Statins and polyneuropathy revisited: case‐control study in Denmark, 1999–2013

Abstract

Aim

In a previous study, we found a positive association between statin use and polyneuropathy risk. Other studies reported equivocal results. The present study aimed to confirm our findings with a design similar to that used in our previous study but with a larger data set.

Methods

We searched medical registry data to identify patients diagnosed with incident polyneuropathy of no known cause (idiopathic polyneuropathy) between 1999 and 2013; we verified diagnoses through medical records. For each case, we recruited 20 general population controls with no previous history of polyneuropathy. Controls were matched to their respective case for age and gender. We ascertained the prior statin use of cases and controls through a prescription registry. Based on this information, exposure to statins was categorized into ‘ever use’ or ‘never use’. Ever use of statins was classified by how recently they had been used (‘current use’ or ‘past use’); current use was further classified into long‐term use (5+ years) and high‐ or low‐intensity use. We used conditional logistic regression to calculate odds ratios (ORs) with 95% confidence intervals (CIs) to examine associations between polyneuropathy and statin use.

Results

We included 370 validated cases and 7400 controls. Ever use of statins was not associated with an elevated risk of polyneuropathy (OR 1.14, 95% CI 0.84, 1.54). Similarly, we found no associations between polyneuropathy risk and current use (OR 1.11, 95% CI 0.79, 1.53), long‐term use (OR 1.13, 95% CI 0.66, 1.92) or high‐intensity statin use (OR 1.05, 95% CI 0.59, 1.84).

Conclusion

Statin use was not associated with an increased risk of idiopathic polyneuropathy.

What is Already Known about this Subject

• Previous studies on the association between polyneuropathy risk and statin use have reported conflicting results.

• Our previous, small Danish study found that patients who used statins long term had an increased risk of chronic polyneuropathy.

• The present, larger but similar study aimed to reassess this issue in a dataset with sufficient statistical power to detect an association between drug exposure and polyneuropathy.

What this Study Adds

• In this large, Danish registry‐based study, statin use was not associated with an increased risk of idiopathic polyneuropathy (i.e. polyneuropathy of no known cause).

• Our findings supported the notion that polyneuropathy is not a major concern among patients who use statins chronically but have no known risk factors for polyneuropathy.

Tables of Links

TARGETS
Enzymes 2
Hydroxymethylglutaryl‐CoA reductase

LIGANDS
Simvastatin	Atorvastatin
Lovastatin	Cerivastatin
Pravastatin	Rosuvastatin
Fluvastatin

These Tables list key protein targets and ligands in this article which are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY 1, and are permanently archived in the Concise Guide to PHARMACOLOGY 2015/16 2.

Introduction

The use of cholesterol‐lowering drugs, known as statins, has increased markedly in recent years. In Denmark, approximately 1% of the population used statins in 1999, and the proportion increased to more than 10% in 2014 3. Given the widespread use of statins, it is important to monitor patients to detect common and rare potential side effects. A number of studies on single cases and case series in the 1990s 4, 5, 6, 7, 8, 9 reported that statin use was a possible cause of polyneuropathy, a disease of the peripheral nerves. Polyneuropathy is estimated to affect approximately 2% of the general population, and its prevalence increases with age 10, 11. Epidemiological studies on the putative association between statin use and polyneuropathy have provided equivocal results 12, 13, 14, 15, 16, 17, 18, 19, 20, 21. We found in a previous study that statin use was positively associated with polyneuropathy risk 12. However, some laboratory studies have reported that statins provided a neuroprotective effect against peripheral nerve injury 22, 23, 24. Furthermore, it remains unclear whether hypercholesterolaemia per se has a deleterious effect on peripheral nerves 25. In the present study, we reassessed the issue of whether statin use affects polyneuropathy risk in a large dataset with more statistical power than that of our original study. To improve the statistical power, we included data collected recently over a long study period. Furthermore, to facilitate comparisons between our present and past studies, and in recognition of several methodological strengths of our previous approach, we performed the current study with a design highly similar to that of our original study.

Methods

The present case‐control study was based on data retrieved from the Odense University Hospital Registry (Patient Registry) and the Odense Pharmacoepidemiologic Prescription Database (Prescription Registry), which covered a geographically well‐defined area in Denmark, Funen County (484 346 inhabitants in 2009) 26. The civil registration number, unique to all Danish citizens, ensured correct linkage of information across registries. Codes were used to identify potential cases, drug exposure and comorbidity (Appendix 1).

