A novel index to assess low energy fracture risks in patients prescribed antiepileptic drugs.

OBJECTIVE: To develop an index assessing the risks of low energy fractures (LEF) in patients prescribed antiepileptic drugs (AED) by exploring five previously suggested risk factors; age, gender, AED-type, epilepsy diagnosis and BMI.
METHODS: In a population-based retrospective open cohort study we used real world data from the Electronic Health Register (EHR) in Region Kalmar County, Sweden. 23 209 patients prescribed AEDs at any time from January 2008 to November 2018 and 23 281 matching controls were followed from first registration in the EHR until the first documented LEF, disenrollment (or death) or until the end of the study period, whichever came first. Risks of LEF measured as hazard rate ratios in relation to the suggested risk factors and in comparison to matched controls were analyzed using Cox regression. The index was developed using a linear combination of the statistically significant variables multiplied by the corresponding regression coefficients.
RESULTS: Data from 23 209 patients prescribed AEDs and 2084 documented LEFs during a follow-up time of more than 10 years resulted in the Kalmar Epilepsy Fracture Risk Index (KEFRI). KEFRI = Age-category x (1.18) + Gender x (-0.51) + AED-type x (0.29) + Epilepsy diagnosis-category x (0.31) + BMI-category x (-0.35). All five previously suggested risk factors were confirmed. Women aged 75 years and older treated with an inducing AED against epilepsy and BMIs of 25 kg/m2 or below had 48 times higher LEF rates compared to men aged 50 years or younger, treated with a non-inducing AED for a condition other than epilepsy and BMIs above 25 kg/m2.
CONCLUSION: The KEFRI is the first weighted multifactorial assessment tool estimating risks of LEF in patients prescribed AEDs and could serve as a feasible guide within clinical practice.

Introduction

With an ageing population worldwide, the number of low energy fractures (LEF) is expected to increase from the 9 million annual fractures registered at the turn of the century [1]. These fractures cause both suffering and generate considerable health care costs [2]. Risk factors for LEFs have been assessed in several epidemiological studies [3, 4], resulting in the development of general risk assessment tools now being used in clinical practice. One of these tools is FRAX®, which integrates eight clinical risk factors in estimating the 10-year major osteoporotic fracture risk [5]. Some of these risk factors, such as age, gender and BMI apply to the population in general. One of the more specific risk factors included in FRAX® is secondary osteoporosis, usually defined as low bone mass in the presence of an underlying disease or drug [6]. Drugs causing bone loss and thus increasing the risk of LEF have been stressed as an important area within drug safety [7]. The only drug class included in FRAX® at present is glucocorticoids.

Contradictory to this drug class selection, increased fracture rates in patients using antiepileptic drugs (AED) have been recognized ever since the 1960s [8] when risks were initially presented in institutionalized patients with epilepsy. The use of AEDs have since disseminated into other medical areas such as psychiatry [9] and pain medicine [10] resulting in a substantial increase in AED consumption. The fracture risk among AED users has been extensively investigated and is evidently multifactorial [11]. Drug side effects including impaired gate stability [12] and influence on bone mineral density [13] are considered contributing factors. The inducing effect on hepatic Cytochrome P450 (CYP) enzymes attributed to some AEDs, have been associated with vitamin D deficiency, and thus as one further underlying cause of both bone impairment and LEF, suggesting fracture risk differences between AED types [14]. In addition to side effects of AEDs, the medical conditions themselves can contribute to the fracture risk, e.g. due to seizure-related falls [14]. An epilepsy-specific risk assessment tool has been requested [15], since neither the epilepsy diagnosis nor use of AEDs are included in the FRAX® definition of secondary osteoporosis. In the recently updated version of QFracture®, another osteoporotic fracture risk tool, epilepsy (either diagnosed or prescribed anticonvulsants) has been added, but merely as a single binary risk factor [16]. The multifactorial nature of the fracture risk among AED users calls for a more differentiated approach in risk assessment.

Data on drug prescriptions, medical diagnoses and events are registered in Electronic Health Records (EHR) together with basic demographic parameters. This type of real world data can act as a source in creating risk prediction models in health care [17]. In this study we developed a multifactorial risk assessment tool for LEFs in patients prescribed AEDs by applying a combination of five previously suggested risk factors; age, gender, BMI, AED-type and epilepsy diagnosis on data from EHR. We also compared the risk for LEFs in patients prescribed AEDs with a matched control group for risk factor confirmation.

Methods

We conducted a retrospective, population-based, open cohort study. Our study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline [18]. The regional ethical review board of Linköping University approved the study plan and deemed it exempt from informed consent because the data was pseudonymized.

Data source

The EHR of Region Kalmar (Cambio COSMIC) was first introduced in 2008 and reached full coverage in 2010. The EHR covers the entire Kalmar County population, varying from 233 400 (in 2008) to 244 700 (in 2018). Relevant data regarding health events, prescriptions, vital parameters, clinical coding, and lab results are included in addition to basic demographic parameters. Both public and private care providers are included in primary as well as secondary care. Both somatic and psychiatric care is included in the EHR.

