Author + information
- Received June 13, 2002
- Revision received December 16, 2002
- Accepted December 18, 2002
- Published online July 2, 2003.
- Eric D Peterson, MD, MPH*,* (, )
- Charles V Pollack Jr, MD, MA†,
- Matthew T Roe, MD, MHS*,
- Lori S Parsons, BS‡,
- Katherine A Littrell, PhD, RN§,
- John G Canto, MD, MSPH, FACC∥,
- Hal V Barron, MD§,
- National Registry of Myocardial Infarction (NRMI) 4 Investigators
- ↵*Reprint requests and correspondence:
Dr. Eric D. Peterson, Duke Clinical Research Institute, 2400 Pratt Street, Room 7009, Durham, North Carolina 27705, USA.
Objectives We sought to identify patient and hospital features associated with early glycoprotein (GP) IIb/IIIa inhibitor therapy for non–ST-elevation (NSTE) myocardial infarction (MI) and to relate this treatment to in-hospital outcomes.
Background Glycoprotein IIb/IIIa inhibitors have improved outcomes in randomized trials of NSTE MI, leading national treatment guidelines to recommend their use. Their actual use, safety, and effectiveness have not been well characterized beyond trial populations.
Methods We studied 60,770 patients with NSTE MI treated between July 2000 and July 2001 at 1,189 hospitals in a U.S. registry. Using logistic regression, we identified patient and hospital features associated with GP IIb/IIIa inhibition within 24 h after presentation. We also compared outcomes by early treatment versus no treatment after adjusting for patient and hospital characteristics and treatment propensity.
Results Only 25% of eligible patients received early GP IIb/IIIa therapy. Elderly patients, women, minority patients, and those without private insurance received such therapy less often than their counterparts. Treated patients had lower unadjusted in-hospital mortality (3.3% vs. 9.6%, p < 0.0001) remaining significantly lower after adjustment for patient risk, treatment propensity, and hospital characteristics (adjusted odds ratio, 0.88; 95% confidence interval, 0.79 to 0.97). Hospitals that adopted early GP IIb/IIIa inhibition more rapidly also had lower adjusted mortality rates than those slower to adopt such therapy.
Conclusions Glycoprotein IIb/IIIa inhibitor therapy appears to be underused in early management of NSTE MI patients. Because this therapy is associated with better outcomes, it represents a target for quality improvement.
More than one million patients are hospitalized in the U.S. each year for acute myocardial infarction (MI) (1). Most of these patients will not have ST-segment elevation on their initial electrocardiogram (ECG) and are classified as having non–ST-elevation (NSTE) MI (2). The outcomes of pa tients with NSTE MI can be improved if they are treated acutely with antithrombotic agents, including aspirin and heparin (3). Multiple randomized clinical trials have also shown that intravenous platelet glycoprotein (GP) IIb/IIIa receptor inhibitors significantly reduce the risk of death or reinfarction in patients with NSTE MI (4). These findings led the American College of Cardiology and American Heart Association (ACC/AHA) in March 2000 to give a class IA treatment recommendation for the use of GP IIb/IIIa inhibitors, in addition to aspirin and heparin, in all patients with positive cardiac markers (2).
Information concerning the use of GP IIb/IIIa inhibition outside of selected clinical trial populations is limited. Are these drugs being used in community practice in a manner consistent with guideline recommendations? If not, where are the gaps in use? Can the results from clinical trials be generalized outside of trial populations? Are these drugs safe and effective when used in an at-large patient population that includes older and higher risk patients?
Using the National Registry of Myocardial Infarction (NRMI) 4, we sought to characterize contemporary patterns of GP IIb/IIIa inhibitor use in the early management of patients with NSTE MI at 1,189 centers across the U.S. We identified patient and hospital features associated with a greater or lesser likelihood of receiving a GP IIb/IIIa inhibitor within 24 h of hospital admission. We also compared unadjusted and adjusted clinical outcomes between those receiving and not receiving early GP IIb/IIIa inhibitor therapy. Finally, we assessed whether hospitals that had more completely adopted the ACC/AHA guideline recommendations for GP IIb/IIIa inhibitor use had better risk-adjusted outcomes than did those who were slower to adopt this practice.
