on 20-Jul-2022 (Wed)

d-separation Implies Association is Causation

Given that we have tools to measure association, how can we isolate causation? In other words, how can we ensure that the association we measure is causation, say, for measuring the causal effect of 𝑋 on 𝑌 ? Well, we can do that by ensuring that there is no non-causal association flowing between 𝑋 and 𝑌 . This is true if 𝑋 and 𝑌 are d-separated in the augmented graph where we remove outgoing edges from 𝑋 . This is because all of 𝑋 ’s causal effect on 𝑌 would flow through it’s outgoing edges, so once those are removed, the only association that remains is purely non-causal association

The CLV models use different strategies for customer behaviour modelling. One of the most reliable ones is using the recency (R), frequency (F), and monetary value (M) variables, called RFM [3], [4], [5]. These variables present some understanding of customer’s behaviour and try to answer the following questions: “How recently did the customer purchase?”, “How often do they purchase?”, and “How much do they spend?” [2]. RFM variables are sufficient statistics for customer behaviour modelling and are a mainstay of the industry because of their ease of implementation in practice [6], [3].

contributions include predictive models and techniques for customer targeting and reactivation timing (Gönül & ter Hofstede, 2006; Simester, Sun, & Tsitsiklis, 2006; Holtrop & Wieringa, 2020), market response models for firm- and/or customer-initiated marketing actions (e.g., Hanssens, Parsons, & Schultz (2003), Blattberg, Kim, & Neslin (2008), Sarkar & De Bruyn (2021)), methods for churn prediction and prevention (e.g., Ascarza (2018), Ascarza, Iyengar, & Schleicher (2016), Lemmens & Gupta (2020)), as well as a growing literature on customer valuation (e.g., McCarthy, Fader, & Hardie (2017), McCarthy & Fader (2018)) and customer prioritizing (Homburg, Droll, & Totzek, 2008). However, none of these qualify as a (Swiss Army knife-like) general-purpose problem solver that generalizes across the described decision tasks of managing customer relationships. This article makes a first step towards this direction. We propose and implement a flexible methodological framework that provides marketing managers with highly accurate forecasts of fine granularity both in the short and in the long run. Our method also captures seasonal peaks and customer-level dynamics and allows to differentiate between different customer groups

A schematic high-level representation of the proposed model architecture is shown in Fig. 2. The structure of the model begins with its input layers for (i) the input variable (i.e., transaction counts) and (ii) optional covariates (time-invariant or time-varying inputs). These variable inputs enter the model through dedicated input layers at the top of the model’s architecture and are combined by simply concatenating them into a single long vector. This input signal then propagates through a series of intermediate layers including a specialized LSTM, or Long Short-Term Memory RNN neural network component.

We demonstrate how firms operating in non-contractual business settings can benefit from the automatic feature extraction capabilities of deep learning models for predictive customer base analysis. Our proposed model informs managers on both short- and long-term forecasts of individual customer behavior and helps to timely uncover business opportunities as well as potential customer defection. As we have shown, it also accurately predicts periods of elevated transaction activity and captures other forms of purchase dynamics that can be leveraged in simulations of future sequences of customer transactions. We highlight our model’s flexibility and performance on two groups of valuable customers: those who keep making more and more transactions with the firm (denoted as ”opportunity” customers) and those who are at risk of defection. We demonstrate that the model also excels at automatically capturing seasonal trends in customer activity, such as the shopping period leading up to the December holidays. In Appendix Section F we provide a further characterization of scenarios where our model performs particularly well and where it does not do so relative to the used benchmark methods. The model brings many practical benefits for the marketing analyst, such as the lack of need for manual encoding of any features in the customer data, a simple optimization objective, and quick estimation on modern computer hardware. We show that incorporating contextual information in the model is straightforward and brings an additional boost in predictive accuracy. However, the model performance is already extremely strong when no context is available beyond the timing of the customer’s transactions. This is welcome news for firms that do not wish to collect personal information on principle, to avoid the questionable ethics of harvesting the ‘‘behavioral surplus” (Zuboff, 2019): our work shows that this is feasible without a big loss of accuracy. We gather evidence from eight diverse real-life settings to demonstrate the model robustness as a flexible, general purpose prediction tool for customer base analysis

