Introduction

Musculoskeletal pain conditions are the leading cause of disability worldwide1. Improving the quality of care for musculoskeletal pain conditions is an urgent necessity. One problem in musculoskeletal pain care is the overuse of opioids2. While overall prescribing of opioids for chronic pain has been reduced as being restricted by guidelines due to their potential harm3,4,5, the prescription of weak opioids continues to remain high6,7,8.

Tramadol is a weak mu opioid receptor agonist and serotonin and norepinephrine reuptake inhibitor9. Due to its multiple analgesic mechanisms and perception as a relatively safe opioid, the use of tramadol has been increasing rapidly around the world over the past few decades7,10,11. However, the evidence of the effects of tramadol on musculoskeletal pain conditions has been inconsistent12,13. Furthermore, there is growing evidence of multiple adverse outcomes by its mechanistic combination14,15,16,17. In particular, tramadol has been prescribed more frequently and for longer durations among older adults patients compared to younger individuals10,11,18. In older adults, chronic tramadol use has been associated with an increased risk of safety events, such as cardiovascular events, falls, hip fractures, emergency room visits, and hospitalization19.

Recent observational studies have reported that long-term opioid use increases the risk of cognitive decline or dementia20,21,22,23. Although animal studies have shown that chronic exposure to tramadol induces hippocampal atrophy and subsequent learning and memory impairment24,25,26,27, the evidence regarding the association of long-term tramadol use with dementia remains scarce. An observational study reported that the tramadol use was not associated with an increased risk of Alzheimer’s disease (AD) compared to non-use28. However, the study had several limitations, including not accounting for the competing risk of death and not controlling for the lifestyle behaviors which could be important confounders in the association between tramadol and dementia. Moreover, the exposure definition of the study was lacking the duration of tramadol usage.

Therefore, this population-based retrospective cohort study, using the Korean National Health Insurance Service (NHIS) database, aimed to investigate the association between cumulative use of tramadol and the risk of all-cause dementia, AD, and vascular dementia (VaD), in older adult patients with musculoskeletal pain. We hypothesized that longer cumulative use of tramadol would be associated with increased dementia risk.

Results

Study population and baseline characteristics

Figure 1 shows the participant selection process for this study. A total of 1,865,827 older adult patients aged 60 years or older with musculoskeletal pain and complete information were identified between January 1, 2003, and December 31, 2007. We excluded patients with cancer (n = 69,918), dementia (n = 52,583), stroke (n = 279,661), opioid use disorder (n = 554), other opioid use (n = 1,221,443), and those who died before the index date (n = 1,330). Participants who were dispensed tramadol were matched with those who were not, resulting in the exclusion of 156,412 unmatched participants. The final analytic cohort consisted of 83,926 participants, including 41,963 tramadol users and 41,963 nonusers. Table 1 shows the baseline characteristics of the study population before and after the propensity score matching. The median (interquartile range) age of the tramadol user group was 67 (63–72) years, and 56.7% were women. Standardized differences were calculated to assess the balance achieved between tramadol users and nonusers after matching. The balance of baseline characteristics across both groups improved with propensity score matching.

Fig. 1
figure 1

Selection process of participants.

Full size image
Table 1 Baseline characteristics of study participants before and after propensity score matching.
Full size table

Association of cumulative duration of tramadol dispensation with incident dementia

During the follow-up period of 14 years, the incidence rates (events per 1000 person-years) of all-cause dementia were 6.1 for nonusers, 6.2 for those with cumulative tramadol use of 1–14 days, 7.7 for those with 15–90 days of use, and 8.0 for those with > 90 days of use. The median (interquartile range) duration of follow-up for tramadol users and nonusers were 3.34 (1.24–7.08) years and 3.86 (1.36–7.86) years, respectively. Table 2 shows a dose–response relationship for all-cause dementia with 5-year cumulative duration of tramadol use (test for trend, P < 0.001). Compared with nonusers, those with cumulative use of 1–14 days (aHR 1.06, 95% CI 0.96–1.17), 15–90 days (aHR 1.14, 95% CI 1.10–1.35), and > 90 days (aHR 1.18, 95% CI 1.00–1.39) had an increased risk for all-cause dementia. Similar patterns were observed for AD, but not for VaD, as shown in Supplementary Table 1.