Case ascertainment and validation

We retrieved information from the Patient Registry on all admissions and outpatient contacts in hospitals in Funen County, from January 1994 to December 2013. We identified patients with codes compatible with polyneuropathy of any type (Figure 1). We then restricted the sample to patients with codes compatible with potentially idiopathic polyneuropathy (i.e. polyneuropathy of no known cause), recorded for the first time in the 1999–2013 period. The date of the first identified record of any of these codes was considered the index date.

Figure 1.

Flowchart of case ascertainment

We excluded patients based on criteria observed before or within the 6 months following the index date. We included the later time frame, owing to the possibility that a comorbidity (e.g. diabetes) might be diagnosed during the polyneuropathy work‐up. The exclusion criteria were: (i) codes for polyneuropathy from a specific cause – e.g. diabetes or alcohol overuse; (ii) admission or outpatient data that included codes for diabetes, hypothyroidism, chronic kidney insufficiency, alcohol overuse or disorders related to alcohol overuse, human immunodeficiency virus, monoclonal gammopathy or cancer (except nonmelanoma skin cancer); and (iii) redeemed prescriptions for antidiabetic agents, thyroid substitution drugs or drugs for treating alcohol overuse (Prescription Registry). Furthermore, we included only patients aged 20–84 years who had resided in Funen County for at least 5 years prior to the index date. Cases identified in the last 3 months of 2013 were excluded, owing to a lack of data on residency status for that period (n = 14). Finally, we excluded patients who had been included in a previous study performed by our group (n = 1) 12. Patients selected based on these criteria were included in the present study as potential cases.

For potential cases, medical records were traced and blinded with regard to information on drug use and personal identifiers. Subsequently, a neurology trainee (T.d.K.S) manually reviewed potential cases under the supervision of a neurologist with a special interest in neuromuscular disorders (S.H.S.). The criteria for accepting a diagnosis of idiopathic polyneuropathy were predefined (Appendix 2). We also collected the dates of symptom onset and diagnosis. In analyses restricted to validated cases, the index date was defined as the earliest of the following: date of symptom onset, date of diagnosis or date of first identified hospital entry of a polyneuropathy code.

Controls

For each potential case, we identified controls from the general population residing in the catchment area. Controls were identified in a local database, which included a demographic module with vital status and history of migration (i.e. dates of moves to and from the geographically well‐defined catchment area) for all residents of the region 26. Controls were selected when, on the index date, they matched a potential case in terms of age (year of birth) and gender. From these eligible controls, after applying the exclusion criteria mentioned above, we retained 20 controls per case. Some cases could be selected as controls prior to their diagnosis date (index date). Thus, the odds ratios (ORs) generated in the present study were unbiased estimates of the incidence rate ratios that would have emerged from a cohort study based on the same source population 27.

Drug exposure

The Prescription Registry provided complete data on all reimbursed prescriptions filled at community pharmacies by residents of Funen County, Denmark, since November 1992 26. The data included the date of dispensation and a full product description, including the anatomical therapeutic code 28, the number of dose units, the dosage form and the total number of defined daily doses (DDDs) contained in the dispensed package. The DDD was defined as the average maintenance dose per day assumed for a drug used for its main indication in adults (for statin DDD values, see Appendix 1) 28. No information was available in the Prescription Registry on the indication for use or the dose prescribed by the physician.

We retrieved from the registry all available information on statin use for all subjects. We classified subjects by their use of statins prior to the index date as follows: ever users (≥1 statin prescriptions) or never users (no statin prescriptions). Subjects were further classified by how recently these drugs had been used, the duration of use and the intensity of use (see Supporting Information for details). We calculated the duration and intensity of statin use for the group with ever use (with the hypothesis that statins have a cumulative effect) and for the group with only current use (with the hypothesis that statins have a more immediate effect).

Analyses

We used conditional logistic regression to calculate the ORs and 95% confidence intervals (CIs) of statin exposure for cases with polyneuropathy compared with control subjects. The ORs were adjusted for the effects of several potential confounders, by design (i.e. matching patients and controls for age, gender and calendar period) and by restriction (i.e. exclusion of subjects with known polyneuropathy risk factors), as described above. Analyses were performed for the group with ever use and groups with current/past use; the group with never use was the reference group. During the study period, simvastatin was the dominant statin, accounting for 74% of total sales 3. Analyses on other single drugs were performed when the numbers permitted.