Study cohort

Patients prescribed at least one AED during the period or having initiated treatment before January 2008 were initially included (Fig 1). For comparative outcome analysis a control group was selected with computer-based randomization based on gender and the age-decade interval at study initiation, ratio 1:1. The control group was defined as having no AED prescriptions before or during the study period. Patients were followed from first registry in the EHR and last follow-up date was defined as either decease date or date of the latest registration in EHR regardless of health condition. The study included data from January 2008 to November 2018 initially comprising 46 775 participants (AED patients = 23 396, and controls = 23 379). Later, 12 patients were excluded because of missing initiation dates (index dates) and 5 patients due to missing age information. Also, 268 patients were excluded because of fracture dates preceding the first AED prescription initiation date.

Fig 1

Flowchart for patient and control group inclusion and primary outcome.

Primary outcome

The primary outcome measure was defined as documented LEF i.e. a combination of ICD-10 codes according to the Swedish National Board of Health and Welfare’s definition of osteoporosis-related fractures [19] and ICD-10 codes for low energy trauma (LET). Osteoporosis-related fractures were thus defined as ICD-10 codes: S22 (Fracture of rib(s), sternum and thoracic spine), S32.1—S32.8 (Fracture of lumbar spine and pelvis), S42.2 (Fracture of upper end of humerus), S42.3 (Fracture of shaft of humerus), S52.5 (Fracture of lower end of radius), S52.6 (Fracture of lower end of ulna), S72.0—S72.4 (Fracture of femur) and S82.1 (Fracture of upper end of tibia). LET was defined as ICD-10 codes for external causes of mortality and morbidity for low energy trauma by codes starting with: W00—W08 (Fall on same level). The registration of LET codes is compulsory in conjunction with a fracture registration in the EHR in Region Kalmar County. The fracture code and LET code had to be registered on the same healthcare event in order to be included.

Variable definitions

For each participant demographic information on age, gender and date of decease were obtained. Furthermore, information on diagnosis, BMI, and drug prescription were extracted.

Age was calculated from date of birth to first registration date in EHR and age was categorized as: < 18, 18–50, 51–74 and ≥ 75 years.

Epilepsy was defined by ICD-10 codes starting with: G40 (Epilepsy and recurrent seizures), G41 (Status epilepticus) or F803 (Acquired aphasia-epilepsy syndrome). Diagnoses registered by medical secretaries and/or physicians were included regardless of level or type of care situation.

BMI was categorized as: ≤ 25 and > 25 kg/m2. The values used were either the registered BMI values or values calculated by registered measurements of length and weight.

Antiepileptic drugs were identified and classified into two groups based on their enzyme inhibiting profile. The group inducing CYP enzymes contained ATC codes: N03AA02 (phenobarbital), N03AA03 (primidone), N03AB02 (phenytoin), N03AB05 (fosphenytoin), N03AF01 (carbamazepine), N03AF02 (oxcarbazepine), N03AF03 (rufinamide) and N03AF04 (eslicarbazepine).

The group non-inducing CYP-enzymes contained ATC codes: N03AD01 (ethosuximide), N03AE01 (clonazepam), N03AG01 (valproate), N03AG04 (vigabatrin), N03AG06 (tiagabine), N03AX03 (sultiame), N03AX09 (lamotrigine), N03AX10 (felbamate), N03AX11 (topiramate), N03AX12 (gabapentin), N03AX14 (levetiracetam), N03AX16 (pregabalin) and N03AX17 (stiripentol). Groups included in the statistical analyses were: inducing, non-inducing, and control group (no AED prescription)

Patients with one or more prescriptions for AED were included. Patients with only one type of AED during the study period were classified accordingly, while patients with both types of AED were classified according to drug exposure. Drug exposure was represented by the total number of prescription days, i.e. all prescription periods were calculated and summarized for the study follow-up period. Patients having prescriptions of both types of AEDs, were categorized in the group with the highest number of prescription days. For 78 patients the number of prescription days were exactly the same, 39 patients were then included by randomization to inducing, and 39 to non-inducing. 23 209 patients prescribed an AED (20497 non-inducing and 2712 inducing) and 23 281 matching controls were thus included for statistical analysis.

Statistical analyses

We used a Cox univariate regression analysis followed by a multivariate analysis (p<0.1). For variables displaying statistical significance in the multivariate analysis, an index was developed for patients prescribed AED using a linear combination of the variables multiplied by the corresponding regression coefficients. Since the age groups < 18 and 18–50 years had identical fracture frequencies (1.7%) the two categories were lumped together in the final Cox regression giving the three categories ≤ 50, 51–74 and ≥ 75 years. The comparative results were presented in Kaplan Meier graphs. The software Statistica v.12 (StatSoft, Inc., Tulsa, OK, USA) was used for all analyses.

