Cilostazol is a unique antiplatelet agent that has been commercially available for two decades. It is a potent antiplatelet agent and possesses vasodilatation and antiproliferative effects. It has been broadly applied in the treatment of peripheral artery disease (PAD) and also for patients with PAD undergoing endovascular procedures. In addition, a randomized placebo controlled trial has investigated the role of cilostazol in patients with coronary artery disease (CAD) undergoing primary coronary intervention (PCI) along with the use of aspirin and/or clopidogrel (Rogers et al., 2015).

Cilostazol has multifactorial pharmacological properties and broad spectrum of pharmacological actions. Its main effect includes selective inhibition of cellular phosphodiesterase (PDE) type III, which augments the effect of cyclic adenosine monophosphate (cAMP) metabolism within the platelet (Pan et al., 1994; Schror, 2002). Since cAMP is involved in the pathway which controls platelet aggregation and vasomotor function, inhibition of PDE III could reversibly inhibit platelet aggregation and results in vasodilation (Schror, 2002). Besides, cilostazol has also been shown to reduce triglycerides, raise high-density lipoprotein levels and increase adenosine accumulation in the interstitium, all of which provide protective effects for patients with stroke, CAD and PAD. Nevertheless, inhibition of PDE III may lead to a positive inotropic effect, meta-analysis for randomized trials of PDE III inhibitors versus placebo in heart failure (HF) patients revealed that PDE III inhibitors were responsible for an increase in mortality rate compared with placebo in patients suffering from chronic HF. Thus, the use of PDE III inhibitors is prohibited in HF patients. The above results were mainly obtained from the trials involving intravenous medication for HF (e.g., milrinone, vesnarinone). Whether the use of cilostazol is associated with development of HF or worsening of HF is still unknown.

Diabetes is a major risk factor for development of PAD. The number of prescriptions for cilostazol in diabetic patients have increased gradually and, therefore, understanding the role of cilostazol in diabetic patients has become even more important. Therefore, this national-wide population-based study aimed to examine whether the use of cilostazol was associated with hospitalization for heart failure (HHF) in the diabetic population.

This study was approved by our hospital’s Research Ethics Committee. The case-crossover study design was used to assess the relationship between cilostazol exposure and risk of HF that led to hospitalization. Instead of using propensity-matched controls, each patient served as his/her own control in the case-crossover study, so that stable confounders, unmeasured or poorly measured, would not influence the results. In addition, because diabetic patients might receive complicated anti-hypertensive and anti-diabetic drug therapy and switch from one class of drug to other classes in order to attain optimal blood pressure and glycemic control, it was difficult to find appropriate comparison groups. The major advantage of this design is that stable risk factors such as cigarette smoking or other cardiovascular risk factors will not confound the analysis results, despite the fact that they are not recorded in the claim database.

A total of 79,100 patients with diabetes who were ≥ 20 years of age and who were first hospitalized for CHF between 2009 and 2011 were identified. After further excluding those who had been hospitalized for any reason 120 days prior to index HHF, a total of 47,506 patients (48.2% male, mean age 72.7 years) were included in the final analysis (Figure 1).

Table 1 summarizes the proportion of patients with comorbidities at index hospitalization. Approximately 63.7% of the cohort had hypertension, 47.3% had ischemic heart disease, 12.4% had a history of myocardial infarction, and 20.5% had chronic kidney disease. The exposure of cilostazol, anti-hypertensive, and anti-diabetic agents, as well as concomitant medications during the case period (1–30 days before index hospitalization) and control period (91–120 days before index hospitalization) is shown in Table 2. A higher proportion of patients took cilostazol, anti-diabetic agents (except for metformin), aspirin, angiotensin converting enzyme inhibitors, beta-blockers, calcium channel blockers, diuretics, nitrates, digitalis, inhaled bronchodilators, aminophylline, systemic antibiotics, and NSAIDs in the case period than in the control period.

Table 3 presents crude and adjusted ORs and their 95% CIs for cilostazol and NSAIDs by conditional logistic regression. A modestly increased risk of hospitalization for HHF was found for the exposure of cilostazol with a crude OR of 1.57 (95% CI: 1.35–1.84). After controlling for potential time-varying confounders including drugs potentially associated with HF, cilostazol was associated with an increased risk of HHF (OR: 1.35, 95% CI: 1.14–1.59) for hospitalized CHF. A significantly elevated OR was also evident for NSAIDs (adjusted OR 1.69, 95% CI: 1.62–1.76).

