Original Contribution

The Dose and Timing of Fentanyl Impacts on Ticagrelor Absorption and Platelet Inhibition During Percutaneous Coronary Intervention: The PACIFY Randomized Clinical Trial

Rakesh R. Goli, BA1;  Khalil Ibrahim, MD2;  Rohan Shah, MD1;  Thomas S. Kickler, MD3;  William A. Clarke, PhD3;   Jon R. Resar, MD2;  Steven P. Schulman, MD2;  John W. McEvoy, MBBCh, MHS4,5,6

Rakesh R. Goli, BA1;  Khalil Ibrahim, MD2;  Rohan Shah, MD1;  Thomas S. Kickler, MD3;  William A. Clarke, PhD3;   Jon R. Resar, MD2;  Steven P. Schulman, MD2;  John W. McEvoy, MBBCh, MHS4,5,6

Abstract: Objective. In this secondary analysis of the PACIFY randomized trial, we assessed whether dose and timing of fentanyl have implications for the pharmacokinetics and pharmacodynamics of ticagrelor loading during percutaneous coronary intervention (PCI). Methods. Among 212 patients undergoing clinically indicated coronary angiography, a total of 70 required PCI and received 180 mg oral ticagrelor. Of these, thirty-two patients received no fentanyl and 38 received fentanyl (with variability in the timing of administration and cumulative dose among those randomized to fentanyl, given that both were provided at the interventional cardiologist’s discretion). A time-weighted cumulative fentanyl exposure variable was calculated based on total dose of fentanyl and proximity in time of fentanyl administrations to the ticagrelor load. Patients were stratified based on receiving above or below the median time-weighted cumulative dose. Outcomes included ticagrelor concentrations by mass spectrometry (24-hour area under the curve [AUC]0-24) and platelet function measured using both VerifyNow platelet reactivity units (PRU) and light-transmission aggregometry (LTA). Results. Unadjusted ticagrelor AUC0-24 was significantly lower across the categories of increasing fentanyl exposure (P=.02). In adjusted regression models, this difference only remained when comparing the no-fentanyl group with the time-weighted cumulative dose above the median group (P=.04). Similarly, with the no-fentanyl group as the reference, adjusted models testing 2-hour PRU and LTA values demonstrated significant differences (with less platelet inhibition for both tests) only among those with time-weighted cumulative fentanyl exposures above the median value (5.1 µg/min). Conclusions. We have previously shown that fentanyl slows absorption of oral ticagrelor, attenuating its effect on platelet inhibition. We now demonstrate this mechanism appears to be dose- and time-dependent.

J INVASIVE CARDIOL 2019;31(9):265-271.

Key words: fentanyl, platelet inhibitor, ticagrelor

Rapid and effective platelet inhibition is an essential component of percutaneous coronary intervention (PCI). This requires a dual-antiplatelet combination of aspirin and an oral P2Y12 inhibitor, with newer oral P2Y12 agents such as ticagrelor and prasugrel increasingly utilized in place of clopidogrel.1

As timely absorption of these orally administered P2Y12 inhibitors is critical in an acute PCI setting, any therapy that could slow gastric emptying and the resulting absorption of oral medications demands scrutiny. A common PCI sedation protocol, used throughout North America but less so in Europe,2 involves opiate administration prior to the start of the procedure in combination with subcutaneous lidocaine and intravenous midazolam, with further as-required doses provided throughout the procedure.3 As opiates are known to slow gastrointestinal motility, concerns have been raised regarding co-administration of opiates and P2Y12 inhibitors.4 Previous morphine-based research has validated such concerns.5-9

With fentanyl, a more rapid-acting and potent intravenous opiate than morphine, being commonplace for procedural analgesia during coronary angiography and PCI, the goal of the Platelet Aggregation with tiCagrelor Inhibition and FentanYl (PACIFY) trial was to analyze the pharmacokinetic and pharmacodynamic interactions between fentanyl and ticagrelor in patients undergoing clinically indicated PCI.10 The main results of the trial demonstrated significantly reduced concentrations of ticagrelor and impaired antiplatelet effects in patients randomized to fentanyl during PCI, with little to no improvement in reported pain and anxiety.11,12

In this secondary analysis of the PACIFY trial, we studied whether the timing and dosage of fentanyl during PCI impacted ticagrelor absorption and resultant platelet inhibition. This was possible because the PACIFY trial protocol did not dictate the dosing or frequency of fentanyl among patients who were randomized to the fentanyl arm; rather, dosing was at the cardiologist’s discretion and as such there was a range of doses and frequencies of fentanyl administration among patients.


