Importance
Tramadol is a weak opioid analgesic whose use has increased rapidly, and it has been associated with adverse events of hypoglycemia.
Objective
To assess whether tramadol use, when compared with codeine use, is associated with an increased risk of hospitalization for hypoglycemia.
Design, Setting, and Participants
A nested case-control analysis was conducted within the United Kingdom Clinical Practice Research Datalink linked to the Hospital Episodes Statistics database of all patients newly treated with tramadol or codeine for noncancer pain between 1998 and 2012. Cohort and case-crossover analyses were also conducted to assess consistency of the results.
Main Outcomes and Measures
Cases of hospitalization for hypoglycemia were matched with up to 10 controls on age, sex, and duration of follow-up. Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated comparing use of tramadol with codeine. A cohort analysis, with high-dimensional propensity score–adjusted hazard ratios (HRs) and 95% CIs, was performed comparing tramadol with codeine in the first 30 days after treatment initiation. Finally, a case-crossover analysis was also performed, in which exposure to tramadol in a 30-day risk period immediately before the hospitalization for hypoglycemia was compared with 11 consecutive 30-day control periods. Odds ratios and 95% CIs were estimated using conditional logistic regression analysis.
Results
The cohort included 334 034 patients, of whom 1105 were hospitalized for hypoglycemia during follow-up (incidence, 0.7 per 1000 per year) and matched to 11 019 controls. Compared with codeine, tramadol use was associated with an increased risk of hospitalization for hypoglycemia (OR, 1.52 [95% CI, 1.09-2.10]), particularly elevated in the first 30 days of use (OR, 2.61 [95% CI, 1.61-4.23]). This 30-day increased risk was confirmed in the cohort (HR, 3.60 [95% CI, 1.56-8.34]) and case-crossover analyses (OR, 3.80 [95% CI, 2.64-5.47]).
Conclusions and Relevance
The initiation of tramadol therapy is associated with an increased risk of hypoglycemia requiring hospitalization. Additional studies are needed to confirm this rare but potentially fatal adverse event.
Tramadol hydrochloride is a weak opioid analgesic whose use has increased steadily worldwide.1,2 Recently, several spontaneous reports have raised concerns that its use might be associated with an increased risk of hypoglycemia.3-8 In a pharmacovigilance study, tramadol-induced hypoglycemia occurred rapidly after initiation—within 10 days of treatment.8 Moreover, there were no known risk factors, such as diabetes mellitus, in more than 40% of the reports.8
Hypoglycemia is a serious clinical event that has been associated with elevated death rates in patients with diabetes.9-11 Furthermore, prolonged and profound hypoglycemia can cause brain death, as well as fatal cardiac arrhythmia.12 With respect to tramadol, it is biologically plausible that it may induce hypoglycemia through its dual effects on μ opioid receptors and inhibitory activity on serotonin-norepinephrine reuptake.13
Given the increasing use of tramadol in the general population,1,2 there is a need to assess whether this drug is associated with an increased risk of hospitalization for hypoglycemia, a potentially fatal outcome. The objective of this large population-based study was to determine whether use of tramadol, when compared with use of codeine, another weak opioid not previously associated with hypoglycemia, is associated with an increased risk of hospitalization for hypoglycemia in individuals with noncancer pain.
This study was conducted using the United Kingdom Clinical Practice Research Datalink (CPRD) linked to the Hospital Episode Statistics (HES) database. The CPRD includes more than 13 million patients from more than 680 practices in the United Kingdom14 and records information on diagnoses, prescriptions, referrals, lifestyle habits, and anthropometric measurements such as body mass index. Information in the CPRD is regularly audited and has been shown to be highly valid (median proportion of patients with confirmed diagnoses is 89%).15,16 Since 1997, the HES database records all hospitalizations in England. The recorded information includes hospitalization dates, primary and secondary diagnoses (coded using International Classification of Diseases, 10th Revision [ICD-10]), and related procedures.
The study protocol was approved by the Independent Scientific Advisory Committee of the CPRD (protocol 14_099R) and by the Research Ethics Board of Jewish General Hospital, Montreal, Quebec, Canada. Patient informed consent was not necessary since the data were anonymized for research purposes.