Results

Characteristics of patients and control group

Since the introduction of the EHR-system in Region Kalmar County in 2008, 23 209 patients having been prescribed AEDs in the EHR (20 497 with non-inducing AEDs + 2712 with inducing AEDs in Table 1) were included. This count also include the 680 patients which had AEDs prescribed solely prior to the EHR roll out, but having been registered in conjunction with the roll out (Fig 1). Of the 23 209 patients, 1546 patients had been prescribed both non-inducing and inducing AEDs during the study period (Fig 1). These patients were classified according to drug exposure i.e. the highest number of prescription days became the dominant AED type. The median age for participants’ first registration in EHR was 55 years and nearly one of two patients prescribed AEDs were male (46.3%). A total of 3921 (16.8%) of the patients prescribed AEDs had an epilepsy diagnosis registered. The vast majority (20 497, 88.3%) of the patients prescribed AEDs, received drugs not considered as CYP enzyme inducers. The three most commonly prescribed non-inducing drugs were gabapentin, pregabalin and lamotrigine (82.5% of the non-inducing drugs prescribed) and the three most commonly prescribed enzyme-inducing drugs were carbamazepine, oxcarbazepine and phenytoin (90.8% of the inducing drugs prescribed). The age- and gender-matched control group consisted of 23 281 patients having a medical visit registered in the EHR, but having not been prescribed any AEDs during the period.

Table 1

Characteristics of patients prescribed antiepileptic drugs according to enzyme inhibiting profile and age- and gender-matched controls.

			AED-type
Variable		Controls	Non-inducing	Inducing	Total
N		23281	20497	2712	46490
Gender
	Male (n; %)	10758 (46.2)	9129 (44.5)	1612 (59.4)	21499 (46.2)
	Female (n; %)	12523 (53.8)	11368 (55.5)	1100 (40.6)	24991 (53.8)
Age (years)
	Mean (SD)	52.0 (20.9)	52.4 (20.6)	50.5 (21.9)	52.1 (20.8)
	Median (range)	55 (0–108)	55 (0–101)	54 (0–100)	55 (0–108)
Age categories (years, n; %)
	<18	1697 (7)	1383 (7)	298 (11)	3378 (7)
	18–50	7980 (34)	7147 (35)	839 (31)	15966 (34)
	51–74	10315 (44)	8942 (44)	1236 (46)	20493 (44)
	≤75	3289 (14)	3025 (15)	339 (13)	6653 (14)
BMI (kg/m2)
	Mean (SD)	27.5 (5.7)	28.7 (6.3)	27.8 (5.9)	28.3 (6.1)
	Median (range)	27 (13–82)	28 (13–70)	27 (14–61)	27 (13–82)
	Missing (n; %)	16188 (69.5)	6783 (33.1)	1331 (49.1)	24302 (52.3)
Alive at period end (n; %)
	Yes	20985 (90.1)	16260 (79.3)	2042 (75.3)	39287 (84.5)
	No	2296 (9.9)	4237 (20.7)	670 (24.7)	7203 (15.5)
Epilepsy diagnosis (n; %)
	Yes	46 (0.2)	2804 (13.7)	1117 (41.2)	3967 (8.5)
	No	23235 (99.8)	17693 (86.3)	1595 (58.8)	42523 (91.5)
Antiepileptic drugs (AED) (n; %)
	Gabapentin (N03AX12)	0 (0.0)	10004 (48.8)	0 (0.0)	10004 (21.5)
	Pregabalin (N03AX16)	0 (0.0)	4313 (21.0)	0 (0.0)	4313 (9.3)
	Lamotrigine (N03AX09)	0 (0.0)	2585 (12.6)	0 (0.0)	2585 (5.6)
	Valproate (N03AG01)	0 (0.0)	1676 (8.2)	0 (0.0)	1676 (3.6)
	Levetiracetam (N03AX14)	0 (0.0)	851 (4.2)	0 (0.0)	851 (1.8)
	Clonazepam (N03AE01)	0 (0.0)	782 (3.8)	0 (0.0)	782 (1.7)
	Topiramate (N03AX11)	0 (0.0)	267 (1.3)	0 (0.0)	267 (0.6)
	Vigabatrin (N03AG04)	0 (0.0)	11 (0.1)	0 (0.0)	11 (0.0)
	Ethosuximide (N03AD01)	0 (0.0)	7 (0.0)	0 (0.0)	7 (0.0)
	Sultiame (N03AX03)	0 (0.0)	1 (0.0)	0 (0.0)	1 (0.0)
	Carbamazepine (N03AF01)	0 (0.0)	0 (0.0)	2173 (80.1)	2173 (4.7)
	Oxcarbazepine (N03AF02)	0 (0.0)	0 (0.0)	165 (6.1)	165 (0.4)
	Phenytoin (N03AB02)	0 (0.0)	0 (0.0)	125 (4.6)	125 (0.3)
	Phenobarbital (N03AA02)	0 (0.0)	0 (0.0)	118 (4.4)	118 (0.3)
	Fosphenytoin (N03AB05)	0 (0.0)	0 (0.0)	98 (3.6)	98 (0.2)
	Primidone (N03AA03)	0 (0.0)	0 (0.0)	27 (1.0)	27 (0.1)
	Rufinamide (N03AF03)	0 (0.0)	0 (0.0)	6 (0.2)	6 (0.0)
Long term epilepsy medication (≥5 yrs) (n; %)
	Yes	0 (0.0)	5609 (27.4)	1248 (46.0)	6857 (14.7)
	No	23281 (100)	14888 (73)	1464 (54)	39633 (85)
LEF during follow up period (n; %)
	Yes	1098 (4.7)	1813 (8.8)	271 (10.0)	3182 (6.8)
	No	22183 (95)	18684 (91)	2441 (90)	43308 (93)

Age and age categories is the age when patients had the first registry in EHR.