Study design and inclusion/exclusion criteria. The PACIFY trial was a single-center randomized trial conducted at the Johns Hopkins Hospital, enrolling 212 patients who presented for a clinically indicated elective coronary angiography with possible PCI. Patients were randomized to receive the procedure either with or without intravenous fentanyl in a 1:1 fashion, with doses determined at the discretion of the interventional cardiology provider and administered as needed throughout the procedure. All study patients received both local anesthetic and intravenous midazolam at the start and as needed throughout the procedure. All patients and assessors of outcomes were blinded to fentanyl allocation status, while treating interventional cardiologists were not. In exceptional circumstances, the study protocol allowed the interventional cardiologist to administer fentanyl to those randomized into the no-fentanyl arm of this trial, but only for the bail-out treatment of active procedural pain.

All patients enrolled in the study were >18 years of age, were undergoing a clinically indicated elective coronary angiography, were able to swallow oral medications, and were able to provide informed consent for the study. Patients were excluded from participation in the study if they met any of the following criteria: use of P2Y12 inhibitors in the past 14 days; known coagulation disorders; preprocedural treatment with an anticoagulant; platelet count <100,000/mm3; pregnancy; impaired renal function, as determined by glomerular filtration rate <45 mL/min/1.73 m2; impaired hepatic function, as determined by previous medical history; prior or planned transcatheter aortic valve replacement; or any contraindication to ticagrelor or opiate administration.

As the research team was almost always unaware of coronary vessel pathology at the time of study consent and randomization preangiography, only 70 of the 212 enrolled patients undergoing coronary angiography met the criteria necessary for a clinically indicated transition to PCI. The remaining patients were either treated medically or were referred for bypass surgery. If transitioning to PCI, study patients were loaded in the cardiac catheterization lab with a 180 mg oral dose of ticagrelor at the conclusion of the diagnostic portion of the coronary angiography, prior to intervention. For these 70 patients undergoing PCI, baseline blood samples were collected before ticagrelor loading and were subsequently collected at 0.5, 1, 2, 4, and 24 hours post loading.

Study outcomes. The primary outcome of the PACIFY trial was ticagrelor plasma concentration, as measured by liquid chromatography-tandem mass spectrometry at 0.5, 1, 2, 4, and 24 hours. Baseline concentrations were assumed to be zero, as any potential patient who had taken a P2Y12 inhibitor in the past 14 days was excluded from the study. Values at each timepoint were used to estimate an area under the concentration-time curve between 0 and 24 hours (AUC0-24).

Pharmacodynamic assessment included analysis of platelet function and high-sensitivity troponin-I (hs-TnI). Platelet function was measured by P2Y12 reaction units (PRU) with the VerifyNow system (Accriva Diagnostics) at baseline (preloading), 0.5, 1, 2, 4, and 24 hours. Platelet function was additionally assessed by light transmission platelet aggregometry at 2 hours after ticagrelor loading, before and after agonism with adenosine diphosphate (ADP, reported as %ADP response). Full details of our platelet function testing can be found in the design paper.10 In prespecified analyses,10,12 patients with 2-hour PRU values ≥235 or 2-hour %ADP response ≥46% were labeled as having high platelet reactivity (HPR).

Myocardial injury was assessed using a research-only hs-TnI assay at 2 hours post PCI, with a limit of detection of 1.2 ng/L and a 99th percentile (upper reference limit) of 26.2 ng/L. Additionally, self-reported pain and anxiety were recorded 2 hours post procedure through a 4-question survey with a Likert-like scale of 1-10 (with 10 being the most severe) and was corroborated with maximum intraprocedural pain, as documented by the procedure-room nurse in the electronic medical record.