We assembled a population-based cohort of patients newly treated with oral formulations of tramadol or codeine between April 1, 1998, and March 1, 2012. Patients initiating treatment with a codeine formulation for cough or diarrhea were not included. Cohort entry was defined by the date of the first prescription for these drugs during the study period. At cohort entry, patients were required to be at least 18 years old and have at least 1 year of baseline medical history in the CPRD and HES. We excluded patients concurrently prescribed other opioids at cohort entry (listed in eTable 1 in the Supplement), thus limiting the cohort to patients using tramadol or codeine only. We also excluded patients who had received a cancer diagnosis (other than nonmelanoma skin cancer) at any time before cohort entry, as well as those previously hospitalized for hypoglycemia (ICD-10 codes E15, E16.0, E16.1, and E16.2, in primary or secondary position) in the year before cohort entry.
Patients were followed from cohort entry until the study outcome of hospitalization for hypoglycemia (defined in the Methods), death from any cause, end of registration with the general practice, or end of the study period (March 31, 2012), whichever occurred first.
A nested case-control analysis was conducted within the aforementioned cohort. This analytic approach was chosen because of the time-varying nature of exposure, the size of the cohort, and the long duration of follow-up.17
Cases were all patients with a first hospitalization for hypoglycemia (recorded within the first 2 days of hospitalization; ICD-10 codes E15, E16.0, E16.1, or E16.2, in primary or secondary position) during the study period. The index date was defined as the time of the case’s hospital admission. Up to 10 controls were randomly selected from the risk sets of each case and matched on age, sex, and duration of follow-up.
Cases and controls were classified into 1 of 3 mutually exclusive categories on the basis of their exposure status at index date: (1) codeine use, defined by a first prescription in the 30 days before index date, with no prior prescriptions of codeine or tramadol in the year before the index date; (2) tramadol use, defined by a prescription in the 30 days before the index date, but no codeine prescriptions in the year before the index date; and (3) other patterns and combinations, consisting of concurrent use of tramadol and codeine, previous use of either tramadol or codeine in the year before the index date, and no use of either medication in the year before the index date. The tramadol group included those who initiated tramadol therapy 30 days or less before the index date, as well as those who initiated tramadol therapy more than 30 days before the index date. This was done to assess whether the risk varies with timing of treatment initiation, an analysis motivated by the pharmacovigilance study that signaled a rapid onset of tramadol-induced hypoglycemia (ie, within 10 days).8 Thus, in a secondary analysis, tramadol use was further classified according to timing of the first prescription in the year before the index date (ie, first prescription ≤30 days and >30 days before index date). The reference category for all analyses consisted of codeine use.
The following predefined baseline covariates were considered in the models: calendar year of cohort entry, body mass index, excessive alcohol use (defined as alcohol-related disorders such as alcoholism, alcoholic cirrhosis of the liver, alcoholic hepatitis, and hepatic failure), comorbidities (chronic renal insufficiency, liver disease, pancreatic disease, other endocrine disease [including adrenal insufficiency and hypopituitarism], and dumping syndrome–inducing surgical procedures [gastrectomy and bypass surgery]; all measured in the year before cohort entry), use of prescription drugs (antidiabetic drugs [insulins, sulfonylureas, metformin hydrochloride, other antidiabetic drugs; entered individually in the models], β-blockers, angiotensin-converting enzyme inhibitors, fluoroquinolones, co-trimoxazole, antiepileptics, antidepressants [selective serotonin reuptake inhibitors, serotonin norepinephrine reuptake inhibitors, other antidepressants], antipsychotics, aspirin, nonsteroidal anti-inflammatory drugs, propoxyphene hydrochloride, and other opioids [listed in eTable 1 in the Supplement]; all measured in the 90 days before cohort entry), and opioid-related indications (headache, abdominal and pelvic pain, musculoskeletal pain, neuralgia, other pain [including chest pain, otolaryngological pain, and unspecified pain], injury or trauma, and surgery; all measured in the 90 days before cohort entry). As additional proxies of health status, the models also included the number of general practice visits and number of hospitalizations in the year before cohort entry, as well as number of prescription drugs received in the 90 days before cohort entry. Variables with missing data were coded with an “unknown” category.