Risk of low energy fractures in relation to previously suggested risk factors

In total, 2084 (1813 with non-inducing AEDs + 271 with inducing AEDs) patients (9.0%) among those prescribed AEDs suffered from LEFs (Table 1) during the median follow-up period of more than 10 years (124.7 months, IQR 87.5 to 130.6 months) after the first registration in the EHR. The hazard rate ratios of LEFs were analyzed with regards to five previously suggested risk factors; age, gender, type of AED, epilepsy diagnosis and BMI (Table 2). In the univariate Cox regression analysis, the statistical differences in individual risk factors are displayed. Advanced age, female gender, having been prescribed an AED, having a registered epilepsy diagnosis and a low BMI all involved higher LEF rates. In the multivariate Cox regression analysis, the five risk factors’ relative importance is presented, taking the variation of the risk factors between groups into account. All five risk factors remained statistically significant in the multivariate regression analysis. Long term medication (≥ 5 years) was also incorporated in the model but was not found statistically significant, and did not alter the regression coefficients included in the index significantly. Patients prescribed CYP enzyme-inducing AEDs had a hazard rate ratio for LEF of 1.63 (1.38 to 1.92) compared to controls, while patients prescribed non-inducing AEDs had a ratio of 1.23 (1.13 to 1.35) compared to controls. Patients prescribed AEDs for epilepsy had a ratio for LEF of 1.40 (1.23 to 1.59) compared to those prescribed AED for non-epileptic conditions. In Figs 2 and 3, the differences in LEF risk associated with type of AED (Fig 2) and epilepsy diagnosis (Fig 3) compared to controls over time are shown in women and men, respectively.

Table 2

Results from Cox proportional hazard regression regarding risks for low energy fractures, analysed for the total cohort.

					Univariate Cox regression		Multivariate Cox regression
Parameter		N total	Fractures	Fractures (%)	HR (95% conf. limits)	p	HR (95% conf. limits)	p
Age categories (years, n)
	<18	3378	57	1.7	0.06 (0.05–0.08)	<0.001	0.03 (0.01–0.11)	<0.001
	18–50	15966	274	1.7	0.06 (0.05–0.07)	<0.001	0.07 (0.06–0.08)	<0.001
	51–74	20493	1552	7.6	0.28 (0.26–0.31)	<0.001	0.34 (0.31–0.37)	<0.001
	≥75	6653	1299	19.5	1.00		1.00
Gender (n)
	Female	24991	2210	8.8	1.00		1.00
	Male	21499	972	4.5	0.53 (0.49–0.57)	<0.001	0.57 (0.52–0.62)	<0.001
Group (n)
	Controls	23281	1098	4.7	1.00		1.00
	Non-inducing	20497	1813	8.8	1.57 (1.46–1.70)	<0.001	1.23 (1.13–1.35)	<0.001
	Inducing	2712	271	10.0	1.88 (1.65–2.15)	<0.001	1.63 (1.38–1.92)	<0.001
Epilepsy diagnosis (n)
	No	42523	2756	6.5	1.00		1.00
	Yes	3967	426	10.7	1.57 (1.42–1.74)	<0.001	1.40 (1.23–1.59)	<0.001
BMI (kg/m2, n)
	≤25	6999	997	14.2	1.00		1.00
	>25	15189	1444	9.5	0.61 (0.56–0.66)	<0.001	0.70 (0.64–0.75)	<0.001

Age category is the age for the patient’s first entry in EHR. HR is hazard rate ratio. The multivariate analyses were reduced from 46490 to 22188 due to missing BMI values.

Significant results are presented in bold.

Fig 2

Low energy fracture-free probability over time in relation to type of antiepileptic drug, compared to controls for men (left) and women (right), respectively.

Fig 3

Low energy fracture-free probability over time in relation to being prescribed antiepileptic drugs and having a registered epilepsy diagnosis or not, compared to controls for men (left) and women (right), respectively.

Visualization of combined risk factors

The differences in LEF risk when combining a number of risk factors are displayed in Fig 4. The group with the lowest fracture risk during the follow-up period in the study was used as reference (LEF risk 1.00) and risks in all other groups were calculated as hazard rate ratios in relation to the reference. Women aged 75 years and older treated with an inducing AED against epilepsy and BMIs of 25 kg/m2 or below thus had 48 times higher LEF rates compared to men aged 50 years and younger, treated with a non-inducing AED for a condition other than epilepsy, and BMIs above 25 kg/m2 (the reference group).