After further adjustment for time-varying co-morbidities including hypertension, ischemic heart disease, cerebrovascular disease, chronic kidney disease, or further adjustment for surrogate measures of glycemic control including number of outpatients visits and the number of A1c test prescribed, the ORs remained essentially the same (Supplementary Table 2).

Sensitivity analyses using different definitions of control period (ranging from 31–60, 61–90, to 121–150 days before index date) yielded adjusted ORs of 1.43 (95% CI: 1.14–1.79), 1.31 (95% CI: 1.09–1.57) and 1.23 (95% CI: 1.06–1.44), respectively (Table 4A). An analysis excluding recent 7 days of drug exposure prior to index date showed similar results suggesting the robustness of our study findings (Table 4A). Additional sensitivity analyses using longer case and control time windows (i.e., 60 and 90 days) showed a similar risk estimate in the analysis using a window length of 60 days; however, the OR were attenuated and became non-significant in the analysis using a window length of 90 days (Table 4B). In contrast, NSAIDs were consistently associated with significantly higher risks, regardless of the definition and window length used for the case and control periods.

The subgroup analysis showed that the increased risk of HHF associated with cilostazol exposure was uniform across pre-defined strata by gender, age, prior history of HF, and chronic kidney disease, suggesting no significant modification effect by these characteristics (Table 5). The same analysis was performed to analysis cilostazol’s effect for overall mortality (Supplementary Table 3). Similar findings were found and recent us of cilostazol was associated with an OR of 3.90 (95% CI: 3.51–4.33) for increased mortality risk (Supplementary Table 4). The results were consistent after adjustment for comorbidities and drug use (Supplementary Table 5).

In this study, we analyzed a nationwide health insurance claims database and found an increased risk of HHF with use of cilostazol in diabetic patients. The effect of cilostazol was only slightly lower than that associated with the use of non-steroid anti-inflammatory drugs (NSAID). We also observed that the elevated risk of HHF associated with cilostazol exposure was uniform across all pre-defined subgroups after controlling for potential risk factors.

PDE3-selective inhibitors such as milrinone and vesnarinone have been used clinically to treat acute congestive HF (Benotti et al., 1978). Blockage of PDE 3 could lead to increased levels of intracellular cAMP and protein kinase A which could result in smooth muscle cell relaxation and attenuation of myosin phosphorylation (McDaniel et al., 1994). On the other hand, elevated cAMP levels enables L-type Ca2+ channels and a component of a delayed-rectifier potassium channel signal transduction pathway in the sinoatrial node (Kodama-Takahashi et al., 2003). Therefore, drugs that inhibit PDE 3 pathways would cause a positive chronotropic effect and inotropic action which have been demonstrated to increase left ventricular ejection fraction and stroke volume (Baim et al., 1983; Jaski et al., 1985; Rauch et al., 1991). However, clinical studies [such as the Prospective Randomized Milrinone Survival Evaluation (PROMISE) (Packer et al., 1991) and the Vesnarinone trial (VEST) (Feldman et al., 1993) studies] have evaluated the long-term effects of PDE 3 administration in congestive HF patients and reported increased cardiovascular mortality. Several studies have also shown that long-term elevation of cAMP could cause adverse effects on chronic HF patients (Movsesian, 2003; Phrommintikul and Chattipakorn, 2006). Increased levels of cAMP trigger early and delayed after-depolarization and enhance the risk of lethal ventricular arrhythmia. Besides, the inotropic effects of PDE 3 inhibitors could cause higher myocardial oxygen consumption and increase various arrhythmia over time (Naccarelli and Goldstein, 1989; Sunderdiek et al., 2000).

