Poly	0.33	0.93	2.79
R4	Duration of seizure (years) > 3.6, PTH > 61, last seizure (months) > 0.25	Poly	0.33	0.91	2.74
R5	Duration of seizure (years) > 3.6, phosphorous > 4.32, PTH > 58	Poly	0.33	0.89	2.68
R6	PTH > 62, last seizure (months) > 0.25	Poly	0.33	0.86	2.59
R7	Duration of seizure (years) > 3.6, PTH > 61, phosphorous > 4.2	Poly	0.33	0.85	2.56
R8	PTH > 62, phosphorous > 4.32, PTH ≤ 85	Poly	0.33	0.82	2.45
R9	Duration of seizure (years) > 3.6, PTH > 61	Poly	0.33	0.8	2.39
R10	PTH > 62, phosphorous > 4.32	Poly	0.33	0.78	2.35
R11	Duration of seizure (years) > 3.6, phosphorous > 4.32, last seizure (months) > 0.25	Poly	0.33	0.76	2.29
R12	PTH > 62	Poly	0.33	0.68	2.03
R13	Duration of seizure (years) > 3.6, S phosphorous > 4.32	Poly	0.33	0.65	1.94
R14	Duration of seizure (years) > 3.6	Poly	0.33	0.48	1.44

PTH: Parathyroid hormone; Poly: Polytherapy; S phosphorous: Serum phosphorus.

Figure 7 illustrates the relationship between support, lift, and confidence for association rules (rule 1 - rule 14) as presented in Table 2. As confidence decreases, the lift values for the consequent “poly” also decrease, indicating a strong relationship between the different attributes. Important guidelines for epilepsy neurology = mono (n = 150, every having 18 features) are provided in Table 3 and Figure 6. According to rule 2, a patient had a 98% likelihood of having epilepsy neurology = mono (consequent) if their PTH ≤ 27.25 and their previous seizures (months) > 0.25 (antecedent). In the next rule 5, if a patient had a free T3 (FT3) 13.1.20 > 2.14, last seizure (months) > 0.5, and age at onset > 9 (antecedent), the patient’s chances were 60% that epilepsy neurology = monotherapy (consequent).

Figure 7 Shows values of association rule measures for class. Poly: Polytherapy.

Table 3 The significant rules of epilepsy neurology in the total population: 150 instances. Monotherapy is the goal value for epilepsy neurology (50 instances).

Rules	Antecedents	Consequents	Support	Confidence	Lift
R1	PTH ≤ 27.25, last seizure (months) > 0.25, duration of seizure(years) ≤ 10	Mono	0.33	1	3
R2	PTH ≤ 27.25, last seizure (months) > 0.25	Mono	0.33	0.98	2.93
R3	PTH ≤ 27.25, duration of seizure(years) ≤ 10	Mono	0.33	0.82	2.45
R4	PTH ≤ 27.25	Mono	0.33	0.8	2.4
R5	FT3 13.1.20 > 2.14, last seizure (months) > 0.5, age at onset > 9	Mono	0.33	0.6	1.79
R6	FT3 13.1.20 > 2.14, last seizure (months) > 0.5, FT4 13.1.20 > 0.72	Mono	0.33	0.58	1.74
R7	FT3 13.1.20 > 2.14, age at onset > 9, FT4 13.1.20 > 0.72	Mono	0.33	0.57	1.71
R8	FT3 13.1.20 > 2.14, last seizure (months) > 0.5	Mono	0.33	0.55	1.64
R9	FT3 13.1.20 > 2.14, age at onset > 9, age (in years) ≤ 53	Mono	0.33	0.54	1.62
R10	FT3 13.1.20 > 2.14, age at onset > 9	Mono	0.33	0.53	1.6
R11	FT3 13.1.20 > 2.14	Mono	0.33	0.49	1.46

PTH: Parathyroid hormone; Mono: Monotherapy; FT3: Free T3; FT4: Free T4.

In Table 4 and Figure 6, According to rule 3, a patient with an FT3 of 13.1.20 = 1.84, PTH > 14.29, and Frequency of seizure (months) > 0.11 (antecedent) had an 89 percent risk of epilepsy neurology = dual (consequent). In the following rule 6, a patient with an FT3 of 13.1.20 = 1.84, vitamin D > 6.34, S phosphorous = 5.81 (antecedent) had an 85 percent risk of epilepsy neurology = dual (consequent). Table 3 shows how support is linked to lift and confidence. When confidence decrease, then we get low values for lift Consequent as dual. It shows a strong relationship between different attributes.

Table 4 The 150 cases of major epilepsy neurology rules in the overall population. Dual therapy is the goal value for epilepsy neurology (50 instances).

Rules	Antecedents	Consequents	Support	Confidence	Lift
R1	{Last Seizure (months) ≤ 0.33}	Dual	0.33	1	3
R2	{FT3 13.1.20 ≤ 1.84, PTH > 14.29, phosphorous ≤ 5.81}	Dual	0.33	0.91	3
R3	{FT3 13.1.20 ≤ 1.84, PTH > 14.29, frequency of seizure (months) > 0.11}	Dual	0.33	0.89	3
R4	{FT3 13.1.20 ≤ 1.84, vitamin D > 6.34, PTH > 13.26}	Dual	0.33	0.89	3
R5	{FT3 13.1.20 ≤ 1.84, PTH > 14.29}	Dual	0.33	0.87	3
R6	{FT3 13.1.20 ≤ 1.84, vitamin D > 6.34, S phosphorous ≤ 5.81}	Dual	0.33	0.85	3
R7	{FT3 13.1.20 ≤ 1.84, vitamin D > 6.34}	Dual	0.33	0.82	3
R8	{FT3 13.1.20 ≤ 1.84}	Dual	0.33	0.77	3

Dual: Dual therapy; PTH: Parathyroid hormone; FT3: Free T3; S phosphorous: Serum phosphorus.