Fig 4

Chart displaying relative low energy fracture risks (as hazard rate ratios) during the follow-up period in relation to the five risk factors (gender, age, epilepsy diagnosis, type of antiepileptic drugs and BMI).

Age group is the age of the patient’s first entry in the EHR.

KEFRI: A tool to assess risk in clinical practice

Using the risk factors in the multiple regression analysis for patients prescribed AEDs in our study, the following equation was developed to act as a clinical risk assessment index. The weighted index is named the Kalmar Epilepsy Fracture Risk Index (KEFRI).

KEFRI = Age-category x (1.18) + Gender x (-0.51) + AED-type x (0.29) + Epilepsy diagnosis-category x (0.31) + BMI-category x (-0.35)

KEFRI-stratification

In Fig 5, patients prescribed AEDs are stratified in quartiles according to the KEFRI index. Patients with an index of below 0.93 are the 25% having the lowest risk of LEFs during the follow-up period from first registry in the EHR. For these low risk patients, the 10-year risk of LEF is approximately 3%. Patients with a KEFRI index of 0.93 to 1.59 have an 8% risk, whereas patients with an index of 1.60 to 2.06 have a 10-year risk of 18%. The quartile with the highest 10-year LEF risk are those with a KEFRI index of above 2.06. For these patients, the risk is 27%.

Fig 5

Low energy fracture-free probability over time in relation to KEFRI-value (quartiles).

Discussion

There is a growing interest for generating clinical evidence using real world data from EHR. This retrospective open cohort study used such real world data to confirm previously suggested risk factors, but also used the weighted importance of these risk factors in order to create an index for assessing risk of LEFs in patients using AEDs. We found that both the type of AED and the combination with an epilepsy diagnosis are indeed important factors, supporting the claim that it is not enough to consider AED-treatment or epilepsy as a single binary risk factor, as in the updated QFracture®, let alone altogether overlook it, as in FRAX®.

Several meta-analyses have been published on the subject of fractures in epilepsy/use of AEDs [20, 21]. These reviews have assessed various studies on risks and associations in different cohorts.

Shen et al. [11] published the most recent systematic review of patients using AEDs regardless of diagnosis in 2014. Fifteen studies were included for analysis. In this meta-analysis the heterogeneity-adjusted relative risk for all fractures was found to be 1.85 (1.62 to 2.12) among AED users globally. If taking only cohort studies into account (9 studies) it was 1.97 (1.31 to 2.96). If including only osteoporosis-related fractures (3 studies) the relative risk was 1.88 (1.40 to 2.53), and among the studies only including patients >50 years (13 studies) it was 1.85 (1.61 to 2.13). When comparing AED types (inducing and non-inducing) to controls (4 and 4 studies, respectively) the differences in fracture risks were 1.60 (1.26 to 2.02) and 1.27 (1.02 to 1.59), respectively. The difference between inducing and non-inducing (4 studies) was 1.18 (1.11 to 1.25). In our study, 20 497 patients with non-inducing and 2712 with inducing AEDs were included. Overall, the patients with AEDs had a doubled unadjusted relative risk of LEFs, and we observed a 23% (13 to 35%) risk increase of LEF in patients with non-inducing AEDs, and a 63% (38 to 92%) risk increase in patients with inducing AEDs compared to controls (in the multivariate analysis) which is well in line with the meta-analysis from Shen et al.

Risks of fractures in patients with epilepsy using AEDs have been specifically reviewed [21]. The three studies included in the review showed mixed results. Two studies (one [22] classified as poor quality and one [23] of good quality) suggested no risk difference between inducing and non-inducing, while the other (classified in the review as having high quality) study found a 22% (12 to 34%) risk difference between AED types among women and a borderline difference (9% (-2 to 20%)) among men [24]. Two of the studies had relatively short follow-ups (about 2 years) and the third study follow-up time was unclear. The study having good quality which showed no AED-type differences did however find that the risk of fracturing increased with cumulative AED exposure among patients with epilepsy [23]. This is also indicated in our study where the mean follow-up time was longer than 10 years. We observed a relative risk of LEF increase of 40% (23 to 59%) among patients with epilepsy in our multivariate analysis. This is a lower risk increase compared to that of the meta-analysis by Vestergaard [14] where 5 studies of measuring any kind of fracture were included, which stated a relative risk of 2.2 (1.9 to 2.5). Our lower risk increase could be explained by the fact that we only included LEF. It is previously known that about a third of the fractures in patients with epilepsy are linked to seizures [14].

Risks in different age groups have been specifically studied. These studies indicate that risks of fractures increase with (most) AED use in children [25], adults and the elderly [26, 27]. In our study 3364 older people with AEDs of 6653 (75 years or older) were included. We observed an increased relative fracture risk of 66% (63 to 69%) in the elderly compared to those middle-aged (51–74 years). In our study there were 1681 with AEDs of 3378 children included. Very few LEF (N = 57) were documented during a follow-up of more than 10 years. This is probably due to the fact that most fractures in children are results of high energy/trauma.

