In a retrospective cohort study, Ruth Brauer and colleagues examine the association between treatment with glitazone and incidence of Parkinson's disease.
Parkinson’s disease (PD) is a common, progressive degenerative neurological disease, with a markedly increased prevalence with older age . It is principally characterised by neurodegeneration of the nigrostriatal neurons and a consequent deficit in striatal dopamine content [2,3]. Until now, no effective treatments have been found to tackle directly the neurodegenerative aspect of PD.
Promising in vitro and in vivo studies with rodents show that peroxisome proliferation-activated receptor gamma (PPARɣ) agonist medications, such as the antidiabetes glitazone (GTZ) drugs pioglitazone and rosiglitazone, may have neuroprotective effects by inhibiting microglial activation [4–6]. There is one ongoing small clinical trial among patients with PD looking at disease progression (Clinicaltrials.gov registration: NCT01280123), but there are no published human data on whether GTZ drugs are protective against PD.
Utilizing electronic health records, we aimed to determine whether there is a protective association between exposure to GTZ antidiabetic medications and PD in a large primary care population of people with diabetes.
Our study protocol was approved by the Independent Scientific Advisory Committee for MHRA database research (Independent Scientific Advisory Committee [ISAC] protocol number 13_016, February 2013). Studies using anonymised CPRD data for ISAC-approved observational research are covered by CPRD’s broad MREC ethics approval, without the need for specific informed consent.
Data Source and Study Design
A cohort study was conducted among people with type 1 and 2 diabetes within the United Kingdom Clinical Practice Research Datalink (CPRD), a well-established research database of electronic health records. CPRD collects and archives the anonymised medical, laboratory, referral, and prescribing records of primary care practices. CPRD data are demographically and geographically representative of the UK population, representing around 8% of it [7,8]. This study used the medical records of patients registered at one of 680 participating practices, comprising 13,291,839 active patients in October 2013 (month of data retrieval). As primary care is free at the point of delivery, it has almost universal coverage in the UK. The database has been described in detail elsewhere .
Selection of GTZ Users and the Comparison Group
The study population was drawn from the entire CPRD population, with follow-up time from the first of January 1999 onwards, when rosiglitazone was introduced. This was a dynamic cohort, with follow up ending at the earliest of the following: the date the patient left the practice, death date, or the latest date of data collection (1August 2013). Exposed patients were selected for inclusion if they had received any prescription for a GTZ (code list in S1 Text), with the first occurrence being at least 12 months after the patient’s start of follow up to ensure incident use. No age restrictions were applied. Each GTZ-exposed patient was matched by age (within 5 years), sex, practice, and treatment stage (see “Definition of Exposure and Treatment Stages”) with one to five patients starting other treatments for diabetes from 1999 onwards, again with at least 12 months prior registration in the database (other antidiabetic drug-exposed group). Individuals in the control pool who were eligible as potential matches for both first-line GTZ users and second-line GTZ users were removed from the potential control pool once selected as a match for first-line GTZ users. Follow up for all patients began at the first prescribing event that made them eligible for inclusion (index date). Patients in the GTZ-using group exposed to other antidiabetic drugs prior to the start of their GTZ follow up were eligible as potential matches pre-GTZ use. Potential participants with any diagnosis of PD prior to their index date were excluded. The time index used in this study was follow-up time in years after study entry. The mean follow-up time was 5.23 years (standard deviation 3.75), with a range of 0 to 14.57 years.
Definition of Exposure and Treatment Stages
Exposure was defined as a recorded prescription for an antidiabetic agent listed in chapter 6.1 of the British National Formulary (BNF). To account for missing information regarding dose and duration of treatment, the median obtained from all other prescriptions of the same drug was imputed if no other information was available. In the primary analysis, exposure was characterized as “ever exposed” versus “never exposed” to GTZ drugs in a time-updated manner: GTZ-exposed individuals at the start of their follow-up time remained ever exposed regardless of any subsequent changes in therapy; nonusers starting a GTZ during follow up changed from never exposed to ever exposed from the date of the change.