Primary Analysis: Nested Case-Control Approach
Descriptive statistics were used to summarize the characteristics of the cohort, cases and matched controls. We calculated the crude incidence of hospitalization for hypoglycemia, along with 95% confidence intervals (CIs) based on the Poisson distribution. Conditional logistic regression was used to estimate odds ratios (ORs) with 95% CIs of hospitalization for hypoglycemia, comparing tramadol use with codeine use. We also conducted 2 secondary analyses. The first assessed the association with tramadol use, categorized according to timing of the first prescription before the index date (≤30 days and >30 days). The second assessed whether the presence of antidiabetic drug use at baseline modified the association between tramadol use and hospitalization for hypoglycemia. For this analysis, effect modification was assessed by including an interaction term between exposure status and the antidiabetic drug covariate. In addition to the matching factors (age, sex, and duration of follow-up) on which the logistic regression was conditioned, the models were adjusted for the aforementioned potential baseline confounders.
We conducted 3 sensitivity analyses to ascertain the robustness of the findings of the nested case-control analysis. In the first, the accuracy of the outcome definition was assessed by restricting the events to those coded in primary position. In the second and third analyses, we assessed whether hospitalizations for any cause in the 30 days before the index date, and surgical procedures in the 90 days before the index date, were effect modifiers. Effect modification was assessed by including interaction terms in the model.
Secondary Analysis: Cohort Approach Among First-Time Users
By design, the nested case-control analysis may have included reinitiators of tramadol and codeine use—patients who may have used these agents in the past but did not experience hypoglycemia. The inclusion of such “survivors” may lead to an overall underestimation of the risk through the depletion of susceptibles phenomenon.18 Furthermore, because of the relatively long follow-up, adjustment for baseline covariates may introduce residual confounding given that they may no longer optimally predict future exposure and outcome. Thus, to address these concerns, we conducted a cohort analysis restricted to first-time tramadol and codeine users (ie, patients identified at cohort entry). These patients were observed for a maximum of 30 days or until a hospitalization for hypoglycemia (as defined in the Methods), a hospitalization unrelated to hypoglycemia, death from any cause, end of registration with the general practice, or end of the study period (March 31, 2012), whichever occurred first.
Kaplan-Meier curves were plotted to compare the cumulative incidence of hospitalization for hypoglycemia of tramadol use with that of codeine use up to 30 days after treatment initiation. Cox proportional hazards models were used to estimate crude and adjusted hazard ratios with 95% CIs of hospitalization for hypoglycemia associated with tramadol use compared with codeine. The model was adjusted for high-dimensional propensity score (HD-PS) quartiles,19 which were calculated using multivariate logistic regression analysis as the probability of being exposed to tramadol vs codeine, conditional on the aforementioned 41 predefined baseline covariates and 500 empirically defined covariates from 7 data dimensions (additional information on this approach is provided in eMethods 1 in the Supplement). Overall, there was excellent overlap between the propensity score distributions, with a c-statistic of 0.69. The proportionality assumption of the Cox model was met and ascertained on the basis of Schoenfeld residuals.
Secondary Analysis: Case-Crossover Approach
As a means to address residual confounding, we also conducted a case-crossover analysis in which cases serve as their own controls. This method implicitly controls for all known and unknown time-independent confounders and can be used to investigate associations between transient exposures and acute outcome events.20
In this analysis, we used the same cases that were identified for the nested case-control approach. For each of these cases, we subdivided the year prior to the index date into 12 consecutive 30-day periods, with a risk period immediately prior to the index date, and 11 control periods (eFigure 1 in the Supplement). Thus, exposure to tramadol in the 30-day risk period prior to the hospitalization for hypoglycemia was compared with that of the 11 previous consecutive 30-day control periods. Because tramadol can be used over the long term by some patients (which can affect the precision of the point estimate by affecting the number of discordant pairs on which the analysis is based), the case series was restricted to those with fewer than 6 exposed control periods, an exposure distribution that corresponds more closely to a transient exposure.21
We conducted an additional analysis to assess whether tramadol use is associated with an increased risk of fatal hypoglycemia, defined as an in-hospital death following a hypoglycemia-related hospitalization. For this analysis, the cases with such in-hospital deaths were similarly restricted to those who transiently used tramadol (<6 exposed control periods). For both hypoglycemia hospitalization and fatal hypoglycemia, conditional logistic regression analysis was used to estimate ORs with 95% CIs comparing exposure in a 30-day risk period immediately before the hospitalization for hypoglycemia with that of 11 previous 30-day consecutive control periods. Overall, there was no evidence of an exposure time-trend in the year before the event date, thus satisfying one of the key assumptions of this approach (additional information regarding the assessment of the exposure time-trend can be found in eMethods 2 in the Supplement).22 All analyses described here were performed using SAS, version 9.3 (SAS Institute Inc).
