香京julia种子在线播放

This study was approved by the “Centre d’épidémiologie sur les causes médicales de décès (CépiDc)–Institut national de la santé et de la recherche médicale (INSERM),” the French data protection authority (“Commission Nationale de l’Informatique et des Libertés”-CNIL), and by the local Ethics Committee (IRB00013412, “CHU de Clermont-Ferrand IRB #1,” IRB number 2022-CF039) with compliance to the French policy of personal data protection.

The data were obtained from the EGB (“Echantillon généraliste des Bénéficiaires”) databases created in 2005 and incorporating several databases such as reimbursements and consumption of care, hospitalizations and consultations in hospitals and specialized medical units, and long-term illness (LTI). The EGB databases represent 1/97th of the population covered by the French national health insurance system (~99% of the French population), i.e., about 700,000 individuals (Bezin et al., 2017).

A control population (without chronic pain) and without dementia history or disorders, was generated by a 3:1 (3 controls for 1 chronic pain) propensity score matching a caliper of 0.005 (using the “PROC PSMATCH” procedure on the SAS Enterprise Guide software). The propensity score incorporated age, sex, comorbidity index [Charlson score, incorporating several chronic disease (Charlson et al., 1987) see Supplementary material for details], and inclusion date (called “index date” as start of follow-up identical to matching individual with chronic pain). The absence of chronic pain was determined by the absence of a diagnosis of chronic pain and the absence of continuity of analgesic treatment for more than 3 months (all analgesics combined).

A chronic pain condition was identified over the period 2006–2010 by the Anatomical Therapeutic Classification (ATC) code of recommended treatment for chronic pain (N02, M01, M02A, N03AX12, N03AX16, N06AA04, N06AA09, N06AX16, N06AX21, N01BB02, and N01BB52), with at least 6 months duration of continuous analgesic prescription (all analgesics combined), or by the medical diagnoses of chronic pain according to the International Classification of Disease (ICD-10-R521, R522, M797) or by the LTI for chronic pain (ICD-10 - R521, R522, R529, M00-M09, M10-M19, M255, M353, M40-M49, M50-M54, M79, M890, G43, and G442), as published previously (Chenaf et al., 2018; Delorme et al., 2021). To avoid chronic pain misidentification, patients with a history of mental health disorders (ICD-10 codes F00 to F99) or epilepsy (G40 and G41) were excluded from the analyzes since antidepressants or antiepileptic drugs, used as first and second line analgesics in chronic pain, are also commonly used in these neurological disorders. For code details, see Supplementary material. All individuals aged 50 years and older with chronic pain between 01/01/2006 and 12/31/2010 were included.

Chronic pain had to be incident (no diagnosis or 6-month continuous analgesic treatment before inclusion). The index date was defined as the date of chronic pain onset (diagnosis or end of the first 6-month continual period of analgesic treatment) starting the follow-up of individual. To avoid an immortality bias due to the 6-month period of analgesic treatment, control individuals who developed ADRD during the first 6 months of follow-up were excluded and replaced (matching was rerun until a ratio of 3 controls to 1 case was reached for all individuals analyzed).

An ADRD condition was identified over the period 2006–2019 by the ATC code of recommended treatment for dementia (N06DA and N06DX) or medical diagnoses according to the ICD-10 or LTI for ADRD (F00, F01, F03, G30, G310, G311 and G319), as published previously (Rochoy et al., 2019b; Elyn et al., 2022). For code details, see Supplementary material. To confirm the presence of ADRD, individuals had to have at least two different dates of diagnosis, or two dispensations of anti-dementia treatment, or have a diagnosis associated with a dispensation of anti-dementia treatment. Diagnosis codes of “dementia with known cause” (F02), such as Parkinson’s disease and toxic dementias, were not considered related to ADRD and are analyzed as comorbidities. Finally, the duration of follow-up was distinguished in three cases:

Pain medication exposure was assessed with dispensation claims recorded in the EGB database. All recommended treatments for chronic pain were considered during the study period. Exposure was described according to cumulative dose, combining both the duration of treatment and the somewhat variable daily dose, for each person, each molecule and each therapeutic class. The cumulative dose was computed during the study time window and converted it into a number of defined daily doses (DDDs) by dividing it by the average recommended daily dose for this product, as previously published by Billioti de Gage et al. (2014). Therefore, one DDD corresponded to an average one-day exposure. Exposure was ascertained up to 13 years follow-up (maximum duration of follow-up) before the onset of ADRD. Cumulative doses were categorized according to level of exposure (null = 0 DDD; mild: DDDs <180 days; moderate: DDDs ≥180 days and ≤ median distribution; and strong: DDDs > median distribution). The median distribution of each molecule and therapeutic class was calculated among people with DDDs ≥180 days. People without any claim for pain medications during the study time window served as references for the analyzes. Therapeutic classes corresponded to “NSAIDs”(M01A), “analgesics” (N02B), “antirheumatic”(M01B and M01C), “antiepileptics” (N03AX12 and N03AX16), “Serotonin-Norepinephrine Reuptake Inhibitors - SNRI”(N06AX16 and N06AX21), “TriCyclicAntidepressant - TCA”(N06AA04 and N06AA09), “triptan” (N02C) and “opioids”(N02A). Treatments used topically and only in medical settings were not included in this analysis.

After case identification and characterization (gender, year of chronic pain onset, age at onset, and Charlson score at onset), controls were selected using 3 nearest neighbor propensity score matching. The propensity score was computed using logistic regression with the group (case / control) as dependent variable, and gender, index date, age and Charlson score as independent variables. Groups were described after matching using frequency and associated percentage for categorical data and mean ± standard deviation, median, interquartile range and minimum-maximum for continuous data (Table 1). The person time incidence rate of ADRD (number of patients developing ADRD / follow-up time) is presented by groups (cases vs. controls) according to gender and age category in Table 2. To take into account the heterogeneity of follow-up durations, we analyzed ADRD onset using survival methods. The event of interest was defined as the onset of ADRD. Because of the advanced age of our study population and thus the significant risk of death, the latter was considered as a competitive event. Time to event was computed as the time from index date to event or latest follow-up (censor). Survival curves were plotted using the Kaplan–Meier method for crude survival (death), cumulative incidence (ADRD + death), and development of ADRD with death as a competing risk (as defined by Fine and Gray; Fine and Gray, 1999). In the univariate analysis, Log-rank and Gray’s tests were used to compare survival curves between chronic pain and control populations. In the multivariable analysis, an adjusted Fine and Gray regression model was performed to assess the risk of ADRD development (with death as competing event), adjusted on factors that were selected upon their clinical relevance and their statistical significance in univariate analysis. Moreover, collinearity between covariate candidates was carried out and in the case of high collinearity, a clinically driven decision was made to choose the covariate to be kept in univariate and multivariable models. This was the case in our study for opioid and NSAID consumption. Thus, only age, sex, hypertension, diabetes, hyperlipidemia, depression, stroke, head injury, Parkinson’s disease, epilepsy, dementia with known cause, and benzodiazepines use were selected for adjustment in the multivariable model. The main factor of interest (chronic pain) was also adjusted, considering groups (chronic pain vs. controls) that were matched using propensity score (see above).

An univariate exploratory analysis of the presence of ADRD (regardless of the delay of onset) in the subgroup of chronic pain patients, taking into account the time of exposure to the different pain medications (according to molecules, therapeutic classes, and the DDDs), was performed to quantify the relationship between ADRD and medication exposure. This analysis was carried out only for exposures to each molecule/therapeutic class with a minimum of 400 individuals, as recommended for odds ratio calculations(Bewick et al., 2005).

The results are shown by frequency, associated percentage and odds ratio with their 95% confidence interval computed in univariate logistic regression (risk of ADRD according to exposure level). Statistics were computed using the SAS software. Tests were two-tailed and a p value <0.05 was considered statistically significant.