Table 2 Risk of all-cause dementia associated with 5-year cumulative duration of tramadol use among older adult patients with common musculoskeletal pain.
Full size table

Subgroup and sensitivity analyses

Table 3 depicts the results from a subgroup analysis of the association between tramadol use and all-cause dementia according to sex, household income, cigarette smoking, alcohol consumption, physical activity, and comorbidities. The association between tramadol use and incident dementia did not differ significantly across these stratifying variables (P for interaction > 0.05).

Table 3 Subgroup analyses for the risk of all-cause dementia associated with 5-year cumulative duration of tramadol use among older adult patients with common musculoskeletal pain.
Full size table

The sensitivity analyses which introduced the lag time window from 1 to 3 years showed results similar to the main analysis. (Supplementary Tables 2, 3, 4). Also, the analyses that defined dementia solely based on ICD-10 codes indicated no appreciable change in risk estimates for all-cause dementia, AD, and VaD (Supplementary Table 5). Furthermore, the analyses excluding tramadol users who received only one prescription also showed results similar to the main analysis (Supplementary Table 6).

Discussion

In this retrospective population-based cohort study of older adults with common musculoskeletal pain aged 60 years or older, tramadol use was associated with an increased risk of all-cause dementia and AD over 14 years of follow-up compared with nonuse, and the risk increased with increasing cumulative duration of tramadol use. Our findings remained consistent and robust across the subgroup and sensitivity analyses.

Tramadol is a weak opioid used for moderate pain in the second step of the World Health Organization pain ladder29. It also has serotonin and norepinephrine reuptake inhibitory activity, contributing to its analgesic effects9. To our knowledge, this is the first study to demonstrate that tramadol is associated with an increased risk of developing dementia. Two recent studies reported an association between long-term opioid use and increased risk of dementia22,23. One prospective cohort study of 91,307 persons using the Israeli nationwide health maintenance organizations database reported that opioid purchase more than 60 days was associated with an increased risk of dementia among individuals aged 75 to 80 compared to nonuse (aHR 1.39, 95% CI 1.01–1.92, P < 0.05)22. Another retrospective cohort study of 41,636 chronic pain patients in China reported that opioid use of 180 defined daily doses per year or more was associated with an increased risk of dementia compared to nonuse (aHR 1.86, 95% CI 1.25–2.09, P < 0.001)23. The findings from observational studies examining the association between serotonin and norepinephrine reuptake inhibitors use and dementia have been mixed30,31. Our study expanded on prior studies by exploring the association between tramadol and incident dementia and found that, tramadol prescription dispensation was associated with a higher risk of dementia. The results from our study suggest that the dementia risk associated with long-term tramadol use (aHR 1.18, 95% CI 1.00–1.39, P < 0.001) is lower than previously reported in studies involving traditional opioids.

One study presented results that differ from the results of our study. A nested case–control study of 53,990 persons with AD using the Finnish nationwide health care register database reported that tramadol prescription was not associated with an increased risk for AD (aOR 1.00 [95% CI 0.97 to 1.02])28. However, the study did not consider the competing risk of death in the analysis. Additionally, the study lacked the duration of tramadol prescription in its exposure definition and did not consider dementia other than AD as outcomes. Moreover, lifestyle behaviors such as physical activity, cigarette smoking, and alcohol consumption, which are important risk factors for dementia32, were not controlled for in the analysis. In our study, further adjustment for lifestyle-related factors did not significantly change the risk estimates, suggesting that these factors may not have a substantial confounding effect on the observed association.

The precise mechanisms through which tramadol may increase the risk of dementia have not been elucidated. In animal studies, chronic tramadol administration have been shown to induce neurodegeneration in various parts of the brain, such as hippocampus, temporal and occipital regions of cerebral cortex, cerebellum, and basal ganglia, leading to impairment in learning, memory, behavior, and motor function24,25,26,27,33,34,35. The neurotoxicity of tramadol was associated with neurodegenerative pathways such as autophagy, apoptosis, and inflammation caused by astrogliosis, microgliosis, and microglial activation24. Furthermore, multiple studies have reported that tramadol use is linked to delirium, particularly in older patients36,37, which is a risk factor for subsequent development of dementia38. O-desmethyltramadol, the potent metabolite of tramadol may cause delirium by inhibiting cholinergic neurotransmission and increasing serotonin and noradrenaline levels39. Additionally, recent observational studies have revealed a substantial link between tramadol use and cardiovascular events including stroke, myocardial infarction, heart failure, and cardiac arrhythmia14,17,19, which are powerful risk factors for dementia40. Moreover, changes in cerebral vasculature, including those affecting the microcirculation of the subcortical white matter, result in the clinical expression of cognitive impairment caused by other pathologies, including AD40. However, further research is necessary to understand the underlying mechanisms.