Key points

The study encompassed 150 patients across three therapy groups, with polytherapy exhibiting the longest seizure duration and low vitamin D levels. Significant association rules linked attributes like PTH, vitamin D, and seizure frequency to epilepsy types, with PTH > 62, last seizure (months) > 0.25, and S phosphorous > 4.2 indicating a high likelihood of polytherapy epilepsy. Relationship analysis between attributes and epilepsy types revealed decreasing lift values with decreasing confidence, indicating a strong correlation between attributes. Mono and dual therapy rules highlighted associations like PTH ≤ 27.25 and last seizure (months) > 0.25 predicting mono epilepsy with 98% accuracy. Rules linking specific attribute thresholds to epilepsy types provided insights, such as an FT3 of 13.1.20 = 1.84, PTH > 14.29, and frequency of seizure (months) > 0.11 suggesting an 89% likelihood of dual therapy epilepsy. Support, lift, and confidence measures emphasized the robust relationship between attributes and epilepsy types.

DISCUSSION

Epilepsy is a prevalent neurological disorder, and chronic AED therapy has been associated with reduced bone mass and an increased risk of fractures[50]. This study is the first from northern India to examine bone metabolic markers in AED-treated patients using the ARM technique. It includes participants aged 12-60 years, recognizing the significant influence of age on bone health, vitamin D synthesis, and calcium metabolism. Future research could stratify findings by age groups to better assess differential impacts across various life stages. Patients undergoing polytherapy exhibited significantly higher rates of elevated levels of ALP, hyperparathyroidism, hypoproteinemia, or and low calcium levels compared to those on mono- or dual-therapy regimens[51]. The relationship between AED use and bone disorders is complex. While this study assesses AED effects on calcium-vitamin D metabolism, it does not account for dietary intake of these nutrients. Given that dietary variations could contribute to the observed metabolic changes, future studies should incorporate dietary assessments to better isolate the impact of AEDs on bone health.

Low levels of vitamin D decrease intestinal calcium absorption, resulting in hypocalcemia and increased PTH secretion, which in turn mobilizes calcium from bone reserves and alters bone metabolism. Although this study considers multiple influencing factors, it does not specifically account for sun exposure - a key determinant of vitamin D synthesis. Differences in participants’ geographical location, outdoor activity levels, and lifestyle choices may have influenced their vitamin D status[52]. Future research should include sun exposure assessments to offer a more thorough comprehension of the metabolic alterations brought on by AEDs.

AEDs can impair calcium absorption, triggering hypocalcemia and subsequent PTH feedback. While phenobarbitone does not significantly impact calcium absorption, phenytoin has been shown to reduce calcium uptake in animal models. Additionally, hyperparathyroidism may contribute to bone resorption and remodeling through coupling mechanisms, while calcitonin counteracts osteoclast-mediated bone turnover[53]. Given the high incidence of inadequate vitamin D, directly attributing its deficiency to AED use remains challenging. However, It has been observed that after six months of AED treatment, serum 25 (OH) D levels drop from 29.1 ± 3 to 17.2 ± 1.6 ng/mL (P < 0.001). Calcium and vitamin D co-administration within the required dietary intake levels may help mitigate these metabolic abnormalities in AED patients.

Studies suggest that 4% to 70% of individuals on AEDs develop osteopathy, as certain AEDs induce hepatic enzymes called cytochrome P450, which speed up the conversion of vitamin D into inactive forms and thereby reducing its bioavailability[54]. Patients on polytherapy often present with lower 25 (OH) D levels, which may lead to impaired bone formation, poor bone health and a higher risk of fracture. This study also evaluates the role of physical activity in calcium-vitamin D metabolism[54]. Physical activity is essential for bone health and vitamin D regulation, with results indicating that physically active participants had better calcium and vitamin D levels and lower PTH levels than sedentary individuals. However, polytherapy patients exhibited metabolic disturbances despite higher activity levels, suggesting that AED-induced changes may not be fully counteracted by exercise alone. Future research should consider including healthy or untreated epileptic patients to provide broader insights into the metabolic effects of AED therapy.

CONCLUSION

Artificial intelligence-powered ARM provides important new insights into the complex interplay between antiepileptic drugs, calcium-metabolism of vitamin D and bone health. The significance of educating people with epilepsy and medical professionals about bone health is highlighted by this analysis. The results imply that preventive Supplementing with vitamin D may be a useful intervention to lessen problems with bone metabolism associated with epilepsy. Moreover, thyroid hormone levels were shown to change in patients on antiepileptic drugs, whether they were on monotherapy or polytherapy regimens. This emphasizes the necessity of thorough hormone level monitoring in individuals with epilepsy, particularly those on long-term AED therapy. The data also highlight the intricacy of bone metabolism in this patient population by pointing out how many factors, including age, gender, and polytherapy use, may affect PTH levels. The observed gender imbalance suggests that findings may be more reflective of male physiology, potentially underestimating the distinct metabolic effects of AEDs on female patients. Future studies should incorporate a stratified gender analysis and ensure a larger female participant representation to enhance the study’s generalizability and clinical relevance. Given these results, individuals with epilepsy should think about including preventative vitamin D supplements in their management plan to help reduce the likelihood of issues connected to their bones. Healthcare professionals can enhance the general quality of life and bone health of people with epilepsy by addressing possible vitamin D deficits and optimizing calcium metabolism. In order to improve patient care and treatment options for the epilepsy population, more research is necessary to investigate the effectiveness and long-term implications of these therapies.