B.2. Network training

When training neural networks, the norm is to monitor the validation loss – a measure of the error the model makes on an unseen part of the data – to assess the progress of model training. The idea is that good performance on unseen data is evidence of the model’s ‘‘general” ability to perform a given task, and by ending the training process at the point when the validation loss stops improving, we prevent the model from overfitting the training data. There is a downside though: when we take out a (randomly sampled) portion of the calibration data to form this validation set, we end up with less data to learn from overall (typically around 10% of the data is used for validation). This trade-off makes sense in scenarios where we do not know which data will be used as model input in the future. This is not our case though: we make predictions for a specific cohort of customers. This means our training procedure is as follows: we train the model as is common using a validation set first, and once the validation loss stops improving for a number of epochs, we restore the model state to the point with the lowest validation loss and perform several ‘‘fine-tuning” training epochs using the entire calibration data set including the samples previously left out as validation, using a large batch size and a reduced learning rate. The idea is to fine-tune the model to the specific cohort, even if this technically means a small degree of ‘‘overfitting”. Note that at no point is any of the holdout period data used during the fine-tuning stage. This way, we also assure a like-for-like comparison of the LSTM with probability models, which also use the entire calibration set for model estimation but do not require a separate validation set.

none of these qualify as a (Swiss Army knife-like) general-purpose problem solver that generalizes across the described decision tasks of managing customer relationships. This article makes a first step towards this direction. We propose and implement a flexible methodological framework that provides marketing managers with highly accurate forecasts of fine granularity both in the short and in the long run. Our method also captures seasonal peaks and customer-level dynamics and allows to differentiate between different customer groups

The CLV models use different strategies for customer behaviour modelling. One of the most reliable ones is using the recency (R), frequency (F), and monetary value (M) variables, called RFM [3], [4], [5]

We demonstrate how firms operating in non-contractual business settings can benefit from the automatic feature extraction capabilities of deep learning models for predictive customer base analysis. Our proposed model informs managers on both short- and long-term forecasts of individual customer behavior and helps to timely uncover business opportunities as well as potential customer defection.

Other [ than ‘‘Buy ’Till You Die” (BTYD) ] options to capture changes between lower- and higher-frequency purchase episodes (as we observe for our customers in Fig. 1), or vice versa, are to adopt a dynamic changepoint model (Fader, Hardie, & Chun-Yao, 2004), a simulation based model of the type presented by Rust, Kumar, and Venkatesan (2011), or to incorporate additional states other than the absorbing, inactive state as in standard BTYD latent attrition models.

To forecast future customer behavior, our model is trained using individual sequences of past transaction events, i.e., chronological accounts of a customer’s lifetime. The example in Table 2 describes one such customer’s transaction history over seven consecutive discrete time periods

This particular individual makes a transaction in the first week, followed by one week of inactivity, then transacting for two consecutive weeks, and so on; in weeks 3 and 4 they also received some form of a marketing appeal.

In one series of 21 patients with elevated cTnI levels and normal coronary angiograms, tachycardia was determined to be the explanation of the troponin elevation in six patients [14]

These reports illustrate that myocardial troponin can be released as a consequence of tachycardia alone in the absence of myodepressive factors, inflammatory mediators, and CHD.

Left ventricular hypertrophy — cTn elevation has been described in the context of left ventricular hypertrophy (LVH). In a series of 74 consecutive patients without clinical evidence of active myocardial ischemia referred for routine echocardiography, 7 of 25 patients in the tertile with the greatest LV mass had an elevated cTnI. In contrast, one patient in the intermediate range, and none of the patients in the lowest tertile had an elevated troponin level [49].

It is well recognized that LVH can lead to occult subendocardial ischemia via increased oxygen demand from increased muscle mass, coupled with decreased flow reserve due to remodeled coronary microcirculation (see 'Demand ischemia' above).