Separate analyses were performed for the following types of cases and their corresponding controls:

1. Potential cases: data based purely on registry information.

2. Cases: cases verified through medical records (criteria in Appendix 2), including:

   1. All cases
   2. Definite cases
   3. Probable or possible cases.

We also performed some supplementary analyses. Based on all available registry data on potential cases and their corresponding controls, we classified subjects according to the type of lipid‐lowering drug used, and we identified the most frequent patterns of lipid‐lowering drug use. We excluded subjects exposed to various drugs reported to increase the risk of polyneuropathy (see Appendix 1), and we restricted the sample to cases defined according to narrow criteria (see Appendix 2), before performing the main analyses. Finally, because both quantitative sensory testing (QST) and skin biopsies had been implemented gradually in our department during the study period, we also analysed verified cases with either skin biopsy or QST results compatible with polyneuropathy, despite no indication of the diagnosis based on nerve conduction studies; this group of patients was not included in the main analyses (Figure 1).

The study was approved by the Danish Data Protection Agency. According to Danish law, approval from an ethics board was not required for registry studies 30.

Results

We identified 3104 patients in the catchment area with a first hospital contact code indicative of polyneuropathy in 1999–2013. Of these, 1146 were classified as potential cases. Statins were, by far, the most frequently used lipid‐lowering drugs among these potential cases and their controls (Table S1). The diagnosis of idiopathic polyneuropathy could be validated in 370 cases, which were included in the main analyses. Another 56 patients with normal nerve conduction studies showed evidence of small fibre neuropathy on a QST or skin biopsy; these patients were only included in supplementary analyses. The most common reasons for study exclusion (n = 720) were a failure to confirm the diagnosis (n = 341) and an identification of non‐idiopathic polyneuropathy (n = 308) (Figure 1).

The characteristics of validated cases included in the main analyses are presented in Table 1. Analyses based on these cases indicated no associations or weak associations between statin use and polyneuropathy risk (Table 2). The group with ever use of statins showed an OR of 1.14 (95% CI 0.84, 1.54) and the group with current statin use showed an OR of 1.11 (95% CI 0.79, 1.53). A restriction to definite cases had little influence on these results (ever use: OR 1.30, 95% CI 0.88, 1.92; current use: OR 1.28, 95% CI 0.84, 1.95). Among the subset of definite cases, the highest estimates were observed for subjects with current statin use for only 1 year or less (OR 1.78, 95% CI 0.81, 3.94) or with a cumulative statin dose of 500–1000 DDD (OR 1.73, 95% CI 0.85, 3.53) (Table 2).

Table 1.

Characteristics of cases with idiopathic polyneuropathy

Characteristic	Statin users (n = 60)	Non‐users (n = 310)
Male gender	47 (78.3)	184 (59.4)
Age, median (interquartile range)	71.1 (64.4‐76.4)	64.5 (55.5‐72.6)
Pain, altered sensation, or numbness	55 (91.7)	287 (92.6)
Absent tendon reflexes a	43 (71.7)	218 (70.3)
Clinical classification
Motor	0	1 (0.3)
Sensory	44 (73.3)	203 (65.5)
Sensorimotor	10 (16.7)	79 (25.5)
Unclassifiable	6 (10.0)	27 (8.7)
Neurophysiology
Mainly axonal	57 (95.0)	269 (86.8)
Axonal and demyelinating	0	20 (6.5)
Demyelinating	1 (1.7)	3 (1.0)
Unclassifiable b	2 (3.3)	18 (5.8)

Statin exposure asserted prior to the index date defined as the earliest of the following: date of symptom onset, date of first hospital contact with polyneuropathy clinical suspicion, or date of diagnosis. Date of symptom onset was available for 282 cases (76%), 40 (67%) statin users and 242 (78%) statin non‐users.

^aInformation missing on six statin users and 33 statin non‐users; all but one classified as probable or possible idiopathic polyneuropathy.

^bThe conclusion of the nerve conduction studies (i.e. results compatible with polyneuropathy) was cited in the medical records but the neurophysiology test results could not be traced.

Table 2.