Since women generally have higher risks of bone impairment, osteoporosis and fractures, fracture risk in postmenopausal women with AEDs have been assessed [28] showing a hazard ratio of 1.44 (1.29 to 1.65) compared to matched women controls without AED treatment. Gender differences in fracture risk among epilepsy patients have also been studied [23] where women have had higher fracture rates in relation to treatment duration. In our study there was a near doubling of risk of LEFs among women compared to men (57% (52 to 62%) of the risk in women in the multivariate analysis).

Although the association between Body Mass Index and fracture risk is complex [29] BMI below 25 kg/m2 is considered a risk factor for major osteoporotic fractures, when unadjusted to BMD. This was confirmed in our study even though we were only able to retrieve BMI values from half of the patients (48%) from the EHR. There was a 30% (25 to 36%) lower risk of LEF in patients with a BMI of >25kg/m2.

Our study has several limitations that should be addressed. First, our control groups were matched only by age and gender. This could possibly result in confounding effects due to differences in socioeconomic factors, hereditary risks and comorbidity. Second, we included all LEFs and AEDs and did not stratify the results in relation to fracture location or individual AED. Furthermore we did neither take multiple AED use nor AED dosage into account. Third, the few patients prescribed osteoporosis treatment (less than 2% in the population) were not excluded. We would have included further risk factors e.g. cigarette smoking and alcohol intake in our model, however in the current EHR there was no standardized way of recording these variables; thus, they were left out. KEFRI is based on merely five previously suggested risk factors and gives a crude estimation of LEF risks over time. While other risk assessment tools have integrated treatment duration in the index or have estimated risk in relation to a predetermined period (e.g. 10-year risk) [5], we found the aspect of temporal associations to be of a complex nature. Patients in our study were included from first registry in EHR and followed until first documented LEF, disenrollment (or death) or until the end of the study period, whichever came first. When including long term treatment (total AED prescription ≥ 5 years) as a risk factor, we found that it did not have a significant influence on KEFRI. The drug exposure (total days of AED prescription) was therefore used to determine the AED type categorization only among patients prescribed multiple types of AED in our study. This implies that patients could have undergone treatment breaks during the follow-up period, this in turn influencing the LEF risk. Furthermore, treatment regimens could differ substantially between diagnoses when prescribing AED perplexing the risk estimation. In spite of the limitations this study is strengthened by the fact that the entire population of patients in a Swedish region prescribed AEDs, regardless of diagnosis, were included, that the follow-up time was longer than 10 years, that we were able to look at LEFs registered by physicians rather than all fracture types, and that the multivariate analysis allowed us to use the weighted risk factors to develop a fracture risk index among the patients in the Kalmar region prescribed AEDs. To determine KEFRIs performance, the algorithm needs to be tested in an external population henceforth.

A recent review by Miziak [30] provided an update on the issue of drugs causing LEFs. Several publications stress investigation with bone mineral density (BMD) and subsequent treatment in patients using AEDs [20, 31, 32]. Despite this, awareness of the risks seem to be low among neurologists and in primary care [33], resulting in the patient group being overlooked [34]. Furthermore, the lack of fracture risk awareness seems to be low among patients with AEDs, whereby only 30% are aware of the association [35].

We think that the KEFRI could act as a future tool in clinical practice to increase the awareness among physicians regarding risks and risk factors in this specific patient group. KEFRI could possibly be integrated in the EHR as a decision support system, when assessing risks of LEFs during consultations. Furthermore KEFRI could create awareness among individual patients using AEDs through physician-patient dialogue.

Transfer Alert

This paper was transferred from another journal. As a result, its full editorial history (including decision letters, peer reviews and author responses) may not be present.

19 Jan 2021

PONE-D-20-40318

A novel index to assess low energy fracture risks in patients prescribed antiepileptic drugs- Using real world data from Swedish EHR to create the KEFRI: Kalmar Epilepsy Fracture Risk Index

PLOS ONE

Dear Dr. Nordqvist,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

The reviewers have noted some serious, overlapping concerns.  Ultimately, they converge on the model and coding exposures, covariates, and outcomes.  The revision must address these issues, and ideally provide a more complete accounting of the logic underlying the choices you've made.

Please submit your revised manuscript by Mar 05 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Robert Daniel Blank, MD, PhD

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: No

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: This study aimed to develop a fracture prediction tool for patients with epilepsy. There are several major limitations that need to be addressed prior to publication:

- The study population was chosen using a 1:1 matching design of AED users and controls by age and gender. All these factors: age, gender and AED are also risk factors used for the development of KEFRI. Such a design is not ideal for the development of a risk prediction model because the distribution of risk factors is regulated by design and almost certainly does not represent the real distribution of these risk factors in the population at risk, thus affecting generalisability and external validation.

- KEFRI index, was developed from a Cox Model but does not have a time specific baseline risk in the equation.