The NRMI is a voluntary, observational study of hospitalized patients with a confirmed diagnosis of acute MI. The fourth generation of data-collection, NRMI 4, began on July 1, 2000, and currently compiles data from one in four acute-care hospitals in the U.S. Hospitals participating in NRMI tend to be slightly larger, and have more cardiac interventional facilities, than nonparticipating hospitals. Characteristics of the NRMI data-collection procedures have been described in detail (5,6). Briefly, participating hospitals are instructed to enroll consecutive MI patients. Hospitals typically screen for MI based on discharge diagnosis code of 410.x1 according to the ninth revision of International Classification of Disease (7). The clinical diagnosis of MI is then confirmed based on one or more of the following: 1) total creatine kinase (CK) or CK-MB level at least twice the upper limit of normal; 2) ECG evidence of acute MI; 3) enzymatic, scintigraphic, or other evidence of MI. All data are abstracted via chart review at the hospital by a trained clinical data abstractor using standardized definitions.
We included patients enrolled between July 1, 2000, and July 31, 2001. From this population, we excluded patients with ST-segment elevation or new bundle-branch block on the baseline ECG, those treated with fibrinolytic therapy, and those who had normal levels of serum cardiac markers. We also excluded those who were transferred out of an NRMI hospital (incomplete follow-up) and those who were transferred to an NRMI hospital >24 h after MI onset (nonacute patients). Finally, we excluded those with potential contraindications for intravenous GP IIb/IIIa inhibitors, including active bleeding, recent surgery, malignancy, uncontrolled hypertension, traumatic cardiopulmonary resuscitation, prior stroke, creatinine > 2.5 mg/dl, or known bleeding disorder, and those for whom the use or timing of GP IIb/IIIa inhibition was unknown.
Study variables and definitions
The NRMI 4 case report form contains demographic and clinical information, including patient age, race, gender, insurance status, medical history, clinical presentation, therapies, and in-hospital outcomes. The completeness and accuracy of patient enrollment in the NRMI database and accuracy of clinical risk factors and in-hospital events has been verified against independent chart abstraction (8).
Outcomes within NRMI are limited to in-hospital events as classified by NRMI investigators, without central adjudication. All-cause mortality was assessed in this analysis. Stroke was defined as a permanent neurological deficit and further classified as hemorrhagic, thromboembolic, thromboembolic with hemorrhagic conversion, or other (subdural hematoma). Major bleeding was defined as bleeding requiring intervention, including hemodynamic instability, transfusion, or surgical intervention. Reinfarction was defined as a second MI event confirmed by new diagnostic ECG changes or (re)elevation of cardiac enzymes.
Our primary goal was to examine the extent to which community use of GP IIb/IIIa inhibitors was in accord with the ACC/AHA recommendations. A secondary goal was to relate these observational care patterns with patient outcomes. During the study period, the ACC/AHA recommendations specified use of GP IIb/IIIa inhibitors for “high-risk” patients (including all those with positive cardiac markers) and in those for whom percutaneous coronary intervention (PCI) was planned (2). Because NRMI 4 does not collect information on “planned PCI,” we defined “early treatment” as administration of intravenous GP IIb/IIIa inhibitors within 24 h of hospital arrival. We compared patients who received early treatment with those not receiving early treatment (including those never treated and those treated only after 24 h, such as after elective PCI). This treatment comparison strategy was conservative, as it favorably bias the no early treatment group by including those who survived to undergo late PCI (and received concomitant GP IIb/IIIa within the no early treatment group).
The associations between baseline characteristics, procedure use, and clinical outcomes among patients who did and did not receive GP IIb/IIIa inhibitors within 24 h of admission were examined using the chi-square statistic for categorical variables and analysis of variance for continuous variables (Table 1). A treatment-propensity analysis was performed to identify multivariable predictors of early GP IIb/IIIa inhibition using stepwise logistic regression analysis (Table 2).
Several analytical techniques were used to adjust patient outcome comparisons for potential treatment selection biases. First, we adjusted our analyses for 13 baseline clinical risk factors using a validated NSTE MI mortality model (Appendix) (9), C-index value in our study population (0.75). We then sequentially adjusted our comparisons of mortality by treatment for treatment propensity score (described in the preceding text) and hospital features (including geographic region, number of beds, on-site cardiac catheterization facilities, and academic affiliation). To test the robustness of these findings, we repeated the analyses: 1) among the subgroup of patients admitted to hospitals with on-site PCI facilities (where patient transfer was limited); 2) after excluding those receiving PCI within the first 24 h; 3) after excluding all PCI patients; 4) after excluding all patients receiving cardiac catheterization; and 5) after excluding all patients who died within the first 24 h of admission.