Coronary vasospasm — Myocardial ischemia caused by coronary vasospasm (Prinzmetal angina) can lead to troponin elevations. This was illustrated in a series of 93 patients with suspected myocardial ischemia in whom coronary angiography revealed no hemodynamically significant lesions [51]. Twenty-three (25 percent) had elevated levels of cTnI. Ergonovine provocation testing showed evidence of coronary vasospasm in 41 patients, and in 17 of the 23 patients with cTnI elevations.

Acute stroke — Both elevated cTn levels and ischemic electrocardiographic changes have been described in the setting of acute stroke or intracranial hemorrhage. In one series of 149 patients with symptoms of acute stroke, 27 percent were found to have elevated serum cTnI [52].

The most likely explanation of troponin elevation and myocardial damage in this setting is an imbalance of the autonomic nervous system, with resulting excess of sympathetic activity and increased catecholamine effect on the myocardial cells [53,55].

The magnitude of troponin elevation in these reports is often less than that seen with acute MI due to coronary artery occlusion. Thus, it is not clear to what extent the LV dysfunction and hemodynamic compromise reported in these case series is due to acute myocardial injury in the setting of the stroke, or reflects new myocardial and hemodynamic stress in patients with underlying cardiovascular disease.

Atrial fibrillation — Elevated concentrations of cTn, in particular when measured with high-sensitivity (hs) assays, have been described in patients with atrial fibrillation in the absence of clinically overt heart failure or demand myocardial ischemia [56-58]. These studies have found that higher concentrations of troponin I or T are independently associated with a higher risk of stroke or systemic embolism, as well as with the risk of MI and cardiac mortality and are substantially more predictive than the CHADS₂ and CHA₂DS₂-VASc scores [56,57].

Direct myocardial injury — Troponin elevation can occur in the setting of myocardial injury by traumatic or inflammatory processes

Thus, a degree of direct cardiac injury, as evidenced by cTnI elevations, is common after blunt chest trauma, although only a minority of patients with troponin elevations in this setting develop significant clinical complications attributable to cardiac injury.

High-dose chemotherapy; troponin levels have been suggested as a method for detecting cardiotoxicity and predicting the development of future LV dysfunction in this population [62].

Heart failure — Troponin elevations tend to be associated with advanced heart failure and an adverse prognosis [66]. The clinical evidence supporting this association is presented separately. (See "Predictors of survival in heart failure with reduced ejection fraction", section on 'Troponins'.)

Heart failure can lead to the release of cTn principally via two related mechanisms: myocardial strain and myocyte death.

Pulmonary disease — Troponin elevations are also described in a variety of pulmonary diseases, usually associated with significant right heart strain.

Serum troponins are elevated in 30 to 50 percent of patients with a moderate to large pulmonary embolism. The presumed mechanism is acute right heart overload, and elevated levels are associated with a significant increase in mortality. The troponin elevations usually resolve within approximately two days after pulmonary embolism in contrast to the more prolonged elevation with acute myocardial infarction.

Chronic kidney disease — Persistent elevation of cTn is frequently observed among patients with end-stage kidney disease. This issue is discussed in detail separately. (See "Cardiac troponins in patients with kidney disease", section on 'Effect of CKD on troponin levels'.)

The cTnT elevations seen in patients with renal failure may be due to structural myocardial abnormalities and are invariably associated with pathological evidence of myocardial injury. Data also suggest an important role for renal clearance when values are low despite the fact that cTn is not detected in the urine [73].

Burns — Severe thermal injury is associated with cardiac contractile dysfunction and elevated cardiac troponin. Elevation of cTn is demonstrated among patients who have sustained >25 percent (total body surface area) of thermal injury. The rise in cTn appears to be related to the extent of burns rather than patient's age, pre-existing medical conditions, or the administration of resuscitation fluid [76].