- There are no risk factors specific to epilepsy: number of seizures, falls

Specific comments:

How was epilepsy diagnosed? Is clinical diagnosis derived from hospitalisations, or emergency department presentations or GP/specialist consultations or all? It is not clear why people with an epilepsy diagnosis were included among controls. Please justify.

Fracture: what happened to MVA fractures or high trauma fracture? Were they excluded as event or as participant? Such details are needed in the methodology.

Variable definition:

- Suggest developing the model based on continuous rather than categorical variables, as they produce more accurate results than the categorical. If not, strong reasons should be provided for the use of categorical data.

Drug exposure:

1. How many prescriptions/days covered are needed to be included in the drug category?

2. The membership to a specific AED type needs to be explained in more details. The methodology just state that the highest number of prescriptions was used as criteria to membership to a specific AED.

More detail is needed:

How many people were classified to the second prescribed AED type?

Was there any wash-out period considered between the stop and start of new AED type?

Can it be assumed that past use of a certain AED type is equal to no use? Please explain if any consideration was given to past use of AED and how the drug intervals were created in participants who transitioned from an AED type to another.

Statistical analysis section is very brief and leaves out needed details:

- How was time of follow-up constructed? What is the start and stop of follow-up times?

- Did the authors use a paired Cox proportional hazards model to account for the case-control design?

- How was the accuracy and predictive power of the model examined?

Results

- The presence of only 16% epilepsy diagnosis among the cases seems to be a strong limitation for a study aiming to develop an index risk for fracture in this population.

- Table 1: need a p-value for the significance between each AED type medication vs controls.

- Fracture risk:

o Two of the fracture risk factors: age and gender were used for matching between AED and controls. How did this study design allow for the assessment of independent effect of age, gender and AED on fracture risk?

- Table 2: I would suggest choosing the group with lowest fracture risk as control in the analysis (i.e. male gender and BMI≤25). It will also be reasonable to collapse age into only 2 groups (younger and older than 50).

- Age and BMI should also be analysed as continuous variables with HR presented as an increment per SD.

- Replace comma with dot in the percentage column

- Table 3: A legend is needed for the colour code used to differentiate the magnitude of fracture risk according to risk factors. Please revise the use of comma for separate integer and decimal places.

- KEFRI equation needs to be revised. A baseline risk (i.e. 1-year, or 5-year) function needs to be added to all predictive models derived from Cox models. Otherwise, the equation will provide the same fracture estimates regardless of the time to fracture.

- Figure 1 needs to be edited. It is not clear who are the eligible participants in box 1 from which the dataset is extracted. The algorithm of matching AED and controls also need to be specified in this first step. I would recommend collapsing all three exclusions steps into one which will state all the exclusion criteria. A last box should indicate the number of fractures for each group.

- For Figures 2, 3, 4 the correct terminology is fracture-free probability and not proportion non-fracture. Both label and caption need to be revised for all figures.

Reviewer #2: This study uses electronic health record data from one geographic region of Sweden to develop a simple index to estimate relative risk of fractures in individuals prescribed anti-epileptic drugs (AED). The strengths of the study are that the study cohort is broad and population based, the study data tease apart the contributions of AED type and diagnosis (indication for AED) to fracture risk, and the index is simple and easily calculated from common EHR variables. A very good rationale for the study was presented in the Introduction.

There are important weaknesses and limitations that in my view need to be addressed before this is accepted for publication, detailed below.

Major Concerns

1. Important covariates were not considered for the model. Most important, prior fracture (before start of index period Jan 2008) and smoking were left out of the model. These are important confounders, since they are both associated with primary exposure variables (AED and/or epilepsy diagnosis) and outcome variable (incident fracture).

2. I am concerned about the potential for misclassification of incident fracture status, since a low energy code (W00 to W08) was also required. At least in some parts of the world, physicians are not required to indicate level of trauma in their documentation of diagnosis. If this is also true in Kalmar, it is possible that in this study many with incident fractures were misclassified as not having had a fracture. This would not be much of a concern if physicians in Kalmar are required in their documentation to indicate level of trauma for the fracture, e.g. that this is a “hard stop” that prevents the physician from completing their documentation until this step is completed. This concern would also be mollified if there is validation data available in Kalmar regarding the accuracy of these diagnosis codes.

3. I am also concerned about potential misclassification of exposure. The first sentence of the Study Cohort section indicates that those having initiated an AED before January 2008 were included. If they had discontinued the AED before the start of the study period, were they still included in the exposed group, excluded, or included in the control group?

4. If individuals discontinued AED treatment during the follow-up period, were they censored, or assigned follow-up time in the control group? This needs to be clarified in the Statistical Analysis part of the Methods.

Minor Issues

5. Absolute fracture risk estimated by FRAX is part of national Swedish guidelines for identifying candidates for osteoporosis. Figure 4 is the potentially really valuable data from this study for application of this index within the context of Swedish guidelines. I suggest the authors consider showing a supplemental table of 10-year absolute risks in subsets defined by age group, BMI, AED type, and diagnosis (similar to table 3 but showing absolute 10 year risks rather than relative risks).