A total of 334 034 patients met the study cohort inclusion criteria, which included 28 110 and 305 924 new users of tramadol and codeine, respectively (eFigure 2 in the Supplement). The use of tramadol increased more than 8-fold during the study period, from 25 334 prescriptions in 1999 to 215 709 prescriptions in 2011 (Figure 1). Overall, tramadol and codeine users were similar with respect to age, sex, comorbidities, and prescription drug use, including antidiabetic drugs. In contrast, tramadol users were more likely to have undergone surgery in the 90 days before cohort entry (eTable 2 in the Supplement).
Primary Analysis: Nested Case-Control Approach
During a mean follow-up of 5.0 years, generating 1 680 000 person-years, there were a total 1105 cases of hospitalization for hypoglycemia (crude incidence, 0.7 [95% CI, 0.6-0.7] per 1000 person-years) in the entire cohort. Of these, 112 (10.1%) were fatal. Table 1 presents the baseline characteristics of the 1105 cases and 11 019 matched controls. Compared with matched controls, cases were less likely to have ever smoked, but more likely to be obese and less healthy, in terms of comorbidities and prescription drug use.
Table 2 presents the results of the primary nested case-control approach. Compared with codeine use, tramadol use was associated with a 52% increased risk of hospitalization for hypoglycemia (adjusted OR, 1.52 [95% CI, 1.09-2.10]). In a secondary analysis, the risk was highest in patients who initiated the treatment within 30 days of the index date (adjusted OR, 2.61 [95% CI, 1.61-4.23]), whereas the OR was closer to the null in users who initiated the treatment more than 30 days before the index date (adjusted OR, 1.17 [95% CI, 0.78-1.75]). Finally, the presence of antidiabetic drug use modified the association between tramadol use and hypoglycemia. Specifically, the OR was higher in nonusers than in users of antidiabetic drugs (adjusted OR, 2.12 [95% CI, 1.18-3.79] vs 1.11 [95% CI, 0.76-1.64], respectively; P for interaction = .02).
There were 507 cases with a hospitalization for hypoglycemia coded in primary position. Restricting the analysis to these cases and their matched controls led to an increase in the OR compared with the primary analysis (adjusted OR, 2.15 [95% CI, 1.33-3.48]). The presence of hospitalizations in the 30 days before the index date did not statistically modify the association, although the OR was higher among patients with no hospitalizations (no hospitalizations: adjusted OR, 1.73 [95% CI, 1.15-2.61] vs hospitalization: adjusted OR, 0.98 [95% CI, 0.57-1.68]; P for interaction = .19). Similarly, there was no statistically significant effect modification by surgical procedures in the 90 days before the index date, although the OR was higher among those with no surgical procedures (no surgical procedures: adjusted OR, 1.66 [95% CI, 1.03-2.67] vs surgical procedures: adjusted OR, 1.11 [95% CI, 0.71-1.73]; P for interaction = .14).
Secondary Analysis: Cohort Approach Among First-Time Users
Overall, tramadol use was associated with a higher cumulative incidence of hospitalization for hypoglycemia than codeine use in the first 30 days after treatment initiation (log rank P < .001) (Figure 2). The crude incidences of hospitalization for hypoglycemia were 3.0 (95% CI, 1.3-6.0) and 0.7 (95% CI, 0.4-1.1) per 10 000 person-months in tramadol and codeine users, respectively. In the HD-PS–adjusted model, the initiation of tramadol use was associated with a more than 3-fold increased risk of hospitalization for hypoglycemia, compared with codeine use (HD-PS–adjusted HR, 3.60 [95% CI, 1.56-8.34]) (eTable 3 in the Supplement).
Secondary Analysis: Case-Crossover Approach
Among the 1105 cases hospitalized for hypoglycemia during the study period, 176 had received at least 1 prescription in the year before the index date, 141 (80.1%) of whom used tramadol transiently (eFigure 3 in the Supplement). Overall, transient tramadol use was associated with an increased risk of hospitalization for hypoglycemia (exposed risk vs control periods: 36.9% vs 13.2%; OR, 3.80 [95% CI, 2.64-5.47]). Transient tramadol use was also associated with an increased risk of fatal hypoglycemia (exposed risk vs control periods: 43.8% vs 9.7%; OR, 6.21 [95% CI, 2.23-17.26]).