While these findings suggest that measuring cTn may be useful to screen seemingly healthy individuals for CVD risk, we do not recommend this for routine practice at present for several reasons. The first is that the values that are prognostic overlap very substantially with individuals who are not at risk. In addition, at these levels, both analytical and biological variation are substantial so that values may change a few ng/L when repeated, moving from a predictive value to one that is not. This has been elegantly shown in chest pain patients as well [81]

The primary goal of STEMI management is to reduce the risk of death and the extent of permanent cardiac injury associated with MI. Because therapy for patients with STEMI becomes less effective with each minute its delivery is delayed (figure 1), another goal of therapy is to rapidly treat patients with STEMI before treatment becomes ineffective.

The rapid diagnosis of STEMI only requires the presence of symptoms suspicious for an ACS (eg, chest discomfort, dyspnea, sudden death) and a confirmatory ECG; it does not require evidence of elevated cardiac biomarkers such as troponin.

Thus, patients with suspected ACS should undergo a focused history, physical, and ECG within ten minutes of hospital arrival to identify the key findings of STEMI ( table 1):

ECGs should be reviewed for signs of severe myocardial ischemia, which include:

• ST-segment elevation with standard lead placement (waveform 1 and waveform 2)

• Newly identified left bundle branch block (waveform 3)

• ST elevation with posterior (waveform 4) or right-sided (waveform 5) lead placement

• Other high-risk ECG findings (eg, de Winter sign, transient ST-segment elevation)

The following signs and symptoms suggest the presence of STEMI:

• Chest pain or chest discomfort

• Dyspnea

• Ventricular arrhythmias, cardiac arrest, or syncope

• Atypical symptoms such as malaise, weakness, and back pain

Shock – Patients with STEMI should be assessed for evidence of shock and, if present, for signs or symptoms (eg, cool extremities, jugular venous distension) that help to characterize the type of shock (ie, cardiogenic, distributive). Patients with shock require specific management of shock as well as appropriate and timely reperfusion. The clinical manifestations and causes of shock, including the mechanical complications of STEMI, are discussed separately. (See "Clinical manifestations and diagnosis of cardiogenic shock in acute myocardial infarction" and "Evaluation of and initial approach to the adult patient with undifferentiated hypotension and shock".)

Heart failure – All patients with STEMI should be assessed for signs and symptoms of heart failure (HF; eg, orthopnea, jugular venous distension, pulmonary edema). The causes, clinical manifestations, and treatment of acutely decompensated HF are covered separately. (See "Approach to diagnosis and evaluation of acute decompensated heart failure in adults".)

Coagulopathy and/or thrombocytopenia – The management of patients with STEMI typically requires treatments that increase the risk of bleeding. Thus, all patients with STEMI should be assessed for chronic use of anticoagulant or antiplatelet medications, history of bleeding or coagulation disorders (eg, uremia, heparin-induced thrombocytopenia), and the presence of abnormal coagulation studies or thrombocytopenia.

The initial management of patients with STEMI requires rapid selection and administration of reperfusion therapy. Patients with STEMI should also receive treatments that prevent further coronary artery thrombosis, minimize myocardial injury, and treat the symptoms of MI.

For patients diagnosed with STEMI, the primary goal of acute management is to rapidly restore blood flow to the acutely occluded coronary artery (ie, culprit artery) with a reperfusion therapy (ie, percutaneous coronary intervention [PCI], fibrinolysis) (algorithm 1). Thus, a reperfusion strategy should be chosen within minutes of arrival

Key factors that influence the choice of reperfusion strategy include:

● Time delay between first medical contact and performance of PCI

● Time from symptom onset to presentation

● Presence of cardiogenic shock

● Diagnostic uncertainty

● Contraindications to PCI or fibrinolysis

● High-risk factors that favor no reperfusion

Regardless of the chosen reperfusion strategy, the following therapies are routinely given to patients with STEMI to slow the progression of coronary artery thrombus formation, minimize the extent of myocardial injury, and treat symptoms (table 1)

Aspirin – All patients with STEMI should receive aspirin as soon as possible. Further details on the use of aspirin in patients with STEMI, including patients with aspirin allergy, can be found elsewhere. (See "Acute ST-elevation myocardial infarction: Antiplatelet therapy", section on 'Aspirin for all patients'.)