6. What proportion of Region Kalmar residents are not included in this dataset? I presume that with a single-payer system this is covering the entire population and none are missed, but it would be helpful for the authors to confirm that.

7. Primary outcome section page 5; please state the skeletal sites that correspond to these fracture codes (can add site in parenthesis after each code)

8. Variable definitions page 5; please add generic drug name after each ATC code.

9. Last row table 1 page 8; indicate these are incident fractures over follow-up period, not percent with prior fracture at baseline.

10. Related to Major Concern #2; please state what level of trauma these LET codes indicate (fall from standing height or less?)

11. Define long-term medication near bottom of page 8 (you state this is ≥ 5 years in the table but it should be restated here)

12. Second to last sentence bottom of page 8, I presume the comparator group are those prescribed AED but for non-epileptic conditions? Would state that for clarity.

13. For Table 3 page 10, define reference group in a footnote (no AED, youngest age group, male, BMI <=25).

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

18 Mar 2021

Please find our response and revisions enclosed.

Submitted filename: Response to Reviewers Ola Nordqvist PLOS ONE.docx

Click here for additional data file.

12 Apr 2021

PONE-D-20-40318R1

A novel index to assess low energy fracture risks in patients prescribed antiepileptic drugs- Using real world data from Swedish EHR to create the KEFRI: Kalmar Epilepsy Fracture Risk Index

PLOS ONE

Dear Dr. Nordqvist,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please see editor and reviewer comments.

Please submit your revised manuscript by May 27 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Robert Daniel Blank, MD, PhD

Academic Editor

PLOS ONE

Additional Editor Comments (if provided):

I agree with the reviewer that the time element must be addressed. There are 2 options, as I see it:

1. revise the proposed equation to incorporate time to event

or

2. make no change to the equation but address the point prominently in the discussion.

Ultimately, it is up to readers to decide what is useful and what is not. The role of the journal is to present the readers with sufficient information about the methods so they can make informed judgments. Transparency is the key, and I will insist on that.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: No

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: No

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: I thank the authors for addressing my comments. The paper is clearer and improved.

However, there is still one point in which I disagree with the authors. I have raised the issue of using time in the KEFRI equation. The authors replied that the risk is constant over time, and this is indeed true when considering the relative risk (ie the risk of AED use vs control). However, Cox model calculates the risk within a certain time frame given the assumption that the person survived up to that point. Thus, an equation derived from Cox Model in which time is ignored, as in KEFRI, would make no difference between people who fracture at 2-year compared to those who fracture at 10-year, despite the latter having a much lower absolute risk.

Indeed, I am not aware of any risk calculator derived from a Cox model that does not have a time function in the equation. Here are a few examples of nomograms predicting absolute risks a 5 or 10-year risk1,2.

Given this considerations, I believe that in the current form KEFRI does not have a great clinical utility.

There are also a couple of minor issues:

Results –Time of follow-up should be in person-years with IQR.

Table 1: bold the significant differences between medication groups and controls

Legend 4- is confusing. I believe that the HR presented is compared to the lowest fracture risk group (men, younger than 50 and low BMI). Is this correct? A clear description of how the HR were calculated needed to be added for clarity.

References

1. Nguyen ND, Frost SA, Center JR, Eisman JA, Nguyen TV. Development of prognostic nomograms for individualizing 5-year and 10-year fracture risks. Osteoporos Int. Oct 2008;19(10):1431-1444.

2. Yang H-I, Sherman M, Su J, et al. Nomograms for Risk of Hepatocellular Carcinoma in Patients With Chronic Hepatitis B Virus Infection. Journal of Clinical Oncology. 2010;28(14):2437-2444.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

8 Jun 2021

Please find our responses to suggestions from editor and reviewer enclosed in the rebuttal letter

Submitted filename: Response to Reviewers and Editor 2 Ola Nordqvist PLOS ONE.docx

Click here for additional data file.

2 Aug 2021

A novel index to assess low energy fracture risks in patients prescribed antiepileptic drugs- Using real world data from Swedish EHR to create the KEFRI: Kalmar Epilepsy Fracture Risk Index

PONE-D-20-40318R2

Dear Dr. Nordqvist,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Robert Daniel Blank, MD, PhD

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: I thank the author for addressing all my comments. I am satisfied with authors's decision to address the temporal association in the discussion.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

18 Aug 2021

PONE-D-20-40318R2

A novel index to assess low energy fracture risks in patients prescribed antiepileptic drugs

Dear Dr. Nordqvist:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Professor Robert Daniel Blank

Academic Editor

PLOS ONE

Age categories (years):	≤ 50	=1
	51-74	=2
	≥ 75	=3
Gender:	Women	=1
	Men	=2
AED type:	Non-inducing	=1
	Inducing	=2
Epilepsy diagnosis:	No	=1
	Yes	=2
BMI-category (kg/m2):	≤ 25	=1
	> 25	=2