GTZ-exposed and other antidiabetic drug-exposed individuals were classified according to their diabetes treatment stage at the index date as shown in Fig 1: stage (1) monotherapy with GTZ only or monotherapy with any other antidiabetic drugs, with no prior pharmaceutical treatment for diabetes received; stage (2) second-line GTZ as monotherapy or a second-line other antidiabetic drug as monotherapy, with the first-line antidiabetic drug discontinued; and stage (3) second-line GTZ in combination with one or more other antidiabetic drugs, versus any other non-GTZ antidiabetic drugs in combination of two or more drugs (see Fig 1).
Identification of Patients with PD
The outcome for this study was the first recording of a PD diagnosis after the index date. People with a first recorded occurrence of PD were identified using clinical and referral records, using the Read code list in S2 Text. Any person with a diagnosis of PD due to known causes (such as drug-induced PD), based on the recording of codes presented in S2 Text, was censored at the date of the diagnosis.
We measured an incidence rate ratio (IRR) for the association between GTZ exposure and incident PD using conditional Poisson regression, conditional on the matched sets to control for gender, age, practice, and treatment stage. GTZ-exposed individuals were compared with patients exposed to alternative treatments. The primary analysis followed all patients from their index date until the earliest of the following: recording of PD, transfer out/death, or last data collection date for the practice.
Adjustments for potential confounding were made in a forward stepwise manner, retaining variables in the model if they had a substantial impact on the estimated rate ratio for GTZ-exposure (a change of 10% or more) and starting with variables associated with PD incidence in univariate analyses. Additionally, we included a model adjusting for all potential confounders. Covariates that were explored at the time of entry to the study because of their potential association with both GTZ exposure and PD were current smoking, prescriptions for hormone replacement therapy and calcium channel blockers, head trauma, calendar year, alcohol consumption, and body mass index (BMI) [10–16]. A protective effect of diabetes, compared to people without diabetes, has recently been reported . Our analysis was restricted to people with diabetes, but the degree of protection may be related to disease severity. As such, we included measures of severity as confounders, including HbA1c level at the index date and diabetes duration to index date. In a secondary analysis, we also adjusted for time-updated age (5-year age bands, except for 50–70 years for which we used 2-year bands) and calendar year (see below).
All data management and analyses were performed using Stata 12 (StataCorp, Texas).
Firstly, we took into account cessation of GTZ therapy by dividing GTZ-exposed follow-up time into current and past use: users were designated as past users 180 days after their last GTZ prescription. The 180 days represent a gradual shift from full exposure to an entirely unexposed state, allowing for potential delayed use of medication. Within this post hoc time-updated analysis, a stratified analysis on length of follow up of current GTZ exposure was conducted to assess whether the treatment association varied with exposure length. Secondly, a stratified analysis was conducted to look separately at the association with pioglitazone and rosiglitazone.
We performed several sensitivity analyses:
Metformin has been suggested to also have a protective effect against PD . To explore any potential effect modification by metformin use, we performed an analysis restricted to current or past metformin users.
A sensitivity analysis was conducted using a stricter definition of PD in which at least two prescriptions for an anti-PD drug were required in addition to a Read code indicating PD, to reduce the likelihood of outcome misclassification. The date of earliest recorded diagnosis remained the date of the outcome.
Post hoc sensitivity analyses were conducted restricted to all users of antidiabetic agents aged 40 and older; in a separate analysis, analyses were restricted to nonsmokers. Lastly, to examine the association with oral antidiabetic agents only, we excluded individuals who were prescribed insulin before or on the index date.