Nitrates – In patients with STEMI, nitrates can reduce the symptoms of chest discomfort and HF as well as treat hypertension. However, nitrates can occasionally produce profound hypotension in patients with right ventricular infarction, aortic stenosis, or who recently used sildenafil.

Beta blockers – Patients with STEMI who do not have shock, HF, bradycardia, or heart block typically receive an oral beta blocker as part of the initial therapy for STEMI. The type of agent, dosing, and evidence for use of beta blockers in ACS are discussed separately. (See "Acute myocardial infarction: Role of beta blocker therapy", section on 'Choice of drug and route of administration'.)

Anticoagulation and additional antiplatelet agents – Most patients with STEMI receive treatment with an anticoagulant and a P2Y₁₂ inhibitor. However, the approach to the use and selection of these agents is determined by the reperfusion strategy and other patient characteristics. Anticoagulant and P2Y₁₂ inhibitor therapy for each reperfusion strategy are reviewed elsewhere in this topic. (See 'Subsequent management by reperfusion strategy' below.)

Statins – In patients with STEMI who do not already take a statin, a statin is typically started soon after presentation to the hospital. The type and dose of statin are discussed separately. (See "Low density lipoprotein-cholesterol (LDL-C) lowering after an acute coronary syndrome", section on 'Inpatient management'.)

In patients with STEMI, we only use morphine to treat chest symptoms refractory to nitrates and other medical therapy; we suggest not to routinely administer morphine to all patients with STEMI. (See 'Routine medical therapy' above.)

The available observational data suggest that routine use of morphine does not reduce mortality, and limited trial data suggest that morphine may diminish the effect of P2Y₁₂ inhibitors [2-5].

In patients with STEMI who have an oxygen saturation ≥94 percent and no signs of respiratory distress, we suggest not routinely treating with supplemental oxygen. Patients with lower oxygen saturation or respiratory distress should be treated with oxygen as needed.

Blood transfusion to treat mild anemia – Red blood cell transfusion may be beneficial in select patients with STEMI. The approach to blood transfusion in patients with ACS is discussed separately. (See "Indications and hemoglobin thresholds for red blood cell transfusion in the adult", section on 'ACS (including MI)'.)

Prophylactic use of antiarrhythmics – During the early phases of acute MI, ventricular arrhythmias (eg, premature depolarizations, nonsustained ventricular tachycardia) are common. However, antiarrhythmic agents (other than beta blockers) are not typically used to prevent ventricular arrhythmias in acute MI.

Patients without obstructive, atherosclerotic coronary artery disease – Some patients with a presentation consistent with STEMI do not have obstructive atherosclerotic CAD at the time of coronary angiography. The diagnosis and management of conditions that can mimic the clinical presentation of STEMI are discussed elsewhere (table 2).

Multivessel coronary artery disease – In patients whose angiography demonstrates multivessel coronary artery disease (CAD) or left main CAD, the culprit lesions are typically treated with immediate PCI. In some patients with residual three-vessel disease or left main CAD, coronary artery bypass graft surgery (CABG) may be performed later.

Patients who undergo specialized coronary interventions (eg, rotational atherectomy) may be at higher risk for complications than patients who undergo standard PCI procedures.

Patients with failed fibrinolysis or hemodynamic instability – Patients with evidence of failed fibrinolysis or hemodynamic instability may benefit from urgent PCI.

Patients with successful fibrinolysis – Patients with STEMI who were initially treated with fibrinolysis may benefit from routine PCI (ie, pharmacoinvasive strategy) in the hours or days following successful fibrinolysis.

Monitoring for complications – Patients who do not undergo reperfusion should undergo standard post-MI evaluation. Notably, patients who are not reperfused in a timely manner are at higher risk of mechanical complications of MI (eg, myocardial wall rupture). These complications are discussed elsewhere. (See "Acute myocardial infarction: Mechanical complications".)

Medical therapy – In patients who do not undergo reperfusion, the approach to treatment with anticoagulation, antiplatelet agents may differ from patients who undergo reperfusion. The approach to medical therapy in this group of patients and the overall approach to nonacute management are discussed separately.