Statistical analysis. To test the association of fentanyl timing and dose with the above endpoints, we derived a time-weighted cumulative dose variable as follows: (1) we recorded the dose of fentanyl provided at each discrete administration during the procedure; (2) we divided this dose by the absolute time difference between ticagrelor loading and each discrete fentanyl administration; and (3) for each patient, we then summed these time-weighted doses during the course of the entire procedure (ie, from catheterization lab entry to exit). For example, a patient who received 25 µg of fentanyl at the start of angiography at 1200 hours, who was loaded with ticagrelor at 1230 hours, and who required 2 further 50 µg doses of fentanyl at 1245 and 1300 hours would have a time-weighted cumulative fentanyl dose value of 5.83 µg/min ([25/30] + [50/15] + [50/30]). This exposure variable was prespecified in our analysis plan, as we hypothesized that the timing of fentanyl administration (ie, its proximity to ticagrelor loading) was important to model, not just the amount of fentanyl received.

Patients who did not receive fentanyl were all assigned a time-weighted cumulative fentanyl dose of zero. Importantly, among the 35 patients randomized to the no-fentanyl arm of this trial, three crossed over into the fentanyl arm during their procedure for bail-out treatment of pain. These 3 crossovers were included in the as-treated fentanyl category of our analysis and were also assigned a time-weighted cumulative fentanyl dose according to the above formula (note that in our primary manuscript we reported that only 2 no-fentanyl patients crossed over to receive fentanyl; however, during the conduct of this analysis, we found 1 more patient who crossed over who had previously been missed).11,12

We then divided our study sample into three groups: (1) those who did not receive fentanyl (time-weighted cumulative dose = 0); (2) those who received fentanyl and had a time-weighted cumulative dose below the median value; and (3) those who received fentanyl and had a time-weighted cumulative dose at or above the median value. Baseline demographic and procedural characteristics among these three groups were then compared using Chi-squared testing for proportions and, depending on the normality of data, t-testing or Wilcoxon rank-sum for continuous variables.

With the no-fentanyl group as the reference, analyses were performed using multivariable linear and logistic regression adjusted for age, sex, hypertension status, diabetes status, smoking status, history of myocardial infarction, aspirin use, baseline international normalized ratio (INR), and baseline (preloading) PRU.

Finally, in sensitivity analyses, we constructed multivariable-adjusted linear models with the time-weighted cumulative fentanyl dose as a continuous exposure variable on our outcomes, rather than as a categorical exposure, and we also modeled the summed total dose of fentanyl provided during the procedure as a continuous exposure (the latter being a simple addition of the sum of fentanyl doses administered without consideration of their proximity in time to ticagrelor administration). All analyses were conducted using Stata, version 15 (Stata, Inc) and two-sided P-values <.05 were considered statistically significant.


A total of 212 patients were enrolled and randomized in this trial, of whom 70 required ticagrelor loading to proceed with clinically indicated PCI at the conclusion of diagnostic angiography. Of these 70 patients, a total of 38 patients received fentanyl at some point during their procedure. The median value of the time-weighted cumulative fentanyl dose variable derived for the 38 patients who received fentanyl was 5.1 µg/min (interquartile range [IQR], 2.4-10.7 µg/min; absolute range, 0.7-35.6 µg/min). Comparing the three time-weighted cumulative fentanyl exposure categories (no fentanyl vs fentanyl dose below median vs fentanyl dose above median), there was a significant difference in the mean summed total fentanyl dose administered (0 µg vs 61.8 ± 36 µg vs 120.8 ± 53 µg, respectively).

There were no statistically significant differences in the amount of midazolam sedation provided in the three fentanyl exposure groups (Table 1). Patient characteristics among those receiving PCI in the three fentanyl groups were statistically different in regard to hypertension status (P=.04), prior myocardial infarction (P=.05), and home use of aspirin (P=.02) (Table 1). There were no significant differences in patient-reported or nurse-reported pain levels among the three fentanyl exposure groups, nor were there differences in patient-reported anxiety levels (Table 2).