Use of statins and risk of idiopathic polyneuropathy

	Based on all cases			Based on definite cases only
Statin use	Casesa , b (n = 370)	Controlsb (n = 7400)	Odds ratioc (95% CI)	Casesa , b (n = 238)	Controlsb (n = 4760)	Odds ratioc (95% CI)
Never use	310	6,309	1 (reference)	201	4,153	1 (reference)
Ever use	60	1,091	1.14 (0.84, 1.54)	37	607	1.30 (0.88, 1.92)
Duration, years
≤1	15	247	1.25 (0.73, 2.16)	9	140	1.36 (0.68, 2.74)
>1, ≤3	18	302	1.23 (0.75, 2.03)	11	187	1.26 (0.66, 2.39)
>3, ≤5	7	223	0.65 (0.30, 1.40)	6	119	1.08 (0.46, 2.52)
>5	20	319	1.30 (0.80, 2.11)	11	161	1.48 (0.77, 2.84)
Intensity of use, tertiles d, e
Low	18	359	1.01 (0.62, 1.66)	13	200	1.36 (0.75, 245)
Middle	19	319	1.20 (0.73, 1.97)	10	175	1.20 (0.61, 2.35)
High	15	327	0.93 (0.54, 1.60)	10	186	1.13 (0.57, 2.21)
Past use	12	188	1.25 (0.67, 2.30)	7	105	1.39 (0.62, 3.10)
Current use	48	903	1.11 (0.79, 1.53)	30	502	1.28 (0.84, 1.95)
Duration, years
≤ 1	9	147	1.23 (0.62, 2.45)	7	81	1.78 (0.81, 3.94)
>1, ≤3	16	267	1.26 (0.75, 2.15)	9	166	1.16 (0.58, 2.34)
>3, ≤5	7	198	0.74 (0.34, 1.61)	6	105	1.23 (0.53, 2.91)
>5	16	291	1.13 (0.66, 1.92)	8	150	1.15 (0.54, 2.43)
Intensity of use, tertiles d, e
Low	14	306	0.92 (0.53, 1.61)	11	171	1.31 (0.69, 2.48)
Middle	15	280	1.08 (0.63, 1.87)	6	159	0.79 (0.34, 1.82)
High	14	282	1.05 (0.59, 1.84)	10	156	1.41 (0.72, 2.79)
Cumulative dose e
<500	10	226	0.92 (0.48, 1.76)	8	140	1.19 (0.57, 2.47)
≥500, <1000	12	220	1.18 (0.65, 2.16)	9	117	1.73 (0.85, 3.53)
≥1000, <2000	9	240	0.81 (0.41, 1.60)	6	132	1.03 (0.44, 2.40)
≥2000	12	182	1.38 (0.74, 2.56)	4	97	0.88 (0.31, 2.48)

^aPatients with a diagnosis of polyneuropathy verified through medical records.

^bSubjects with any of the following were excluded: admission or outpatient contacts with codes for diabetes, hypothyroidism, chronic kidney insufficiency, alcohol overuse or disorders related to alcohol overuse, human immunodeficiency virus, monoclonal gammopathy or cancer (except nonmelanoma skin cancer); claimed prescriptions for antidiabetic agents, thyroid substitution drugs or drugs used to treat alcohol overuse.

^cAdjusted for age, gender and calendar year through matching.

^dStatin use among controls used to determine cut‐off values.

^eOnly subjects with 2+ statin prescriptions in period of interest used to calculate intensity of statin use and cumulative dose.

We also performed a number of supplementary analyses. Analyses based exclusively on registry data (i.e. potential cases and their controls) showed slight to moderate associations between risk of polyneuropathy and some measures but the results were inconsistent (Table 3). A restriction to cases that fulfilled the narrow definition of a neuropathy diagnosis had little impact on the estimates (Tables S2). Additionally, risk estimates were similar when the analyses were restricted to subjects with no previous exposure to other drugs associated with an increased risk of polyneuropathy (Table S3). Analyses of the subset of 56 cases with paraclinical test results compatible with small fibre neuropathy and their 1120 controls showed an OR of 1.65 (95% CI 0.65, 4.24) for subjects currently using statins and an increased risk of neuropathy (OR 4.20, 95% CI 1.28, 13.79) for subjects with past statin use (Table S4).

Table 3.