Unstable angina (UA) and acute non-ST elevation myocardial infarction (NSTEMI) differ primarily in whether the ischemia is severe enough to cause sufficient myocardial damage to release detectable quantities of a marker of myocardial injury (troponins):

● UA is considered to be present in patients with ischemic symptoms suggestive of an ACS and no elevation in troponins, with or without electrocardiogram changes indicative of ischemia (eg, ST segment depression or transient elevation or new T wave inversion).

● NSTEMI is considered to be present in patients having the same manifestations as those in UA, but in whom an elevation in troponins is present.

Since an elevation in troponins may not be detectable for hours after presentation, UA and NSTEMI are frequently indistinguishable at initial evaluation. As a consequence, initial management is the same for these two syndromes. For this reason, and because the pathogenic mechanisms of the two conditions are similar, they are often considered together.

Our approach to the management of elderly patients with NSTEACS is broadly similar to that in younger individuals. Adjustments to treatment may be made based on the patients' general health, weight, and renal function.

It is estimated that 60 to 65 percent of MIs occur in patients ≥65 years of age and 33 percent occur in patients ≥75 years of age [7,8]. In addition, as many as 80 percent of all deaths related to MI occur in persons ≥65 years of age.

Although patients age 75 and older have been underrepresented in clinical trials of ACS, the following observations concerning acute MI in older adults compared to younger patients are generally accepted [9]:

● Elderly patients are more likely to have an NSTEMI rather than an ST elevation MI

● Elderly patients more frequently have an atypical presentation, including silent or unrecognized MI due to presenting symptoms of syncope, weakness, or confusion (delirium) [9,10]. Delays in diagnosis have been well-documented and often lead to delays in therapy [9].

● Patients ≥75 years of age have a higher in-hospital mortality (19 versus 5 percent in one series) [11]. Both bleeding and recurrent MI are also more frequent [9]. (See 'Importance of dosing' below.)

● Elderly patients are more likely to have heart failure associated with the MI (40 versus 14 percent in one report), the risk for which increases progressively in each successive age group from 36 percent in those 65 to 69 years of age to 65 percent in those ≥85 years of age [12].

Therapies such as beta blockers, percutaneous coronary intervention, or coronary artery bypass grafting are all utilized to a lesser degree in elderly patients [11-13].

A retrospective review of almost 57,000 patients with a non-ST elevation ACS found that after adjustment, patients (including those ≥75 years of age) who received more recommended therapies had lower in-hospital mortality rates than those who did not [14]. Thus, although concern about risks and side effects is appropriate and it is not clear that adjustment accounted for all risk factors, older age alone should not be an indication to withhold recommended therapy

Cocaine-associated myocardial infarction — MI is a well-described complication among patients presenting with cocaine-induced ischemic symptoms. Beta blockers should be avoided due to the possibility of exacerbation of coronary artery vasoconstriction in the setting of acute MI in patients whose MI might have been triggered by cocaine.

Patients with unstable angina (UA) or acute non-ST elevation myocardial infarction (NSTEMI) should be treated with an early medical regimen similar to that used in an acute ST elevation MI (STEMI) with one exception: There is no evidence of benefit (and possible harm) from fibrinolysis. Initial therapy, which should be instituted within 20 minutes of presentation, is discussed in detail separately.

Drug metabolism is more likely to be reduced in elderly patients, particularly with regard to drugs that are excreted by the kidney. Dose adjustment is necessary with glycoprotein IIb/IIIa inhibitors and unfractionated or low molecular weight heparin, but not with aspirin and clopidogrel [9].

A pathophysiologic basis for the potential of harm with supplemental oxygen in patients with normoxia has been articulated [21-23]. Hyperoxia, which might occur with the administration of oxygen to normoxic individuals, has been shown to have a direct vasoconstrictor effect on the coronary arteries [21].