We also performed a “matched pair” propensity analyses (10,11), in which we paired patients with similar baseline likelihood for treatment where one received early GP IIb/IIIa treatment and one did not. After matching, analyses of baseline characteristics were repeated to ensure that that the populations were comparable. Conditional logistic regression was then used to determine the effect of early GP IIb/IIIa inhibition on mortality in this matched-pair sample.
Finally, the association between early GP IIb/IIIa inhibition and mortality was assessed at the hospital level. For this analysis, institutions contributing fewer than 10 patients per year (18% of the hospitals) were excluded to ensure stable center estimates. Hospitals were subdivided into quartiles based on the proportion of their patients with NSTE MI who received GP IIb/IIIa inhibitors early. Unadjusted and risk-adjusted mortality rates then were compared among hospital subgroups. All statistical analyses were performed using SAS software (version 8.0, SAS Institute, Cary, North Carolina).
The NRMI database contained 110,590 NSTE MI patients with acute MI treated at 1,189 U.S. hospitals from July 2000 to July 2001. From this, we excluded 20,594 patients who were transferred out (19%), 5,802 patients who were transferred in more than 24 h after hospitalization (5%), 23,038 patients with a potential contraindication for GP IIb/IIIa inhibition (21%), and 386 patients for whom the time when GP IIb/IIIa inhibition began was missing (<1%). This left a sample of 60,770 patients with NSTE MI who were eligible for early GP IIb/IIIa inhibitor therapy. Of these, 15,379 patients (25%) received such therapy within 24 h of admission. The median time to administration of a GP IIb/IIIa agent was 6 h. The average duration of infusion was 26.3 ± 18 h. Among the 45,391 (75%) not receiving early GP IIb/IIIa inhibition, 4,372 (10%) received a GP IIb/IIIa blocker at some point >24 h after admission, most commonly with elective PCI.
Baseline characteristics and predictors of early treatment
Patients receiving early GP IIb/IIIa inhibitor treatment were significantly younger and were more likely to be white and male and have private insurance (instead of Medicare, Medicaid, or self-pay) (Table 1). Treated patients were less likely to have had a prior MI or most comorbid illnesses, but were more likely to have had a coronary revascularization procedure. Patients who received early GP IIb/IIIa inhibitor therapy were more likely to present with chest pain or ST-segment depression, and were more likely to be admitted with a presumed diagnosis of acute MI, compared with patients not treated, but the time from symptom onset to hospitalization was similar between groups. Treated patients were also more prevalent among larger hospitals, hospitals with catheterization laboratories and bypass surgery facilities, and teaching hospitals.
Table 2displays selected independent predictors of treatment in our multivariable propensity model. In total, 23 factors were independently associated with the propensity for GP IIb/IIIa inhibitor treatment, with a model C-index of 0.78. The factor most strongly associated with GP IIb/IIIa inhibition was chest pain on presentation, followed by the hospital’s capacity to perform PCI.
Compared with those not treated with early GP IIb/IIIa inhibition, treated patients were significantly more likely to have a cardiology consultation (96% vs. 79%, p < 0.0001) and to undergo in-hospital PCI (45% vs. 15%, p < 0.0001). In contrast, the use of coronary bypass surgery was similar between treated and untreated patients (11% for each). Within 24 h of hospital admission, patients who received early GP IIb/IIIa blockade were also more likely to receive aspirin (91% vs. 80%, p < 0.004) and beta-blockers (75% vs. 60%, p < 0.001).
The overall in-hospital mortality rate was 8.0%. Patients treated early with GP IIb/IIIa inhibitors had significantly lower unadjusted mortality than did patients not receiving such therapy (3.3% vs. 9.6%, p < 0.001). Likewise, the rates of death or (re)MI were lower in those receiving early therapy versus not (4.5% vs. 10.3%, p < 0.001). Unadjusted rates for nonfatal complications of MI (shock, cardiac arrest, reinfarction) were similar or higher in those receiving a GP IIb/IIIa inhibitor versus not. In terms of safety, patients receiving a GP IIb/IIIa inhibitor had a higher rate of major bleeding (10% vs. 9.5%, p = 0.04), but a lower risk for stroke (0.7% vs. 1.2%, p < 0.001) and a similar hemorrhagic stroke rate (0.1% vs. 0.1%, p = 0.38) (Table 3).