Despite its strengths, this study has several limitations. First, we were unable to evaluate and adjust for the severity of pain. Since tramadol is indicated for managing moderate to severe pain, confounding by indication could have affected study findings. Additionally, our exposure measure relied on the cumulative duration of the prescription dispensation of tramadol. This could instead reflect anticipated chronicity of pain. Previous studies have shown conflicting evidence regarding the effect of chronic pain on the development of dementia41,42. Second, the dispensation of a prescription does not necessarily guarantee actual medication consumption. Additionally, tramadol use status could change over the follow-up period, which were not captured in our analysis. Therefore, misclassification of tramadol exposure is possible. Third, since the outcome measurement of this study was based on recorded ICD-10 codes, there is a possibility of under ascertainment of dementia diagnoses in the registers used. Some individuals may have had undiagnosed dementia that did not yet need medical care. Additionally, because the diagnoses of AD and VaD were not based on biomarkers or neuroimaging data, discrepancies may exist between clinical diagnoses and those recorded in claims data, potentially impacting the reliability of our findings. Fourth, tramadol can induce delirium, which may obscure or delay the diagnosis of dementia in some patients, potentially leading to an underestimation of our dementia risk estimates. Fifth, we could not consider time-varying confounders. Finally, despite we considered a wide range of potential confounding factors, given the observational design of the study and the possibility for residual confounding, we could not prove causality.

In conclusion, the results of this population-based retrospective cohort study with a follow-up of 14 years suggested that the use of tramadol was associated with an increased risk of dementia among older adults aged 60 years or older with common musculoskeletal pain, and the risk increased with higher cumulative tramadol usage. Taking this potential risk into account, clinicians should carefully weigh potential benefits and risks when prescribing tramadol to older adults. Further research, including pain severity measurement and consideration of underlying pain conditions, is needed to clarify the association between tramadol use and dementia risk.

Methods

Data sources

The NHIS databases between 2002 and 2021 were used. The NHIS is a single non-profit health insurer in South Korea, providing insurance coverage to the entire population. It manages a computerized database containing healthcare-related information of individuals, which includes inpatient and outpatient reimbursement claims data, drug prescriptions, diagnoses, procedures, and treatments provided to each patient. Upon request, the NHIS provides researchers a customized database with de-identified information pertaining to the study population43. We also used the National Health Screening Program dataset. All NHIS beneficiaries 40 years or older are provided biannual mandatory health screening examinations that includes a self-reported questionnaire for lifestyle and health behavior, anthropometric measurements, blood pressure, and laboratory tests44. The validity of NHIS database has been described in detail elsewhere43,44.

This study was approved by the Institutional Review Board of National Health Insurance Ilsan Hospital, and the requirement for informed consent was waived due to the anonymized nature of the NHIS database (IRB No. NHIMC 2023-02-011). All methods were performed in accordance with the relevant guidelines and regulations of Helsinki declaration of 1964 and its amendments.

Study design and population

We conducted a retrospective cohort study of older adults aged 60 years or older who were diagnosed with common musculoskeletal pain conditions (low back pain, neck pain, shoulder pain, and knee pain) between January 1, 2003, and December 31, 2007. The International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) codes for the identification of common musculoskeletal pain conditions are provided in Supplementary Table 71,2. From this initial pool, we identified those who were or were not newly dispensed tramadol for common musculoskeletal pain conditions. ‘New dispensation’ was defined by applying a 1 year washout period (January 1, 2002-December 31, 2002), during which cases having tramadol dispensation were removed. We excluded individuals who had used opioids other than tramadol, those with a history of opioid use disorder, cancer, or stroke, and those diagnosed with dementia or died on or before the index date (January 1, 2008). The list of opioids is provided in Supplementary Table 8.