The potential benefit from the use of hyperbaric oxygen therapy was evaluated in a 2015 meta-analysis of six small studies with 665 participants with MI or severe angina. While a reduction in the risk of death was found (relative risk 0.58, 95% CI 0.36-0.92), methodologic limitations of the studies prevent us from having confidence in the use of such therapy [24].

Nitroglycerin — Sublingual nitroglycerin is administered to patients presenting with ischemic type chest pain, followed by intravenous nitroglycerin in patients with persistent pain after three sublingual nitroglycerin tablets, hypertension, or heart failure.

Morphine — In the setting of acute myocardial infarction, intravenous morphine should be avoided if possible and reserved for patients with an unacceptable level of pain since a large but retrospective study suggests its use is associated with an adverse effect on outcome. We give intravenous morphine sulfate at an initial dose of 2 to 4 mg, with increments of 2 to 8 mg repeated at 5- to 15-minute intervals.

In a study of 24 healthy subjects who received a loading dose of 600 mg of clopidogrel and either 5 mg of intravenous morphine or placebo, morphine significantly delayed clopidogrel resorption and reduced the area under the curve levels of its active metabolite by 52 percent [27]. Platelet inhibition, as measured by multiple tests, was less pronounced in those given morphine.

Beta blockers — Controlled trials have repeatedly documented the beneficial effects of beta blockers in patients with acute MI; however, there have been no randomized trials specifically addressing the efficacy of these drugs in non-ST elevation ACS.

Nevertheless, given the proven efficacy in unselected patients with an acute MI and the absence of harm in NSTEMI or UA, we start beta blocker therapy in all patients without contraindications within 24 hours [15]. In our view, treatment should include early use of intravenous beta blockade in patients without contraindications who have ongoing chest pain, hypertension, or tachycardia not caused by heart failure. A cardioselective agent (metoprolol or atenolol) is preferred.

However, based upon the results of the COMMIT/CCS2 trial, the largest placebo-controlled trial ever performed with beta blockers in acute MI, it seems reasonable to defer intravenous beta blockers in patients who are hemodynamically compromised in whom mortality may actually be increased by such therapy.

Importance of dosing — Excessive dosing of antithrombotic and antiplatelet agents is common and associated with an increase in bleeding risk. This issue is discussed separately. (See "Anticoagulant therapy in non-ST elevation acute coronary syndromes", section on 'Anticoagulant Use'.)

Transfusion thresholds — Red blood cell transfusion is generally reserved for severe or symptomatic anemia, such as hemoglobin <8 g/dL or hemoglobin 8 to 10 g/dL with hemodynamic instability or ongoing ischemia. Clinical judgment is required to determine if transfusion is likely to improve oxygen delivery or if there are other reasons to consider transfusion such as active bleeding or trauma.

The rationale for using a restrictive transfusion strategy (limiting transfusions, transfusing for a lower rather than a higher hemoglobin level) includes avoiding risks of transfusion reactions or transfusion-transmitted infection and limiting burdens and costs while using evidence-based thresholds for optimal outcomes. While anemia correlates with worse outcomes, this is likely to be an association rather than causation.

Potassium and magnesium — There are no clinical trials documenting the benefits of electrolyte replacement in acute MI. We recommend maintaining the serum potassium concentration above 4.0 meq/L and a serum magnesium concentration above 2.0 meq/L (2.4 mg/dL or 1 mmol/L). Much of the evidence for this recommendation was derived from studies before the routine use of beta blocker and the use of reperfusion in many patients.

Prophylactic intravenous or intramuscular lidocaine to prevent VT/VF in the acute MI patient is not recommended [31]. Recommended prophylactic measures include early administration of a beta blocker and treatment of hypokalemia and hypomagnesemia. Treatment of ventricular tachyarrhythmias in the setting of acute MI is discussed separately.

Clinical trials have identified a number of factors predicting high risk and a benefit from an early invasive strategy [33-37]. These include the presence and extent of ST segment depression, elevated cardiac biomarkers, evidence of hemodynamic instability, and persistent chest pain despite appropriate medical therapy. These variables have been used to create risk scores such as TIMI, GRACE, and PURSUIT.