Figure 1shows the in-hospital mortality rate for patients treated versus not with an early GP IIb/IIIa inhibitor stratified by NSTE MI risk score. Among all risk strata, in-hospital mortality rates were lower in patients treated with a GP IIb/IIIa inhibitor than in those not so treated. In particular, the absolute treatment differences tended to be widest among those with intermediate to high baseline risk.
Table 4provides the results from our matched-pair propensity analysis. After matching on treatment propensity, baseline characteristics and clinical features were similar among those treated with versus not. This matched sample had a lower risk profile resulting in a lower overall mortality rate than the overall study population. Despite lower risk in this matched subset, early GP IIb/IIIa inhibitor therapy remained associated with lower in-hospital mortality compared with those not treated (3.5% vs. 3.9%, p < 0.03).
Table 5presents the mortality comparisons after sequentially adjusting for baseline patient risk (NSTE MI risk score), treatment propensity, and hospital factors. Adjustment for patient risk and treatment propensity reduced the association between early GP IIb/IIIa inhibition and in-hospital mortality. Further adjustment for provider factors (hospital location, facilities, and academic affiliation) and the use of acute therapies (aspirin and beta-blocker use within 24 h) had limited impact on these conclusions. After full adjustment, patients receiving a GP IIb/IIIa inhibitor early remained significantly less likely to die compared with those not treated (adjusted odds ratio [OR], 0.88; 95% confidence interval [CI], 0.79 to 0.97).
This risk-adjusted analysis was repeated among important patient subsets. The mortality reductions associated with early GP IIb/IIIa use were similar magnitude in men and women after adjusting for baseline risk and treatment propensity, mortality OR in males 0.85 (0.73 to 0.97) versus mortality OR in females 0.88 (0.74 to 1.01). Among NRMI hospitals with on-site PCI facilities, the use of early GP IIb/IIIa inhibitor therapy was significantly associated with lower patient mortality (OR, 0.82 [95% CI, 0.73 to 0.92]). Similar survival benefits of early GP IIb/IIIa inhibition also were seen after excluding those receiving PCI within 24 h (OR, 0.82 [95% CI, 0.73 to 0.93]) and after excluding all patients receiving PCI at any time (OR, 0.85 [95% CI, 0.76 to 0.95]) or after excluding all patients receiving cardiac catheterization (OR, 0.68 [95% CI, 0.60 to 0.78]). Excluding all patients who died within the first 24 h of hospitalization reduced the overall mortality rate by 14%, yet patients receiving early GP IIb/IIIa treatment still tended to have a lower in-hospital mortality risk than those not treated (OR, 0.91 [95% CI, 0.83 to 1.01].
Figure 2displays the association between hospitals’ use of early GP IIb/IIIa inhibitors in patients with NSTE MI and their overall unadjusted and risk-adjusted mortality rates. Centers with the highest rate use of early GP IIb/IIIa (used in 30% to 85% of patients) had nearly 50% lower unadjusted and risk-adjusted mortality rates compared with the lowest quartile of use. This hospital-level relationship likely results from multiple factors including other processes of care and structural features beyond purely the effects of GP IIb/IIIa treatment. Regardless, adoption of this newer NSTE MI care process appears to be good marker for overall hospital outcomes and quality.
This study represents the first nationwide assessment of the contemporary use of GP IIb/IIIa inhibitors in patients with NSTE MI. Despite published randomized trials and national guideline recommendations, we found that three of four eligible NSTE MI patients failed to receive GP IIb/IIIa inhibitor therapy within 24 h of admission. When studied in a higher-risk, community-based population, the early use of GP IIb/IIIa inhibition appears generally safe and was associated with a reduction in mortality risk similar to that seen in randomized studies.
Randomized clinical trials have evaluated the efficacy of GP IIb/IIIa inhibitors in more than 30,000 patients with acute coronary syndromes (4,12). The trial populations, however, were younger and had overall lower risk than the patients in our community-based NRMI cohort. For example, the mean age of patients enrolled in the three largest trials of these agents (13–15)ranged from 61 to 64 years (vs. 70 years in NRMI). These and other differences led to significantly higher in-hospital mortality rate in our study patients relative to trial populations (8% in NRMI vs. 1.5% to 3% in trials).