Pharmacokinetics. Analysis of ticagrelor concentrations revealed significant differences in crude AUC0-24 mean values between the fentanyl exposure groups (P=.02), with lower ticagrelor concentrations among those with higher levels of fentanyl exposure (Table 3; Figure 1). After adjustment for potential confounders, there was no longer any difference in ticagrelor AUC0-24 between the no-fentanyl group and the below-median fentanyl group (P=.30). However, compared with the reference category, an independent association remained between lower ticagrelor concentrations and fentanyl exposure above the median, with an average adjusted difference of -2008 ng•h−1 mL between the groups (P=.04) (Table 4).

Pharmacodynamics. As expected, platelets became progressively more inhibited in the 4 hours after the ticagrelor load in all three fentanyl exposure groups. However, consistent with the above differences in pharmacokinetics, platelet inhibition at the 1-hour, 2-hour, and 4-hour timepoints tended to be less pronounced among those with the highest levels of fentanyl exposure (Table 3). For example, statistically more patients in the uppermost fentanyl exposure category had HPR according to the PRU definition at the 1-hour (P=.05), 2-hour (P<.01), and 4-hour marks (P=.04). In terms of continuous PRU values according to fentanyl exposure category, there was a trend toward lower average platelet inhibition at 1 hour (P=.08) and a more convincing difference at 2 hours (P<.01), with similar statistically significant differences at 2 hours when platelet function was measured continuously using %ADP response by platelet light aggregometry (Table 3).

When comparing between-group differences in platelet function at 2 hours, both by PRU and by %ADP response, only the group with time-weighted cumulative fentanyl exposures above the median differed statistically (ie, there were no significant differences in platelet function between the no-fentanyl group and the below-median fentanyl group; Figure 2). In regression models adjusting for potential confounders, with the no-fentanyl group as the reference, the lower levels of platelet inhibition in the above-median fentanyl group remained independently significant at the 2-hour mark (Table 4).

Furthermore, crude analysis of mean high-sensitivity troponin-I levels at 2 hours post ticagrelor loading demonstrated monotonic increases in myocardial injury according to time-weighted cumulative fentanyl exposure category (Table 3). However, these troponin differences at 2 hours were no longer significant in multivariable adjusted models (Table 4).

Sensitivity analyses. These analyses evaluated the associations between our outcomes and the following: (1) the time-weighted cumulative fentanyl dose variable as a continuous exposure; or (2) the summed total fentanyl dose as a continuous exposure; before (3) then contrasting the strength of associations between these two exposures and our outcomes (ie, modeling the fentanyl dose exposure variable with or without weighting for time of administration relative to the ticagrelor load). In unadjusted linear regression analyzing 2-hour PRU values, there was a 3.5 unit increase in PRU (ie, less platelet inhibition) for every 1 µg/min increase in time-weighted cumulative fentanyl dose administered (95% confidence interval [CI], 0.5-6.6; P=.02) (Supplemental Figure S1A; supplemental materials available at www.invasivecardiology.com). By light aggregometry, there was a 1% increase in %ADP response (again consistent with less platelet inhibition) for every 1 µg/min increase in time-weighted cumulative fentanyl dose (95% CI, 0.1-2.0; P=.03) (Supplemental Figure S1B). Consistent with these pharmacodynamic findings, every 1 µg/min increase in time-weighted cumulative fentanyl dose was associated with a 104 ng•h−1 mL lower value for ticagrelor AUC0-24 (95% CI, 7-201; P<.01) (Supplemental Figure S1C). In the adjusted regression models, these findings remained significant or borderline significant for 2-hour PRU (P=.04), 2-hour %ADP response (P=.04), and ticagrelor AUC0-24 (P=.06) (Supplemental Figure S1).

When using the summed total fentanyl dose as a continuous exposure variable in linear models, the adjusted association between fentanyl dose and our outcomes was robust for 2-hour PRU (P=.02) and ticagrelor AUC0-24 (P=.04), but was no longer statistically significant for 2-hour %ADP response (P=.08) (Supplemental Figure S2). In addition, the R2 statistics from models using the summed total fentanyl dose as the exposure variable were generally lower than for models where the time-weighted cumulative fentanyl dose was used as the exposure variable, indicating that model fit appeared better with the latter (Supplemental Figures S1 and S2).