Of the 60,224 patients with a recorded GTZ prescription from 1999 onwards, 44,597 met the inclusion criteria (see Fig 2). 13,988 GTZ-exposed individuals had less than 12 months of follow-up time or were exposed to GTZ drugs before the start of follow up. 1,494 GTZ-exposed individuals could not be matched to users of other antidiabetic drugs because of a lack of available matches with the same age, gender, treatment stage, and practice and were dropped from the analyses, as were 145 GTZ users diagnosed with PD before GTZ use. For the primary analysis, the GTZ-exposed individuals were matched to 120,373 other antidiabetic drug-exposed individuals. The average follow-up time of GTZ-exposed individuals and those prescribed other antidiabetic treatments from the case index date within the matched set until the overall end of follow up in CPRD was similar (6.1 versus 5.9 years). 81% of all patients in the GTZ-exposed groups (n = 36,341) and 79% (n = 95,331) of those prescribed other treatments for diabetes were followed up from the start date until the final data collection date. Four percent of all study participants (1,561 individuals exposed to GTZ drugs and 4,946 individuals exposed to other treatments for diabetes) were followed up until they transferred to another primary care practice, and the remaining patients (6,695 GTZ-exposed and 20,096 individuals exposed to other treatments for diabetes) were followed up until their recorded death date. Most of the GTZ-exposed group (94%) were prescribed second-line GTZ as dual therapy in combination with metformin (55%) or sulphonylureas (39%). The characteristics of the GTZ-exposed and the matched other antidiabetic drug-exposed group are presented in Table 1.
GTZ-exposed individuals were more likely than those prescribed other antidiabetic drugs to have BMI > 25 kg/m2 and to have higher HbA1c levels at baseline. The average age of PD diagnosis in GTZ users was similar to non-GTZ users (73.2 versus 73.5 years). See S1 Table for a comparison of individuals with and without PD.
Results Primary Analysis
The incidence rate (IR) of PD in the GTZ-using group (6.4 per 10,000 patient years, 95% confidence interval [CI] 5.5–7.4) was lower than the IR of PD in the matched comparison cohort (8.8 per 10,000 patient years, 95% CI 8.1–9.6). The IRR, accounting for age, gender, practice, and treatment stage by matching, showed strong evidence of a protective association between use of GTZ drugs and the onset of PD (IRR 0.72, 95% CI 0.60–0.87, p-value < 0.001). The results of the primary analysis are presented in Table 2. A similar association was found when adjusting for all potential confounders (fully adjusted IRR 0.75, 0.59–0.94, p-value = 0.011; see Table 3).
A simple test for interaction by age (above and below the median age) as a post hoc analysis showed no evidence for effect modification by age (p-value = 0.9729).
Results Secondary Analyses
When the follow-up time of the GTZ–exposed group was updated to differentiate between the current and past GTZ-exposed groups, a strong protective association was found for current GTZ-exposure (IRR 0.59, 95% CI 0.46–0.77, p-value < 0.0001), but we observed a weaker, nonsignificant association with past use (IRR 0.85, 95% CI 0.65–1.10, p-value = 0.209). A stratified analysis on length of follow up of current GTZ exposure suggested no important variation in rates of PD depending on duration of use (see Table 2).
The point estimate for the estimated protective association was stronger for rosiglitazone compared to other antidiabetic drugs (IRR = 0.65, 0.52–0.82, p-value < 0.0001) than pioglitazone compared to other antidiabetic drugs (IRR = 0.89, 0.65–1.24, p-value = 0.496), but numbers of events were limited in the latter group, CIs overlapped, and a formal test suggested no evidence of an interaction (p-value for interaction = 0.42).
Results Sensitivity Analyses
Restricting the cohort or changing the definition of PD did not have a material impact on the main results: the association between exposure to GTZ drugs and the lower incidence of PD remained (see Table 2).
In a large population-based cohort study, we have shown for the first time that in individuals with diabetes, a prescription for GTZ is associated with a 28% lower rate of clinical presentation of PD compared to those prescribed other antidiabetic agents. The estimated association was similar when established risk factors, such as age and smoking, and possible risk factors, such as head injury and BMI, were adjusted for. Results allowing for changes in GTZ exposure showed strong evidence of a protective association between current GTZ exposure and PD compared with other antidiabetic drug exposure (IRR 0.59, 95% CI 0.46–0.77). No significant association was found for past exposure (IRR 0.85, 95% 0.65–1.10). Our results suggest that the protective association is limited to periods of GTZ treatment, with little or no longer-lasting benefit. This is supported by animal study data in which withdrawal of PPARγ therapy negates any neuroprotective effect .