To our knowledge, this is the first epidemiological study investigating the association between tramadol use and hospitalization for hypoglycemia. We found that tramadol use is associated with an increased risk of hospitalization for hypoglycemia, with the risk highest around the time of treatment initiation. These results were corroborated in cohort and case-crossover analyses, which also associated tramadol use with a more than 3-fold increased risk of hospitalization for hypoglycemia. Overall, these results remained robust in several secondary analyses, including among patients not using any antidiabetic drugs, as well as in sensitivity analyses.
The findings of this study confirm the reported signals of tramadol therapy potentially increasing the risk of hospitalization for hypoglycemia. Three recent case reports have described tramadol-induced hypoglycemia, which included patients with and without diabetes using the drug at the recommended doses.3,5,6 Hospitalization for hypoglycemia has also been reported in a woman with intentional tramadol overdose.7 An in-depth analysis of spontaneous reports from a French pharmacovigilance database identified 43 tramadol-associated hypoglycemia cases between 1997 and 2010.8 Most of these events occurred soon after initiation of tramadol therapy (77% within 10 days of treatment) and were more frequent in the elderly. Overall, our findings corroborate these signals because the increased risk seemed to be limited to the first 30 days of use and remained statistically significant in patients with no history of treated diabetes. The rarity of this outcome, approximately 7 per 10 000 per year, may explain why it was not observed in randomized clinical trials, which were underpowered to detect such events (a total of 1019 and 1378 tramadol-treated patients in osteoarthritis and lower back pain randomized clinical trials, respectively).23,24
The association between tramadol use and hypoglycemia is biologically plausible and may relate to its pharmacodynamic properties. Tramadol mainly acts through 2 mechanisms: the activation of µ opioid receptors and the inhibition of central serotonin and norepinephrine reuptake.13 Serotonin pathways are known to have complex effects on peripheral glucose regulation,25 with animal studies reporting that serotonin induces low glucose levels in diabetic mice and rats.26,27 Moreover, use of antidepressants such as those acting through serotonin and norepinephrine reuptake inhibition has been previously associated with an increased risk of hypoglycemia.28,29 In addition to its effects on serotonin pathways, the activation of µ opioid receptors by tramadol may also increase the risk of hypoglycemia. In rats with streptozotocin-induced diabetes, a dose-dependent glucose-lowering effect was observed with tramadol.30 This effect persisted when these rats were depleted in serotonin with p-chlorophenylalanine, suggesting a serotonin-independent effect. Furthermore, this effect was weaker in rats previously treated with naloxone hydrochloride (a µ opioid receptor agonist), suggesting a strong implication of the µ pathways.30 Given the novelty of this association, these hypotheses remain speculative and will require additional investigation.
This study has several strengths. First, we assembled a large population-based cohort of patients initiating tramadol or codeine therapy. Second, the use of the CPRD and HES databases allowed us to control for a large number of potential confounders. Third, the use of a new-user design eliminated biases related to the inclusion of prevalent users.31 Fourth, confounding by indication was likely minimized by using codeine therapy as an active comparator. Indeed, tramadol and codeine users were similar on nearly all baseline potential confounders, likely owing to the fact that both agents have similar indications. Finally, we observed consistent results with the cohort and case-crossover approaches, which addressed concerns related to possible residual confounding.
This study has some limitations. First, CPRD prescriptions represent those written by general practitioners, and thus treatment adherence is unknown, although this misclassification tends to bias the point estimates toward the null hypothesis. Second, to our knowledge, HES-defined hypoglycemia has not been formally validated, although it has been used as an outcome in a previous unrelated study.32 However, similar results were obtained after the case definition was limited to those diagnoses in primary position. Third, because of the observational nature of the study, residual confounding needs to be considered. Reassuringly, we observed consistent results using different study design and analytic approaches (such as cohort analysis adjusted for HD-PS quartiles and case-crossover analysis). Moreover, it is important to note that, given the strong observed associations (point estimates ranging between 1.52 and 3.80), any unmeasured or unknown confounder would need to be strongly associated with both the exposure and outcome to completely confound the observed association. It is unclear whether such a confounder exists beyond those considered in the models. Finally, despite the large sample size, the rarity of the outcome led to wide confidence intervals in secondary analyses, and thus these should be interpreted with caution.