The results from this secondary analysis of the PACIFY trial support a dose-dependent action of fentanyl on P2Y12 receptor inhibitor absorption and platelet inhibition during PCI, with higher summed doses of fentanyl being more associated with delayed ticagrelor absorption and reduced platelet inhibition. In addition, using a time-weighted cumulative fentanyl exposure variable, which also accounted for the proximity of fentanyl administration to the ticagrelor load, our results also suggest that the closer the fentanyl administration to ticagrelor loading, the stronger the effect on impaired absorption of this oral medication.

While the PACIFY trial was the first trial to demonstrate the negative effects of fentanyl administration on P2Y12 inhibitor function, previous research has consistently shown attenuation of the antiplatelet effects of P2Y12 inhibitors with administration of morphine, both in healthy subjects and in patients with coronary artery disease.5,8 However, whether the association between opioid administration and reduced P2Y12 inhibitor absorption has a dose-response nature has been previously unknown. The hypothesized mechanism of opioid-induced reduction in gastric motility provides an intuitive explanation as to why the dose and timing of fentanyl might have implications for ticagrelor absorption.

Because the dose and timing of fentanyl were non-randomly assigned in the PACIFY trial, it was important to address potential confounding. The above conclusions are buttressed by highly robust associations between fentanyl dose and timing and our 2-hour platelet outcomes in multivariable regression analyses. Furthermore, in sensitivity analyses, we used linear regression to model two forms of fentanyl exposure (both on continuous numeric scales); first, the time-weighted cumulative fentanyl dose (which accounted for both the amount of fentanyl provided and the proximity of fentanyl dosing to the ticagrelor load); and second, the summed total fentanyl dose (which was a simple addition of all fentanyl doses during the procedure without respect to their timing). Comparison of these two modeling strategies demonstrated that the time-weighted cumulative fentanyl exposure was more strongly associated with our endpoints – supporting the initial hypothesis that timing of the fentanyl dose does appear to matter from a pathophysiological perspective. However, the difference between these two modeling strategies was marginal, so we believe it is reasonable to conclude that the total dose (rather than the exact timing) of fentanyl is the more important parameter to consider.

Study limitations. As with all trials, the results of this study do not come without important caveats. Importantly, the trial was not powered or designed to test the dose-response implications of fentanyl on our outcomes (rather it was designed simply to test whether fentanyl administration [yes/no], without respect to dose, delays ticagrelor absorption), and because doses were provided at the discretion of the interventional cardiologist in a non-randomized fashion, we cannot claim causality for our results. Nevertheless, we are confident in our results, as there is strong biologic plausibility to our dose-response findings and they were almost all robust to rigorous adjustment for confounders. Furthermore, despite the relatively small numbers of patients enrolled in PACIFY, our highly significant crude P-values suggest that type 1 error (false positive findings by chance) is unlikely. Confirmatory studies will nonetheless be necessary. Finally, while it is tempting to suggest that either a time-weighted exposure above the median of 5 µg/min (ie, no more than 25 µg of fentanyl within 5 minutes of ticagrelor loading or no more than 50 µg of fentanyl within 10 minutes of loading, and so on) or, more simply, a summed total dose over the median of 100 µg of fentanyl should be avoided and that lower doses might be safer for timely absorption of ticagrelor, it must be emphasized that this study was not powered to test whether a certain dose threshold exists below which fentanyl is safe to administer after ticagrelor loading.


The use of fentanyl in the setting of PCI inhibits the absorption and action of P2Y12 inhibitors in a dose-dependent fashion. Interventional cardiologists should further consider the necessity of opiates in cardiac catheterization and, if needed, avoid administration temporally proximate to ticagrelor loading and, more importantly, minimize the dose of these agents to the lowest dose that is reasonably achievable, particularly when rapid platelet inhibition is desired.

Acknowledgments. The investigators thank the patients who participated in this study for their critical contribution. Drs McEvoy and Schulman also thank their research coordinator, Ms Frances A. Kirkland, without whom this study would not have been possible.


1. Tantry US, Bonello L, Aradi D, et al. Consensus and update on the definition of on-treatment platelet reactivity to adenosine diphosphate associated with ischemia and bleeding. J Am Coll Cardiol. 2013;62:2261-2273.