To our knowledge, this is the first population-based investigation of the relationship between treatment with GTZ drugs and the incidence of PD. A clinical trial is currently investigating pioglitazone as a therapy to slow disease progression in early-stage PD in 216 participants (Clinicaltrials.gov registration: NCT01280123), but this trial is not designed to assess an effect on disease incidence. Our study provides unique evidence to support further investigation of GTZ use in PD.
Evidence was found of an association between the use of both pioglitazone and rosiglitazone and the onset of PD. Rosiglitazone was introduced to the market in 1999 but was eventually suspended because of cardiovascular safety concerns . However, in late 2013 the FDA removed the prescribing and dispensing restrictions for rosiglitazone medicines that were put into place in 2010 . Whilst pioglitazone use has waned because of concerns over the risk of bladder cancer, recent studies show that the association may be partly mediated by proteinuria testing [22,23].
Stratifying by year of GTZ initiation did not change the main results, as strong associations were found for relatively short follow-up periods of use.
Regarding the mechanisms of neuroprotection, glitazones have been shown to reduce the expression of matrix metalloproteinases (MMP 3 and 9) by traumatised neurons, which are known microglial activators as well as a direct anti-inflammatory on microglia . GTZ drugs have been shown to have a number of other neuroprotective properties in addition to their anti-inflammatory effects. GTZ treatment, through its ability to up-regulate PGC-1alpha expression, has been shown to increase brain mitochondrial biogenesis and the expression of antioxidant enzymes and antiapoptotic factors [5,24–27]. Underexpression of peroxisome proliferator-activated receptor-gamma coactivator (PGC)-1α has been identified as a potential link for manifestations of defects in mitochondrial biogenesis and energetics . Selective loss of dopamine neurons in the substantia nigra can be reversed by PGC-1α expression in both in vitro and in vivo models [28,29]. This could be potentially important since mitochondrial deficits and oxidative stress have also been implicated in the pathogenesis of PD. Additional peripheral mechanisms of neuroprotection may be through adiponectin, or adipocyte complement-related protein 30 (acrp30), which is elevated following the administration of PPAR agonists and has anti-inflammatory properties in its own right . Short-term treatment with GTZ drugs in animal models produced dramatic neuroprotection, larger than what is often observed by many neuroprotective compounds. Hence, even short-term treatment with GTZ drugs in the presymptomatic phase of PD, when the nigrostriatal system will be fairly intact, may be able to break the cycle of neurodegeneration by inhibiting the neuroinflammation process and preventing or slowing disease onset .
Strengths and Limitations
Our cohort included a large study population of >160,000 diabetes patients from a health care database with representative population coverage, in a setting with near-universal access to medical care across the demographic and socioeconomic spectrum. As the electronic primary care record is used to generate patient prescriptions, medication data are complete and accurate. Repeat prescriptions are triggered by the patient; thus, our measures of drug use likely reflect actual use. Misclassification of exposure periods could have occurred if patients were not adhering completely to prescribed medicines, but diabetes is a serious condition for which patients are likely to be motivated to take their medication, and any such misclassification is likely to have been nondifferential with respect to PD status. Similarly, confounders such as smoking status and alcohol use will be subject to some imprecision, but this is unlikely to be differential according to GTZ user status.
Diagnoses of PD in CPRD have a positive predictive value of 81% (primary definition) to 90% (secondary definition) [13,30]. The results of a secondary analysis with a more stringent definition of PD (clinical diagnosis plus at least two prescriptions for an anti-PD drug) gave similar results (0.82 versus 0.72 in the primary analysis). Given the overall low prevalence of PD, it is unlikely that a substantial number of patients were missed. As we did not take into account any future treatment breaks or changes, we may have misclassified the exposure status of individuals in our primary analysis. However, in a secondary analysis we allowed for treatment changes and breaks (current versus past exposure to GTZ drugs) and found a strong association with current use. Moreover, in a post hoc analysis, we compared current GTZ exposure with current exposure to other treatments for diabetes, allowing for treatment breaks in both groups: the strong association with current GTZ exposure remained (IRR 0.56, 95% CI 0.43–0.72). Any misclassification in our primary analysis is likely to have been nondifferential and could have caused bias towards the null, or a potential underestimation of the protective association.