Use of statins and risk of polyneuropathy based exclusively on register data

Statin use	Potential casesa , b (n=1,146)	Controlsb (n=22,920)	Odds ratioc (95% CI)
Never use	928	19,339	1 (reference)
Ever use	218	3,581	1.32 (1.12‐1.55)
Duration, years
≤ 1	63	834	1.61 (1.23‐2.11)
>1, ≤3	56	907	1.32 (1.00‐1.76)
>3, ≤5	33	693	1.02 (0.71‐1.47)
>5	66	1,147	1.25 (0.95‐1.64)
Intensity of use, tertiles d, e
Low	65	1,080	1.27 (0.98‐1.67)
Middle	55	1,081	1.09 (0.82‐1.45)
High	79	1,112	1.53 (1.19‐1.97)
Cumulative dose e
<500	70	950	1.56 (1.20‐2.00)
≥500, <1000	36	734	1.05 (0.75‐1.49)
≥1000, <2000	47	878	1.14 (0.84‐1.55)
≥2000	46	711	1.39 (1.00‐1.91)
Past use	50	637	1.68 (1.23‐2.28)
Current use	168	2,944	1.24 (1.03‐1.48)
Duration, years
≤ 1	41	528	1.64 (1.18‐2.28)
>1, ≤3	40	772	1.13 (0.82‐1.57)
>3, ≤5	26	599	0.94 (0.62‐1.40)
>5	61	1,045	1.26 (0.95‐1.67)
Intensity of use, tertiles d , e
Low	47	886	1.12 (0.83‐1.53)
Middle	46	925	1.07 (0.79‐1.47)
High	68	992	1.51 (1.15‐1.97)
Cumulative dose e
<500	53	700	1.60 (1.20‐2.14)
≥500, <1000	26	645	0.88 (0.59‐1.31)
≥1000, <2000	37	803	0.98 (0.70‐1.38)
≥2000	45	655	1.51 (1.09‐2.09)

^aPatients recorded with first time ever diagnosis code of polyneuropathy in Patient Registry, i.e., unverified register diagnosis.

^bSubjects with any of the following were excluded: admission or outpatient contacts with codes for diabetes, hypothyroidism, chronic kidney insufficiency, alcohol overuse or disorders related to alcohol overuse, HIV, monoclonal gammopathy, or cancer (except non‐melanoma skin cancer); claimed prescriptions for antidiabetics, thyroid substitution drugs, or drugs used to treat alcohol overuse.

^cAdjusted for age, sex, and calendar year through matching.

^dStatin use among controls used to determine cut‐off values.

^eOnly subjects with 2+ statin prescriptions in period of interest used to calculate intensity of statin use and cumulative dose.

Discussion

We found no evidence of an association between statin use and the risk of idiopathic polyneuropathy in the present study. We expect this finding to be highly reassuring both for patients and care providers.

Our findings are most strikingly at odds with the results of our own previous study, conducted in the same setting, where we found a strong association between statin use and idiopathic polyneuropathy (e.g. OR 4.6, 95% CI 2.1, 10‐0, for patients who currently used statins) 12. It is unlikely that design issues could explain the differences between our previous and current results because we used identical data sources and highly similar designs in the two studies. Furthermore, interobserver variation could only account for a minor part of the discrepancy in study results because in both studies we used similar predefined case definitions, assessed medical record information blinded for medication exposure, and provided supervision by the same expert in neuromuscular disorders (S.H.S.). However, the present study included information on more than twice as many cases as our previous study, and the subjects had much higher levels of statin exposure. Consequently, the present study had much higher precision than the former study, as reflected by the ratio between the upper and lower confidence limits of our main estimate (4.2 in the former study and 1.8 in the present study). Thus, although chance findings may be at play, the present study was less likely than the former study to have been affected by chance. Of note, we found no clear dose–response patterns that pointed to an association between statin use and the risk of idiopathic polyneuropathy. This negative finding further supported the lack of a causal association between exposure to statins and the development of idiopathic polyneuropathy. Our former study 12 was criticized for not being sufficiently rigorous because the subjects did not exclusively include strictly distal and symmetrical cases of polyneuropathy 30. Despite that criticism, we chose to retain the same broad criteria in the present study to facilitate a straightforward comparison with our former study. Moreover, the ability to apply anatomical criteria was somewhat limited with our retrospective design because symptoms and findings were based on medical record information with variable completeness. We note that, in the present study, including only cases with distal symmetrical symptoms clearly stated in the medical records (see Appendix 2) had little impact on the results. The higher exposure frequency among controls compared with that in the previous study provided the present study with more power; however, it also implied that, in recent years, statins might have been prescribed for patients in a wider array of categories, compared with the subjects included in our original study. Thus, the strong signal we observed in our first Danish study may have indicated that statin use was a strong marker for risk factors of polyneuropathy – i.e. confounding factors associated with dyslipidaemia or other cardiovascular risk factors 25. In recent times, statins have probably been prescribed more liberally 31 and may thereby be less strongly tied to particular patient characteristics. Indeed, in the present study, past statin use was more strongly associated with polyneuropathy risk than was current statin use. Moreover, an appreciation of the results of our former study may have led clinicians to discontinue statins at the first sign of polyneuropathy. The delays in referrals and diagnoses for this disorder could have led to the observed association between a first diagnosis of polyneuropathy and past statin use. Nevertheless, we consider this explanation unlikely because we received only a small number of referrals that specifically mentioned statin exposure as a possible cause of polyneuropathy (Gaist and Sindrup, personal communication). Likewise, although acute polyneuropathy has been reported in connection with statin use 32, we believe that this potential side effect is extremely rare and unlikely to explain our findings concerning past statin use. Overall, we consider our current findings to be consistent with those of other studies that have found slight to moderate associations 14, 16 or no association 19, 20 between statin use and polyneuropathy. Consequently, we strongly urge that our results of a potential association between statin use and small fibre neuropathy, which were based on small numbers, should only be regarded as hypothesis generating.