Areas for improvement in performance
While we found overall low use of early GP IIb/IIIa inhibitor therapy, this was more profound in certain groups. Older patients, women, minorities, and those without private insurance all were significantly less likely to receive early GP IIb/IIIa inhibition than were their counterparts. Disparity in the use of evidence-based care by age, gender, and socioeconomic status is disturbing, yet it is consistently seen in many areas of healthcare (16–19). Patients with higher baseline risk were also paradoxically less likely to receive early GP IIb/IIIa inhibitor treatment. For example, whereas 45% of low-risk patients (expected risk < 2%) received early GP IIb/IIIa inhibition, only 9% of the high-risk patients (expected mortality > 15%) received treatment. This pattern of conservative care in high-risk patients is also seen in other areas of medicine (20,21), but is inconsistent with maximizing absolute benefits from therapy (Fig. 1). It is, therefore, incumbent upon physicians to use available risk assessment tools to guide care decisions (22–24).
Finally, patients were more likely to receive a GP IIb/IIIa inhibitor if they were treated at larger centers, centers with academic affiliations or with interventional capability, or if cared for by cardiologists. Combined, these patterns appear to indicate that randomized studies of GP IIb/IIIa inhibition and the resultant ACC/AHA treatment recommendations have yet to diffuse to all healthcare providers and specialties.
Safety and effectiveness of GP IIb/IIIa inhibition
While being observational in nature, this study also provides insight into the safety, effectiveness, and appropriate timing of GP IIb/IIIa inhibition in “real-world” patients. Although the NRMI cohort was older and higher risk than prior trial populations, we found a <1% absolute increase in the risk of major bleeding and no increased risk of stroke or intracranial hemorrhage associated with GP IIb/IIIa inhibition.
With regard to effectiveness, early GP IIb/IIIa therapy was associated with significantly lower in-hospital mortality rates, an effect that persisted after adjustment for patient and provider factors and treatment selection (Table 5). The plausibility of this finding is increased by its robustness. The estimated benefit of GP IIb/IIIa inhibition was constant across an array of risk-adjustment techniques and patient subgroups. The magnitude of mortality benefit also was remarkably consistent with those estimated from a meta-analysis of the major GP IIb/IIIa inhibitor trials (4). In an overview of the major GP IIb/IIIa inhibitor trials, the 30-day mortality risk reduction with treatment was 0.87 (95% CI, 0.74 to 1.02) (25)versus the current study estimate of 0.88 (95% CI, 0.79 to 0.97). Our larger patient population and higher event rates, however, provided greater statistical power to detect a significant association between early treatment and lower mortality.
The interrelation between the benefits of GP IIb/IIIa therapy and interventional strategies deserve further comment. In our study, patients receiving early GP IIb/IIIa therapy were significantly more likely to receive cardiac catheterization and PCI than those not, raising the possibility that the benefits demonstrated were related to the combination of early GP IIb/IIIa therapy and PCI. In fact, this combined aggressive, interventional strategy is supported by the recent Treat Angina with Aggrastat and Determine Cost of Therapy with an Invasive or Conservative Strategy (TACTICS-TIMI 18) randomized trial and has been incorporated into subsequent ACC/AHA guidelines recommendations.
A related question is whether the benefits from early GP IIb/IIIa inhibitor therapy are limited only to those receiving an interventional approach. This possibility was raised by the negative results of the Effect of Glycoprotein IIb/IIIa Receptor Blocker Abciximab on Outcome in Patients with Acute Coronary Syndromes without Early Coronary Revascularization (GUSTO IV ACS) when a GP IIb/IIIa was tested in a predominately conservatively treated ACS population (13). However, a subgroup analysis from the six major GP IIb/IIIa trials (including GUSTO IV ACS) reported that both PCI patients and non-PCI patients had trends towards mortality benefits with GP IIb/IIIa therapy OR 0.83 (0.53 to 1.29) and 0.91 (0.81 to 1.04), respectively (12). Our current observational analysis was consistent with this trial overview finding that the benefits of early GP IIb/IIIa therapy persisted after excluding those who had PCI or cardiac catheterization. While intriguing, further study of the potential benefits of GP IIb/IIIa therapy among medically treated patients appears warranted.