The onset of PD in the context of the current study refers to the first recorded clinical diagnosis of PD in the CPRD. Patients are likely to have experienced symptoms before the date of their clinical diagnosis. We compared the age of onset of PD between GTZ users and nonusers but found no difference in average age, suggesting that GTZ drug do not simply delay the appreciation of symptoms. A key clinical consideration is whether GTZ can slow PD progression after onset. Unfortunately, we cannot address this question with these data, as objective markers of PD progression do not exist and measures of symptom progression are subjective and poorly captured in the primary care record. Furthermore, increases in dosage of anti-PD medication or the addition of added anti-PD therapies, e.g., catechol-O-methyltransferase (COMT) inhibitors (used in patients with motor fluctuations), are not reliable markers of disease progression. When we attempted to compare the time to first PD treatment after the clinical PD diagnosis, we found that a large number of GTZ users stopped their GTZ before PD diagnosis. GTZ drugs, like all medications, have known side effects, and we are not making any recommendations about their use in treating PD, since the balance of any potential benefits and risks is entirely unknown at this stage.
Whilst we tried to select comparator groups that were as similar as possible to the GTZ users, we cannot rule out unknown or unmeasured differences between the groups. The main reason for nonmatches was the lack of similar patients per treatment stage, and this is a matching criterion that cannot be weakened in a meaningful way. GTZ drugs are most often prescribed as a second-line treatment to patients who have responded less successfully to other treatments for diabetes. We matched these patients to other users of second-line treatment at the date of receipt of the add-on treatment. We controlled for the date of onset of diabetes and HbA1c levels in a fully adjusted analysis. As the results were very similar to the results of the primary analysis, we do not believe that GTZ patients in this study had a different underlying risk of developing PD compared with users of other antidiabetic drugs. Adjusting for other known predictors of PD (most notably smoking) did not appreciably change the estimated association with GTZ use. As none of the potential risk factors were strongly associated with the decision to prescribe a GTZ, only the matching variables were retained in the final model.
The study was conducted entirely within a patient population with diabetes, which may affect the generalisability of our results. However, adjusting for duration of diabetes did not have an effect on the results of the main analysis (IRR 0.72, 95% CI 0.59–0.87), and we believe that it is unlikely that any pharmacological effect of GTZ drugs relevant to PD is limited to people with diabetes only.
We found strong evidence for a protective association between GTZ exposure and the IR of clinical presentation with PD in humans, consistent with previous evidence from animal and in vitro studies. Our findings indicate that interventions based on the same mechanisms as PPARγ agonist activity may be fruitful targets for future research in PD.
1. Pringsheim T, Jette N, Frolkis A, Steeves TD. The prevalence of Parkinson's disease: a systematic review and meta-analysis. Mov Disord. 2014 Nov;29(13):1583–90. doi: 10.1002/mds.25945 24976103
2. Dexter DT, Jenner P. Parkinson disease: from pathology to molecular disease mechanisms. Free radical biology & medicine. 2013 Sep;62:132–44.