The initiation of tramadol use was significantly associated with a more than 2-fold increased risk of hospitalization for hypoglycemia, when compared with codeine use. Although rare, tramadol-induced hypoglycemia is a potentially fatal adverse event. The clinical significance of these novel findings requires additional investigation.
Accepted for Publication: October 13, 2014.
Corresponding Author: Samy Suissa, PhD, Centre for Clinical Epidemiology, Lady Davis Institute, Jewish General Hospital, McGill University, Montreal, QB H3T 1E2, Canada (samy.suissa@mcgill.ca).
Published Online: December 8, 2014. doi:10.1001/jamainternmed.2014.6512.
Author Contributions: Dr Suissa had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Fournier, Azoulay, Suissa.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Fournier, Suissa.
Critical revision of the manuscript for important intellectual content: Fournier, Azoulay, Yin, Montastruc, Suissa.
Statistical analysis: Fournier, Azoulay, Yin, Suissa.
Obtained funding: Suissa.
Administrative, technical, or material support: Suissa.
Study supervision: Suissa.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was funded in part by research grants from the Canadian Institutes of Health Research and Canada Foundation for Innovation. Dr Suissa is the recipient of the James McGill Chair.
Role of the Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
2.Manchikanti
L, Helm
S
II, Fellows
B,
et al. Opioid epidemic in the United States.
Pain Physician. 2012;15(3)(suppl):ES9-ES38.
PubMedGoogle Scholar 3.Grandvuillemin
A, Jolimoy
G, Authier
F, Dautriche
A, Duhoux
F, Sgro
C. Tramadol-induced hypoglycemia: 2 cases [in French].
Presse Med. 2006;35(12, pt 1):1842-1844.
PubMedGoogle ScholarCrossref 4.Tavassoli
N, Lapeyre-Mestre
M, Sommet
A, Montastruc
J-L; French Association of Regional Pharmacovigilance Centres. Reporting rate of adverse drug reactions to the French pharmacovigilance system with three step 2 analgesic drugs: dextropropoxyphene, tramadol and codeine (in combination with paracetamol).
Br J Clin Pharmacol. 2009;68(3):422-426.
PubMedGoogle ScholarCrossref 5.Taugourdeau
S, Chiche
L, Rouby
F,
et al. Severe hypoglycemia induced by tramadol: two new cases of an unlisted side effect [in French].
Rev Med Interne. 2011;32(11):703-705.
PubMedGoogle ScholarCrossref 6.Jonville-Bera
A-P, Marie
A, Magba
D, Gedon
E, Autret-Leca
E. Tramadol-induced hypoglycemia in a diabetic patient [in French].
Therapie. 2010;65(5):499-500.
PubMedGoogle ScholarCrossref 8.Bourne
C, Gouraud
A, Daveluy
A,
et al; French Association of Regional Pharmacovigilance Centres. Tramadol and hypoglycaemia: comparison with other step 2 analgesic drugs.
Br J Clin Pharmacol. 2013;75(4):1063-1067.
PubMedGoogle ScholarCrossref 9.Bonds
DE, Miller
ME, Bergenstal
RM,
et al. The association between symptomatic, severe hypoglycaemia and mortality in type 2 diabetes: retrospective epidemiological analysis of the ACCORD study.
BMJ. 2010;340:b4909.
PubMedGoogle ScholarCrossref 10.Zoungas
S, Patel
A, Chalmers
J,
et al; ADVANCE Collaborative Group. Severe hypoglycemia and risks of vascular events and death.
N Engl J Med. 2010;363(15):1410-1418.
PubMedGoogle ScholarCrossref 11.McCoy
RG, Van Houten
HK, Ziegenfuss
JY, Shah
ND, Wermers
RA, Smith
SA. Increased mortality of patients with diabetes reporting severe hypoglycemia.
Diabetes Care. 2012;35(9):1897-1901.
PubMedGoogle ScholarCrossref 13.Raffa
RB, Friderichs
E, Reimann
W, Shank
RP, Codd
EE, Vaught
JL. Opioid and nonopioid components independently contribute to the mechanism of action of tramadol, an ‘atypical’ opioid analgesic.
J Pharmacol Exp Ther. 1992;260(1):275-285.
PubMedGoogle Scholar 14.García Rodríguez
LA, Pérez Gutthann
S. Use of the UK General Practice Research Database for pharmacoepidemiology.
Br J Clin Pharmacol. 1998;45(5):419-425.