Participants who were dispensed tramadol underwent propensity score matching with those who were not. A caliper width equal to 0.2 of the standard deviation of the logit of the propensity score was used for 1:1 matching between tramadol users and nonusers45. The following variables were included in the matching process: age, sex, household income, Charlson Comorbidity Index (CCI), nonsteroidal anti-inflammatory drugs (NSAIDs) use, acetaminophen use, antidepressants use, cigarette smoking, alcohol consumption, physical activity, body mass index, systolic blood pressure, fasting serum glucose, and total cholesterol.

Exposures

The main exposure was 5 year cumulative duration of tramadol use, computed by summing all days covered by tramadol from January 1, 2003, to December 31, 2007. The cumulative duration was categorized into three groups: 1–14 days, 15–90 days, and more than 90 days. The cutoff points were determined based on the clinical interpretability and exposure distribution in the sample5. We excluded prescriptions after the index date from the analysis.

Study outcomes

We defined dementia cases as individuals prescribed acetylcholinesterase inhibitors (donepezil, galantamine, rivastigmine) or and N-methyl-D-aspartate receptor antagonist (memantine) under ICD-10 codes F00, F01, F02, F03, and G30, based on data from hospitals and primary health care inpatient and outpatient visits. To identify AD and VaD, ICD-10 codes F00 and G30 and ICD-10 code F01, were used, respectively. Prescribing acetylcholinesterase inhibitors or N-methyl-D-aspartate receptor antagonists for dementia requires evidence of cognitive impairment according to the National Health Insurance Reimbursement guidelines. Validation needs a Mini-Mental State Examination score of 26 or below, along with either a Clinical Dementia Rating of 1 or higher or a Global Deterioration Scale score of 3 or more46. Starting from the index date, the participants were followed up until the date of dementia diagnosis, date of death, or December 31, 2021, whichever came first. Additionally, the use of other opioids was applied as a censoring criterion during follow-up.

Covariates

The study population was characterized at the index date considering the following potential confounders: age (years), sex (female and male), household income (lower half and upper half, derived from insurance premiums), CCI (0, 1, and ≥ 2), NSAIDs use (days), acetaminophen use (days), antidepressants use (days), cigarette smoking (non-smoker and smoker), alcohol consumption (non-drinker, 1–2 days/week, and ≥ 3 days/week), physical activity (none, 1–2 days/week, and ≥ 3 days/week), body mass index (kg/m2), systolic blood pressure (mmHg), fasting serum glucose (mg/dL), and total cholesterol (mg/dL). The use of NSAIDs, acetaminophen, and antidepressants was quantified in cumulative days of use and adjusted as continuous variables in the analysis. The body mass index was calculated by dividing the participant’s weight in kilograms by the square of their height in meters.

Statistical analyses

We used the Fine-Gray subdistribution hazard model, accounting for the competing risks of death, to determine the subdistribution hazard ratio for the association between cumulative duration of tramadol use and incident dementia47. Three adjustment models were employed to estimate the adjusted subdistribution hazard ratios (aHRs) and 95% confidence intervals (CIs) for all-cause dementia, AD, and VaD according to the cumulative duration of tramadol use. The first adjustment model included age and sex, and the second adjustment model included household income, CCI, NSAIDs use, acetaminophen use, and antidepressants use in addition to the covariates in the first model. The third adjustment model included cigarette smoking, alcohol consumption, physical activity, body mass index, systolic blood pressure, fasting serum glucose, and total cholesterol in addition to the covariates in the second model.

In secondary analyses, we conducted a prespecified subgroup analysis and interaction tests on the cohort, categorizing by sex (female and male), household income (upper half and lower half), cigarette smoking (non-smoker and smoker), alcohol consumption (non-drinker and drinker), physical activity (physically active and physically inactive), and CCI (0, 1, and ≥ 2). We also performed prespecified sensitivity analyses. We included the lag time window from 1 to 3 years and excluded patients with outcomes that occurred during this period from the analysis to diminish reverse causation bias48. Additionally, recognizing that patients with dementia might not always be treated with medications, we conducted sensitivity analyses by defining dementia solely based on ICD-10 codes from hospitals and primary health care inpatient and outpatient visits. Furthermore, considering that individuals who continue to fill prescriptions are more likely to use the drugs compared to those who fill only one prescription, we performed sensitivity analyses excluding participants who were prescribed tramadol only once. Statistical significance was considered at P < 0.05 in a two-sided manner. All data analyses were conducted using the SAS software (version 9.4; SAS Institute Inc., Cary, NC, USA).