3. Jenner P, Olanow CW. The pathogenesis of cell death in Parkinson's disease. Neurology. 2006 May 23;66(10 Suppl 4):S24–36. 16717250
4. Carta AR, Frau L, Pisanu A, Wardas J, Spiga S, Carboni E. Rosiglitazone decreases peroxisome proliferator receptor-gamma levels in microglia and inhibits TNF-alpha production: new evidences on neuroprotection in a progressive Parkinson's disease model. Neuroscience. 2011 Oct 27;194:250–61. doi: 10.1016/j.neuroscience.2011.07.046 21839812
5. Sadeghian M, Marinova-Mutafchieva L, Broom L, Davis JB, Virley D, Medhurst AD, et al. Full and partial peroxisome proliferation-activated receptor-gamma agonists, but not delta agonist, rescue of dopaminergic neurons in the 6-OHDA parkinsonian model is associated with inhibition of microglial activation and MMP expression. Journal of neuroimmunology. 2012 May 15;246(1–2):69–77. doi: 10.1016/j.jneuroim.2012.03.010 22498097
6. Hetzel M, Walcher D, Grub M, Bach H, Hombach V, Marx N. Inhibition of MMP-9 expression by PPARgamma activators in human bronchial epithelial cells. Thorax. [Research Support, Non-U.S. Gov't]. 2003 Sep;58(9):778–83. 12947137
7. Herrett E, Thomas SL, Schoonen WM, Smeeth L, Hall AJ. Validation and validity of diagnoses in the General Practice Research Database: a systematic review. British journal of clinical pharmacology. 2010 Jan;69(1):4–14. doi: 10.1111/j.1365-2125.2009.03537.x 20078607
8. Garcia Rodriguez LA, Perez Gutthann S. Use of the UK General Practice Research Database for pharmacoepidemiology. British journal of clinical pharmacology. 1998 May;45(5):419–25. 9643612
9. Williams T, van Staa T, Puri S, Eaton S. Recent advances in the utility and use of the General Practice Research Database as an example of a UK Primary Care Data resource. Therapeutic Advances in Drug Safety. 2012 Apr;3:89–99. doi: 10.1177/2042098611435911 25083228
10. van der Mark M, Nijssen PC, Vlaanderen J, Huss A, Mulleners WM, Sas AM, et al. A case-control study of the protective effect of alcohol, coffee, and cigarette consumption on Parkinson disease risk: time-since-cessation modifies the effect of tobacco smoking. PLoS ONE. [Research Support, Non-U.S. Gov't]. 2014;9(4):e95297. doi: 10.1371/journal.pone.0095297 24788751
11. Morens DM, Grandinetti A, Reed D, White LR, Ross GW. Cigarette smoking and protection from Parkinson's disease: false association or etiologic clue? Neurology. [Research Support, U.S. Gov't, P.H.S.Review]. 1995 Jun;45(6):1041–51. 7783862
12. Dye RV, Miller KJ, Singer EJ, Levine AJ. Hormone replacement therapy and risk for neurodegenerative diseases. International journal of Alzheimer's disease. 2012;2012:258454. doi: 10.1155/2012/258454 22548198
13. Becker C, Jick SS, Meier CR. Use of antihypertensives and the risk of Parkinson disease. Neurology. [Comparative Study]. 2008 Apr 15;70(16 Pt 2):1438–44. doi: 10.1212/01.wnl.0000303818.38960.44 18256367
14. Marras C, Hincapie CA, Kristman VL, Cancelliere C, Soklaridis S, Li A, et al. Systematic review of the risk of Parkinson's disease after mild traumatic brain injury: results of the International Collaboration on Mild Traumatic Brain Injury Prognosis. Archives of physical medicine and rehabilitation. [Research Support, Non-U.S. Gov't Review]. 2014 Mar;95(3 Suppl):S238–44. doi: 10.1016/j.apmr.2013.08.298 24581909
15. Litvan I, Halliday G, Hallett M, Goetz CG, Rocca W, Duyckaerts C, et al. The etiopathogenesis of Parkinson disease and suggestions for future research. Part I. Journal of neuropathology and experimental neurology. [Review]. 2007 Apr;66(4):251–7. 17413315
16. Hu G, Jousilahti P, Nissinen A, Antikainen R, Kivipelto M, Tuomilehto J. Body mass index and the risk of Parkinson disease. Neurology. [Comparative Study Research Support, Non-U.S. Gov't]. 2006 Dec 12;67(11):1955–9. 17159100
17. Lu L, Fu DL, Li HQ, Liu AJ, Li JH, Zheng GQ. Diabetes and risk of Parkinson's disease: an updated meta-analysis of case-control studies. PLoS ONE. 2014;9(1):e85781. doi: 10.1371/journal.pone.0085781 24465703
18. Wahlqvist ML, Lee MS, Hsu CC, Chuang SY, Lee JT, Tsai HN. Metformin-inclusive sulfonylurea therapy reduces the risk of Parkinson's disease occurring with Type 2 diabetes in a Taiwanese population cohort. Parkinsonism & related disorders. [Research Support, Non-U.S. Gov't]. 2012 Jul;18(6):753–8.