The present study had several strengths. First, we used population‐based registries to screen for potential cases and to choose general population controls. Second, because the information on drug exposure in our study was based on registry data, we eliminated recall bias. Third, we only accepted cases where a polyneuropathy diagnosis was confirmed by the results of nerve conduction studies.

The study also had a number of potential limitations. In an effort to reduce confounding, we restricted cases to patients with no known risk factors for polyneuropathy – such as diabetes and high alcohol intake. Therefore, our results on the influence of statins on the development of polyneuropathy may not be generalizable to all patients with polyneuropathy. Furthermore, because the symptoms and signs of chronic polyneuropathy develop insidiously, a longer time frame than that employed in the present study might be necessary to capture an association between statin use and slowly evolving statin‐induced polyneuropathy; these cases would not have been ascertained in the present study. In addition, we may have missed some patients with polyneuropathy, either because they were not suspected of having the disorder 33 or because they were referred to local private practice specialists. However, in the catchment area studied, nerve conduction studies are primarily conducted at our hospital; therefore, we assumed that the present study included the majority of recognized polyneuropathies in this area. More importantly, we did not suspect systematic biases related to where patients were examined for diagnosing polyneuropathy because all medical healthcare in Denmark is tax financed and free of charge, independent of income. However, we could not rule out the possibility that, in some instances, polyneuropathy might have been diagnosed without using neurophysiology tests; this possibility may particularly apply to mild cases of the disorder diagnosed in nonhospital settings. Nevertheless, this type of misclassification should not be related to whether a person was treated with or without statins.

Finally, because the present study was observational, we could not rule out the effects of poorly measured or unmeasured confounders.

In conclusion, the present study provided evidence that statin use is not associated with the risk of idiopathic polyneuropathy (i.e. polyneuropathy of no known cause).

Competing Interests

All authors have completed the Unified Competing Interest Form at http://www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: T.d.K.S., P.N.H. and L.A. have no disclosures. L.A.G.R. is employed by CEIFE, which has received research grants from Bayer Pharma AG, Germany, and he has also served as an advisory board member for Bayer Pharma AG, Germany; J.H. and S.H.S. have received grants from Pfizer unrelated to this work; D.G. has received fees from Astra Zenica (Sweden) for participation as a coinvestigator in a research project unrelated to this work.

This study received support from Odense University Hospital (OUH). Dr Gaist was funded by OUH. Dr Svendsen was funded by OUH, the Region of Southern Denmark, and the University of Southern Denmark.

Contributors

D.G. and T.d.K.S. conceived and designed the study; acquired, analysed and interpreted data; and drafted the manuscript. L.A.G.R. and J.H. designed the study; acquired, analysed and interpreted data; and critically revised the manuscript. S.H.S. conceived and designed the study; acquired, analysed and interpreted data; and critically revised the manuscript. P.N.H. and L.A. acquired and interpreted data; and critically revised the manuscript. All authors provided approval to publish the final version of the manuscript, and agreed to be accountable for all aspects of the work and to ensure that questions relating to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Supporting information

Table S1 Use of lipid‐lowering drugs among potential cases and controls

Table S2 Use of statins and risk of idiopathic polyneuropathy based on narrow definition of neuropathy diagnosis

Table S3 Use of statins and risk of idiopathic polyneuropathy All cases and controls with use of other drugs associated with increased risk of polyneuropathy excluded (see Appendix 1 for list of drugs)

Table S4 Use of statins and risk of idiopathic polyneuropathy. Cases restricted to patients with polyneuropathy paraclinically verified through QST test or skin biopsy (i.e. nerve conduction study not compatible with neuropathy)

Appendix 1.