With regard to nonfatal events, our study was not able to replicate trial findings of a reduction in reinfarction with GP IIb/IIIa therapy. There are several potential reasons for this. Most important, our NSTE MI population could not assess the benefits of prevention of infarction in those with unstable angina, only assessing “re-infarction.” This end point is difficult to measure in registries due to inconsistency in drawing sequential cardiac markers and the lack of central event adjudication. As a result, the reported rate of reinfarction in NRMI’s higher risk patients was less than half the rate in randomized trials.
This study was based on a voluntary registry of MI care at 1,000+ U.S. centers. Although this represents about 25% of U.S. hospitals, participating hospitals tend to be larger and have more interventional facilities compared with non-NRMI hospitals. Thus, our reported results likely underestimate the current gap between clinical practice and guideline-recommended treatment. Contraindications to GP IIb/IIIa agents were based on chart review, which may lead to under-coding. However, we doubt that this can fully explain the three-fourths of patients not receiving treatment.
Our treatment comparisons were based on observational data and are subject to the limitations of all nonrandomized analyses. Although we applied multiple techniques to adjust for potential patient and provider differences, we acknowledge that unmeasured treatment-selection biases could persist. Issues of early mortality may further complicate treatment comparison. However, the relative benefits of GP IIb/IIIa therapy tended to persist after conservatively excluding patients who died within 24 h. Finally, as discussed above, outcome differences both at the patient and provider levels (Fig. 2) may have resulted from the effects of care other than GP IIb/IIIa inhibitor therapy. Although the benefits of GP IIb/IIIa inhibition persisted after adjustment for other acute therapies (acute PCI, aspirin, and beta-blockers), fully adjusting for downstream management differences is challenging due to survivor bias issues.
This study shows that the safety and efficacy of early GP IIb/IIIa inhibition in patients with NSTE MI appear to be generalizable to a national sample. Importantly, despite ACC/AHA care guidelines, we found these agents were underused, particularly in high-risk patient subgroups. We hope that ongoing provider feedback and quality-improvement programs will help hasten the dissemination and adoption of guideline recommendations into clinical practice (6,26–28).
☆ The National Registry of Myocardial Infarction is funded by Genentech, Inc., and Drs. Littrell and Barron are employees of Genentech. This analysis was funded by an unrestricted grant from Genentech, Inc.
- American College of Cardiology
- angiotensin-converting enzyme
- American Heart Association
- creatine kinase
- GUSTO IV ACS
- Effect of Glycoprotein IIb/IIIa Receptor Blocker Abciximab on Outcome in Patients with Acute Coronary Syndromes without Early Coronary Revascularization
- myocardial infarction
- National Registry of Myocardial Infarction
- odds ratio
- percutaneous coronary intervention
- Received June 13, 2002.
- Revision received December 16, 2002.
- Accepted December 18, 2002.
- American College of Cardiology Foundation
- American Heart Association
- Braunwald E.,
- Antman E.M.,
- Beasley J.W.,
- et al.
- Kong D.F.,
- Califf R.M.,
- Miller D.P.,
- et al.
- Rogers W.J.,
- Bowlby L.J.,
- Chandra N.C.,
- et al.
- Rogers W.J.,
- Canto J.G.,
- Lambrew C.T.,
- et al.
- World Health Organization
- Every N.R.,
- Frederick P.D.,
- Robinson M.,
- et al.
- Dong W.,
- Cheeks M.,
- Littrell K.,
- et al.
- Batchelor W.B.,
- Peterson E.D.,
- Mark D.B.,
- et al.
- Boersma E.,
- Pieper K.S.,
- Steyerberg E.W.,
- et al.
- for the TACTICS-TIMI 18 Investigators,
- Cannon C.P.,
- Weintraub W.S.,
- Demopoulos L.A.,
- et al.
- Kong D.F.,
- Califf R.M.,
- Miller D.P.,
- et al.
- Eagle KA, Gallogly M, Mehta RH, et al. Taking the national guideline for care of acute myocardial infarction to the bedside: developing the guideline applied in practice (GAP) initiative in southeast Michigan. J Comm J Qual Improv 2002;28:5–19