During the inclusion period (2006–2010), 15,136 individuals over 50 years old and with chronic pain were identified. After matching, 13,596 chronic pain and 40,788 control individuals were analyzed. As expected from the matching procedure, cases and controls were comparable in terms of age (66 years), sex (62% female), and Charlson score (0.4). Chronic pain and control individuals had a median follow-up time of 10.2 years. For the two groups and with similar proportions, the most common comorbidities were cancer, cerebrovascular disease, heart disease, and diabetes (Table 1). The vast majority of people with chronic pain were undergoing therapeutic management of their pain (94%), but the precise etiology of the chronic pain was not known in 82% of cases. Therapeutic pain management included non-opioid analgesics (~85%, mainly paracetamol), opioids (64%, mainly dextropropoxyphene), and anti-inflammatory drugs (topical and oral, 58 and 74% respectively; Table 3).

Survival analyzes with the Kaplan–Meier method showed that the death rate was greater (p < 0.001) in the chronic pain group [2,777 (20.4%)] than the control group [5,644 (13.8%)]. The same result was observed for the ADRD rate (and ADRD + deaths; Figure 1).

Therefore, Kaplan–Meier survival curves (univariate analysis), according to the Fine and Gray model with death as competing factor, were obtained (Figure 2). Among people with chronic pain, 1,416 (10.4%) developed ADRD with a median time to onset of 3.8 years (1.6–7.2), for a median age of onset of ADRD at 81 years, and an incidence rate of 11.3 cases per 1,000 patient-years. In control people, 3,692 (9.1%) developed dementia with a median time to onset of 4.1 years (2.1–6.8), with a median age at onset of ADRD at 82 years. The comparison of incidence between groups showed that control people had a lower incidence rate than people with chronic pain, with 9.5 cases per 1,000 patient-years (Gray’s test 18.67 and p < 0.0001). The analysis of ADRD onset in survival analysis, taking death as competing event, showed that people with chronic pain are at greater risk of ADRD onset (Gray’s test p < 0.0001).

Sensitivity analysis showed that the incidence of ADRD was higher for women (12.5 cases per 1,000 person-years vs. 9.2 cases for men) and increased significantly with age (50–59 year-olds, 1.6–2.2 cases per 1,000 person-years, +90 year-olds: 54.8–76.9 cases per 1,000 person-years, and according to sex; Table 2).

Univariate analysis of confounding factors related to the development of ADRD showed that chronic pain increased the risk of developing ADRD [HR = 1.15 (1.08–1.22), p < 0.0001]. The same observation was made for confounders, such as Parkinson’s disease, dementia with known cause, stroke, benzodiazepine use, head injury, female sex and hypertension. In contrast, cancer, lung disease, and hyperlipidemia were protective against the development of ADRD (Figure 3). All of these confounders were therefore used for the adjusted model. Diabetes, depression, and epilepsy had no effect on the incidence of ADRD.

Multivariable analysis confirmed the previous results with an HR of 1.23 [1.15–1.32] of developing ADRD for the chronic pain group (Figure 4).

Exploratory analyzes were performed to assess the relationship between exposure to classes of pain medications [NSAIDs, opioids, non-opioids (mainly paracetamol), antiepileptics, tricyclic antidepressants, serotonin and noradrenaline reuptake inhibitors, and anti-rheumatic] and the presence of ADRD. The analysis of DDDs showed that only the use of NSAIDs was correlated with a lower presence of ADRD and depended on the duration of exposure [mild exposure = OR 0.93, 95%CI (0.82–1.05); moderate exposure = OR 0.67, 95%CI (0.54–0.85); and strong exposure = OR 0.72, 95%CI (0.58–0.89)]. In contrast, only the use of non-opioid analgesics and opioids was correlated with a more frequent occurrence of ADRD, depending on the duration of exposure [non-opioid analgesics: mild exposure = OR 1.76, 95%CI (1.39–2.23); moderate exposure = OR 2.54, 95%CI (2.00–3.24); and strong exposure = OR 2.83, 95%CI (2.23–3.60); opioids: mild exposure = OR 1.10, 95%CI (0.97–1.24); moderate exposure = OR 1.21, 95%CI (1.00–1.47); and strong exposure = OR 1.24, 95%CI (1.02–1.49)]. Looking at all analgesic treatments, only continuous analgesic use over a 24-month period seems to be linked to the presence of ADRD [OR 1.23, 95%CI (1.10–1.38)] (Table 4). No relationship was found between the presence of ADRD and exposure to other therapeutic classes.