Codes used in this study to identify potential cases, statin exposure and comorbidity

Hospital discharge codes according to International Classification of Diseases (ICD)‐10
Polyneuropathy
Polyneuropathy, potential idiopathic cases
G60.3 – progressive idiopathic neuropathy
G60.8 – other hereditary and idiopathic neuropathies
G60.9 – unspecified hereditary and idiopathic neuropathies
G62.0 – drug‐induced polyneuropathy
G62.9 – unspecified polyneuropathy
Polyneuropathy, ‘other’
G60‐G63 codes excluding G60.3, G60.8, G60.9, G62.0, G62.9
Other disorders (covariates)
E10‐E14 – Diabetes mellitus
E03 – Hypothyroidism
N18, N19 – Chronic kidney insufficiency
F10, G31.2, G62.1, G72.1, I42.6, K29.2, , K70, K86.0, R78.0, T51, Z72.1 – alcoholism and related disorders
D47.2 – Monoclonal gammopathy
B20‐B24 – Human immunodeficiency virus infection and disorders related to HIV infection
C1‐C43, C45‐C97 – Cancer, excluding nonmelanoma skin cancer

Anatomical therapeutic classification codes
Statins (defined daily dose value)
C10AA01 – Simvastatin (30 mg)
C10AA02 – Lovastatin (45 mg)
C10AA03 – Pravastatin (30 mg)
C10AA04 – Fluvastatin (60 mg)
C10AA05 – Atorvastatin (20 mg)
C10AA06 – Cerivastatin (0.2 mg)
C10AA07 – Rosuvastatin (10 mg)
Other drugs (covariates)
A10 – Antidiabetics
H03AA – Thyroid substitution drugs
N07BB – Drugs used to treat alcohol dependency
Other drugs associated with increased risk of polyneuropathy
A02BA01 – Cimetidine
A11HA02 – Pyridoxine
C01BD01 – Amiodarone
C02DB – Hydralazine
D01BA01 – Griseofulvin
J04AC, J04AM, J04AK02 – Isoniazid and/or rifampicin
J04BA02 – Dapsone
J01XE01 – Nitrofurantoin
L01 – Chemotherapy
L04AX02 – Thalidomide
M01CB – Gold
M04AC01 – Colchicine
N03AB02 – Phenytoin
P01BA01 – Chloroquine
P01AB01, J01XD01 – Metronidazole
S01AA01 – Chloramphenicol

Appendix 2.

Diagnostic criteria

Criteria for diagnosis of polyneuropathy Distal sensory symptoms, motor symptoms or both Symmetric symptoms in lower extremities Bilateral diminished or absent ankle reflexes Nerve conduction test findings: abnormal conduction (velocity or compound action potential) in two or more peripheral nerves, at least one of which being a leg nerve. When nerve conduction test results cannot be retrieved, the medical record has to report that the test results were compatible with a peripheral neuropathy diagnosis. Skin biopsy compatible with polyneuropathy Quantitative sensory testing compatible with polyneuropathy Broad definition: Criterion Ia and at least one item in criteria II Narrow definition: All items in criteria I and at least one item in criteria II

Criteria for ascertaining the cause of polyneuropathy Verified cases of polyneuropathy were classified according to the completeness of information on the potential causes of the disorder; i.e. the degree of confidence that the cause was idiopathic: Definite: sufficient information to rule out all exclusion criteria Probable: information only sufficient to rule out diabetes, alcohol overuse, renal insufficiency and cancer Possible: insufficient information to rule out diabetes, alcohol overuse, renal insufficiency and cancer

Svendsen, T. K. , Nørregaard Hansen, P. , García Rodríguez, L. A. , Andersen, L. , Hallas, J. , Sindrup, S. H. , and Gaist, D. (2017) Statins and polyneuropathy revisited: case‐control study in Denmark, 1999–2013. Br J Clin Pharmacol, 83: 2087–2095. doi: 10.1111/bcp.13298.