2. Shah R, Kirtane AJ, McEvoy JW. Opiate use in the cath lab. Eur Heart J. 2018;39:642-645.

3. Lavi S, Jolly SS, Bainbridge D, Manji F, Randhawa V, Lavi R. Sedation, analgesia, and anaesthesia variability in laboratory-based cardiac procedures: an international survey. Can J Cardiol. 2014;30:627-633.

4. McCarthy CP, Mullins KV, Sidhu SS, Schulman SP, McEvoy JW. The on- and off-target effects of morphine in acute coronary syndrome: a narrative review. Am Heart J. 2016;176:114-121.

5. Hobl EL, Stimpfl T, Ebner J, et al. Morphine decreases clopidogrel concentrations and effects: a randomized, double-blind, placebo-controlled trial. J Am Coll Cardiol. 2014;63:630-635.

6. Hobl EL, Reiter B, Schoergenhofer C, et al. Morphine decreases ticagrelor concentrations but not its antiplatelet effects: a randomized trial in healthy volunteers. Eur J Clin Invest. 2016;46:7-14.

7. Hobl EL, Reiter B, Schoergenhofer C, et al. Morphine interaction with prasugrel: a double-blind, cross-over trial in healthy volunteers. Clin Res Cardiol. 2016;105:349-355.

8. Kubica J, Adamski P, Ostrowska M, et al. Morphine delays and attenuates ticagrelor exposure and action in patients with myocardial infarction: the randomized, double-blind, placebo-controlled IMPRESSION trial. Eur Heart J. 2016;37:245-252.

9. Thomas MR, Morton AC, Hossain R, et al. Morphine delays the onset of action of prasugrel in patients with prior history of ST-elevation myocardial infarction. Thromb Haemost. 2016;116:96-102.

10. Ibrahim K, Goli RR, Shah R, Resar JR, Schulman SP, McEvoy JW. Effect of intravenous fentanyl on ticagrelor absorption and platelet inhibition among patients undergoing percutaneous coronary intervention: design, rationale, and sample characteristics of the PACIFY randomized trial. Contemp Clin Trials. 2018;64:8-12.

11. Ibrahim K, Shah R, Goli RR, et al. Fentanyl delays the platelet inhibition effects of oral ticagrelor: full report of the PACIFY randomized clinical trial. Thromb Haemost. 2018;118:1409-1418.

12. McEvoy JW, Ibrahim K, Kickler TS, et al. Effect of intravenous fentanyl on ticagrelor absorption and platelet inhibition among patients undergoing percutaneous coronary intervention: the PACIFY randomized clinical trial (platelet aggregation with ticagrelor inhibition and fentanyl). Circulation. 2018;137:307-309.

From the 1Department of Medicine, Johns Hopkins University School of Medicine, Baltimore, Maryland; 2Division of Cardiology, Department of Medicine, Johns Hopkins University School of Medicine, Baltimore, Maryland; 3Department of Pathology, Johns Hopkins University School of Medicine, Baltimore, Maryland; 4Johns Hopkins Ciccarone Center for the Prevention of Heart Disease, Baltimore, Maryland; 5National University of Ireland and National Institute for Preventive Cardiology, Galway, Ireland; and the 6Division of Cardiology, Department of Medicine, Saolta University Healthcare Group, University College Hospital Galway, Galway, Ireland.

Funding: PACIFY was fully funded by an institutional grant to Dr McEvoy from The Johns Hopkins Magic That Matters Research Grant.

ClinicalTrials.gov Identifier: NCT02683707

Disclosure: The authors have completed and returned the ICMJE Form for Disclosure of Potential Conflicts of Interest. The authors report no conflicts of interest regarding the content herein.

Manuscript submitted March 13, 2019, accepted March 25, 2019.

Address for correspondence: John W. McEvoy, MBBCh, MHS, Saolta University Healthcare Group, University College Hospital Galway, Newcastle Rd, Galway, H91 YR71 Ireland. Email: johnwilliam.mcevoy@nuigalway.ie  Twitter: @johnwmcevoy