A second analysis was also carried out according to exposure to different analgesic molecules (see Supplementary material; Supplementary Table S1 for details). Nevertheless, the low rate of patients receiving certain drugs did not allow us to interpret reliably this results for all pain medications, only paracetamol and dextropropoxyphene widely prescribed were interpreted. We observed that paracetamol (OR from 1.75 to 2.80 according to exposure) and dextropropoxyphene (OR from 1.29 to 1.90 according to exposure) were correlated with the presence of ADRD and dependent on exposure time (Table 5). Nevertheless, these results are only exploratory and need to be confirmed by a dedicated study on more patients.

First, the identification of people suffering from chronic pain can only be done by medical diagnosis, LTI or the presence of a continuity of pain medications. This implies that individuals who have not been diagnosed for their chronic pain and who are not treated by reimbursed painkillers are not identified, with a potential underestimation of chronic pain patients. Second limitation is the reliability of the data reported in the administrative databases, with risks of error or non-coding of some pathologies or the care received. This is the case for chronic pain (Lacasse et al., 2020) and ADRD (Carcaillon-Bentata et al., 2016), leading to the potential underestimation of theses pathologies. Nevertheless, the algorithms used to identify chronic pain and ADRD individuals seemed reliable. Indeed, in a previous study (Chenaf et al., 2018) the algorithm used for identifying chronic pain showed a prevalence of chronic pain in France similar to that found in a large French national cohort study (Bouhassira et al., 2008), indirectly validating the algorithm. Concerning the algorithm of identification of ADRD, it has been validated and published previously (Rochoy et al., 2019a; Elyn et al., 2022). Finally, the most important limitation was the time between the identification of chronic pain (index date) and the diagnosis of ADRD. In a similar study of Billioti de Gage et al. (2014) the minimal delay used before dementia onset was 5 years. In our study, the average time from identification of chronic pain to diagnosis of ADRD was approximately 4 years. However, because of a lack of power (not enough patients with a minimum of 5 years between chronic pain and ADRD), we did not impose a delay between the index date and the diagnosis of ADRD. This lack of patients is explained by the advanced age of our population at inclusion (65 years), leaving little time for the onset of ADRD. Moreover, as the EGB database only had data available from 2005 onwards, the follow-up time was too limited to include younger patients and to allow the appearance of ADRD. This limitation raises the question of follow-up bias. Indeed, it is possible that individuals with chronic pain have a more optimal medical follow-up than individuals without chronic pain, and indirectly have an earlier detection of dementia disorders and therefore an earlier diagnosis. However, this follow-up bias needs to be modulated, as control individuals, like those with chronic pain, have similar comorbidities, some of which require extensive medical follow-up. These limitations raises questions about the real impact of chronic pain on the incidence of ADRD. Moreover, certain mechanisms involved in pathological central nervous system aging toward ADRD start very early even before the mild cognitive impairment stages (Jack et al., 2010; Bateman et al., 2012). Moreover, the delay in diagnosis is estimated to be between 2 and 5 years after the onset of symptoms, with more than two-thirds of diagnoses already in moderate to severe stages (Löppönen et al., 2003). Thus, we can hypothesize that chronic pain could be considered as a risk factor, but not the direct cause, of a neurodegenerative pathology and its clinical revelation. Also, symptoms of chronic pain may be a prodromal of ADRD. These questions remain complex and difficult to answer. A larger study, including a minimum delay of 5 years between chronic pain and the diagnosis of ADRD, should be conducted to confirm our results.

In conclusion, our results suggest that chronic pain is a risk factor for the development of an ADRD independently of various confounders. Combined with the fact that chronic pain can lead to decreased mobility, interfere with daily activities, and increase the risk of falls and additional injuries, our study reinforces the importance of prevention, diagnosis, and management of chronic pain with adapted pharmacological and non-pharmacological strategies to limit resulting comorbidities, such as dementias and Alzheimer’s disease. This is especially important in the elderly, who may have difficulty expressing their pain and are at greater risk of developing dementias. Nevertheless, further studies are needed to clarify the pathophysiological mechanisms involved in relationship between chronic pain, long-term use of pain medications and dementia (Cao et al., 2019).