19. Sadeghian M, Camarata MA, Broom L, Bindra H, Davis JB, Medhurst AD, et al. Microglial activation in the 6-hydroxydopamine Parkinson's disease model is associated with a time dependent increase in neuronal expression of MMP-3 and MMP-9 and a dissociation of the CD200R neuronal/microglial complex. J Neuroinflammation.
20. Singh S, Loke YK, Furberg CD. Long-term risk of cardiovascular events with rosiglitazone: a meta-analysis. JAMA: the journal of the American Medical Association. [Meta-Analysis]. 2007 Sep 12;298(10):1189–95.
21. U.S. Food and Drug Administration. FDA requires removal of some prescribing and dispensing restrictions for rosiglitazone-containing diabetes medicines FDA Drug Safety Communications [serial on the Internet]. [cited October 2014: http://www.fda.gov/downloads/Drugs/DrugSafety/UCM381108.pdf.
22. Azoulay L, Yin H, Filion KB, Assayag J, Majdan A, Pollak MN, et al. The use of pioglitazone and the risk of bladder cancer in people with type 2 diabetes: nested case-control study. BMJ. [Research Support, Non-U.S. Gov't]. 2012;344:e3645. doi: 10.1136/bmj.e3645 22653981
23. Lewis JD, Habel L, Quesenberry C, Mamtani R, Peng T, Bilker WB, et al. Proteinuria testing among patients with diabetes mellitus is associated with bladder cancer diagnosis: potential for unmeasured confounding in studies of pioglitazone and bladder cancer. Pharmacoepidemiology and drug safety. 2014 Jun;23(6):636–45. doi: 10.1002/pds.3619 24764283
24. Randy LH, Guoying B. Agonism of Peroxisome Proliferator Receptor-Gamma may have Therapeutic Potential for Neuroinflammation and Parkinson's Disease. Current neuropharmacology. 2007 Mar;5(1):35–46. 18615152
25. Pagel-Langenickel I, Bao J, Joseph JJ, Schwartz DR, Mantell BS, Xu X, et al. PGC-1alpha integrates insulin signaling, mitochondrial regulation, and bioenergetic function in skeletal muscle. The Journal of biological chemistry. 2008 Aug 15;283(33):22464–72. doi: 10.1074/jbc.M800842200 18579525
26. Miglio G, Rosa AC, Rattazzi L, Collino M, Lombardi G, Fantozzi R. PPARgamma stimulation promotes mitochondrial biogenesis and prevents glucose deprivation-induced neuronal cell loss. Neurochemistry international. 2009 Dec;55(7):496–504. doi: 10.1016/j.neuint.2009.05.001 19442697
27. Sauerbeck A, Gao J, Readnower R, Liu M, Pauly JR, Bing G, et al. Pioglitazone attenuates mitochondrial dysfunction, cognitive impairment, cortical tissue loss, and inflammation following traumatic brain injury. Experimental neurology. 2011 Jan;227(1):128–35. doi: 10.1016/j.expneurol.2010.10.003 20965168
28. Zheng B, Liao Z, Locascio JJ, Lesniak KA, Roderick SS, Watt ML, et al. PGC-1alpha, a potential therapeutic target for early intervention in Parkinson's disease. Sci Transl Med. [Meta-Analysis Research Support, N.I.H., Extramural Research Support, Non-U.S. Gov't]. 2010 Oct 6;2(52):52ra73.
29. Shin JH, Ko HS, Kang H, Lee Y, Lee YI, Pletinkova O, et al. PARIS (ZNF746) repression of PGC-1alpha contributes to neurodegeneration in Parkinson's disease. Cell. [Research Support, N.I.H., Extramural Research Support, Non-U.S. Gov't]. 2011 Mar 4;144(5):689–702. doi: 10.1016/j.cell.2011.02.010 21376232
30. Hernan MA, Logroscino G, Rodriguez LA. A prospective study of alcoholism and the risk of Parkinson's disease. Journal of neurology. 2004 Oct;251 Suppl 7:vII14–7. 15505749