Cilostazol also inhibits PDE 3 selectively. It has potent vasodilating and antithrombotic effects and has been used in the treatment of PAD (Dawson et al., 1998). Therefore, the FDA contraindicated the use of cilostazol in HF of any severity. Nevertheless, no clinical study has directly investigated the effects of cilostazol in HF patients. Although cilostazol is a PDE 3 inhibitor, the effect on myocytes differs from other PDE 3 inhibitors. The weaker inotropic effect of cilostazol compared with milrinone lessens long-term cardiac side effects (Cone et al., 1999). Cilostazol also blocks adenosine uptake, and, subsequently, the interstitial adenosine elevation in the heart could attenuate cAMP response and limit the long-term harmful effects of cAMP activation (Wang et al., 2001). In addition, in smooth muscle and platelets, elevated interstitial and circulating adenosine levels by cilostazol could potentiate its cAMP-raising effect through inhibiting PDE3 and, thereby, augmenting antiplatelet and vasodilatory effects which could be beneficial for prognosis of congestive HF (Liu et al., 2001). For subjects with myocardium ischemic/reperfusion, cilostazol could decreased plasma IL-6, IL-1beta and TNF-alpha levels, and activate the PPARγ/JAK2/STAT3 pathway. Through the above pathway, cilostazol is possible to limit myocardial inflammation and apoptosis after ischemic/reperfusion injury. These results may provide a rationale for new therapy with cilostazol in patients with ischemic cardiomyopathy and may possible restore cardiac function (Li et al., 2017). Taken together, the use of cilostazol in distinct groups should be reinvestigated.

The biological mechanism(s) behind the increased risk of HHF associated with cilostazol is unclear. Atarashi et al. (1998) demonstrated that cilostazol increased heart rate and improved symptoms in patients with symptomatic bradyarrhythmia. There is no cardiac electrophysiological study that has investigated the strength of cilostazol. In contrast, there is evidence for its possible undesirable tachyarrhythmia effects (Barta et al., 2008; Shinohara et al., 2010). Elevated resting heart rate was associated with increased risk for incident HF in asymptomatic high risk patients and higher heart rate was an independent risk factor for HHF in the subclinical group (Opdahl et al., 2014). Few studies specifically evaluated the effect of cilostazol on HF in the patients with diabetes. One recent meta-analysis evaluated 10 randomized case-control trials in the diabetic population, mostly with reported outcomes of major adverse cardiovascular events or platelet aggregation function (Bundhun et al., 2015). However, all the above trials failed to provide information about the risk of HHF associated with cilostazol use.

Our current study compared recent use of cilostazol (1–30 days) with controls (remote use) to document the influence of the drug over a short period. In the sensitivity analysis, we excluded the recent 7 days of drug exposure and found similar risk for the drug. We also used different definitions of control periods or longer case and control time windows (i.e., 60 and 90 days), assuming the statistical significance associated with discordant cilostazol use might diminish. In contrast, the effects of NSAIDs were consistently associated with significantly higher risks, regardless of the definition or window length used for the case and control periods. These findings suggested that exposure to cilostazol caused an accumulated HHF risk with a duration about 2 months but not longer. Instead, the well-known risks for NSAID in diabetic patients are consistent around 3 months. As we have discussed, cilostazol has a much weaker inotropic effect than other PDE3 inhibitors and could lessens long-term cardiac side effects. However, cilostazol might also increase heart rate and hence lead to higher HHF risk during the initial 2-month period. On the other hand, our data was consistent with the previous conclusion that NSAID use is associated with a higher risk of myocardial infarction, stroke, and hospitalizations for both ischemia and HF (Kohli et al., 2014) which further validated the reliability of our cohort study.

A very recent study used five population-based healthcare databases from four European countries and showed that current use of any NSAID (use in preceding 14 days) was found to be associated with a 19% increased risk of HHF (adjusted OR: 1.19, 95% CI: 1.17–1.22), compared with past use of any NSAIDs (use > 183 days in the past). The results were consistent with our findings of an association between NSAID use and HHF. Hence, we verified the higher risk of developing HHF associated with NSAID use compared with that of cilostazol in diabetic patients (Arfe et al., 2016).

In conclusion, our study found that cilostazol use was associated with a modestly increased risk of HHF. Further studies are warranted to explore underlying mechanisms and to confirm the association.

C-HC had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of data analysis. C-KW and J-WL participated in the study concept and design. L-CW acquired the data. C-KW and J-WL also contributed to the analysis and interpretation of data.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.