PubMedGoogle ScholarCrossref 15.Herrett
E, Thomas
SL, Schoonen
WM, Smeeth
L, Hall
AJ. Validation and validity of diagnoses in the General Practice Research Database: a systematic review.
Br J Clin Pharmacol. 2010;69(1):4-14.
PubMedGoogle ScholarCrossref 16.Khan
NF, Harrison
SE, Rose
PW. Validity of diagnostic coding within the General Practice Research Database: a systematic review.
Br J Gen Pract. 2010;60(572):e128-e136.
PubMedGoogle ScholarCrossref 17.Suissa
S. Novel approaches to pharmacoepidemiology study design and statistical analysis. In: Strom
BL, ed. Pharmacoepidemiology.4th ed. Chichester, England: John Wiley & Sons, Ltd; 2006:811-829.
18.Miettinen
OS, Caro
JJ. Principles of nonexperimental assessment of excess risk, with special reference to adverse drug reactions.
J Clin Epidemiol. 1989;42(4):325-331.
PubMedGoogle ScholarCrossref 19.Schneeweiss
S, Rassen
JA, Glynn
RJ, Avorn
J, Mogun
H, Brookhart
MA. High-dimensional propensity score adjustment in studies of treatment effects using health care claims data.
Epidemiology. 2009;20(4):512-522.
PubMedGoogle ScholarCrossref 20.Maclure
M. The case-crossover design: a method for studying transient effects on the risk of acute events.
Am J Epidemiol. 1991;133(2):144-153.
PubMedGoogle Scholar 21.Hebert
C, Delaney
JAC, Hemmelgarn
B, Lévesque
LE, Suissa
S. Benzodiazepines and elderly drivers: a comparison of pharmacoepidemiological study designs.
Pharmacoepidemiol Drug Saf. 2007;16(8):845-849.
PubMedGoogle ScholarCrossref 23.Cepeda
MS, Camargo
F, Zea
C, Valencia
L. Tramadol for osteoarthritis: a systematic review and metaanalysis.
J Rheumatol. 2007;34(3):543-555.
PubMedGoogle Scholar 24.Chaparro
LE, Furlan
AD, Deshpande
A, Mailis-Gagnon
A, Atlas
S, Turk
DC. Opioids compared to placebo or other treatments for chronic low-back pain.
Cochrane Database Syst Rev. 2013;8:CD004959.
PubMedGoogle Scholar 26.Yamada
J, Sugimoto
Y, Kimura
I, Takeuchi
N, Horisaka
K. Serotonin-induced hypoglycemia and increased serum insulin levels in mice.
Life Sci. 1989;45(20):1931-1936.
PubMedGoogle ScholarCrossref 27.Chi
T-C, Ho
Y-J, Chen
W-P,
et al. Serotonin enhances beta-endorphin secretion to lower plasma glucose in streptozotocin-induced diabetic rats.
Life Sci. 2007;80(20):1832-1838.
PubMedGoogle ScholarCrossref 28.Derijks
HJ, Meyboom
RHB, Heerdink
ER,
et al. The association between antidepressant use and disturbances in glucose homeostasis: evidence from spontaneous reports.
Eur J Clin Pharmacol. 2008;64(5):531-538.
PubMedGoogle ScholarCrossref 29.Derijks
HJ, Heerdink
ER, De Koning
FH, Janknegt
R, Klungel
OH, Egberts
AC. The association between antidepressant use and hypoglycaemia in diabetic patients: a nested case-control study.
Pharmacoepidemiol Drug Saf. 2008;17(4):336-344.
PubMedGoogle ScholarCrossref 30.Cheng
JT, Liu
IM, Chi
TC, Tzeng
TF, Lu
FH, Chang
CJ. Plasma glucose-lowering effect of tramadol in streptozotocin-induced diabetic rats.
Diabetes. 2001;50(12):2815-2821.
PubMedGoogle ScholarCrossref 32.Khalid
JM, Raluy-Callado
M, Curtis
BH, Boye
KS, Maguire
A, Reaney
M. Rates and risk of hospitalisation among patients with type 2 diabetes: retrospective cohort study using the UK General Practice Research Database linked to English Hospital Episode Statistics.
Int J Clin Pract. 2014;68(1):40-48.
PubMedGoogle ScholarCrossref