TIMI risk score — Analysis of data from the TIMI 11B and ESSENCE trials found that seven variables at presentation were independently predictive of outcome in patients with unstable angina or an acute non-ST elevation MI; a value of one was assigned when a factor was present and 0 when it was absent (calculator 1) [40]:

● Age ≥65 years

● Presence of at least three risk factors for coronary heart disease (hypertension, diabetes, dyslipidemia, smoking, or positive family history of early MI).

● Prior coronary stenosis of ≥50 percent

● Presence of ST segment deviation on admission electrocardiogram

● At least two anginal episodes in prior 24 hours

● Elevated serum cardiac biomarkers

● Use of aspirin in prior seven days (which is probably a marker for more severe coronary disease) [41]

Patients are considered to be at low risk with a score of 0 to 2, intermediate risk with a score of 3 to 4, and high risk with a score of 5 to 7 (figure 1).

GRACE risk score — The simplified GRACE calculation uses eight parameters to predict death and MI in hospital and at six months: age, heart rate, systolic blood pressure, creatinine, Killip Class congestive heart failure, cardiac arrest, ST segment deviation, and elevated cardiac enzyme markers [42,43].

Avoidance of fibrinolysis — Prospective trials have demonstrated that fibrinolytic therapy is not beneficial in patients with a non-ST elevation acute coronary syndrome (ACS) [33,44]. We do not recommend the routine use of fibrinolytic agents in patients with a non-ST elevation ACS [15].

Immediate angiography and revascularization — Patients who have a non-ST elevation ACS and one or more of the following characteristics are at extremely high risk of an adverse cardiovascular event in the short term:

● Hemodynamic instability or cardiogenic shock

● Severe left ventricular dysfunction or heart failure

● Recurrent or persistent rest angina despite intensive medical therapy

● New or worsening mitral regurgitation or new ventricular septal defect

● Sustained ventricular arrhythmias

MI with normal coronary arteries — In clinical trials, 9 to 14 percent of patients with a non-ST elevation ACS who undergo early angiography have no significant coronary stenosis [34,35,46-48]. Possible mechanisms include rapid clot lysis, vasospasm, myocarditis, and coronary microvascular disease. Such patients have a much better short-term prognosis than those with a culprit lesion [47,48].

By convention, acute disease is distinguished from chronic disease at an arbitrary time point of two weeks from initial clinical presentation (hyperacute: <24 hours, acute: 1 to 14 days, subacute: >14 to 90 days, chronic: >90 days) and typically manifests with symptoms.

Intimal tear without hematoma, penetrating aortic ulcer, aortic intramural hematoma, and periaortic hematoma are variants of the classically described aortic dissection ( figure 1) [1-3].

For spontaneous dissection, it is uncertain whether the initiating event is a primary rupture of the intima with secondary dissection of the media or primary hemorrhage within the media and subsequent rupture of the overlying intima [4].

A higher mean pressure in the false lumen can cause dynamic or static compression and occlusion of the true lumen with malperfusion of the branches of the aorta, resulting in end-organ ischemia (coronary, cerebral, spinal, extremity, visceral) [6].

Aortic regurgitation, coronary ischemia, and cardiac tamponade can occur if the dissection progresses proximally to involve the aortic valve and coronary arteries or the pericardial sac, respectively.

Iatrogenic or traumatic — Aortic dissection that does not occur spontaneously can be due to instrumentation or trauma. Traumatic tears typically involve the descending thoracic aorta just distal to the subclavian artery and are reviewed in detail elsewhere [16]. Iatrogenic or traumatic injury (eg, intra-aortic balloon pump placement, rapid deceleration motor vehicle accident) was responsible for 6 percent of cases of aortic intramural hematoma in one review [17].

In different series, aortic intramural hematoma accounted for 5 to 20 percent of patients with symptoms consistent with an aortic dissection (eg, acute aortic syndrome) [16-20].

For intramural hematoma, the outer media (toward the adventitia) is thinner, which may explain the higher risk of rupture for intramural hematoma compared with acute dissection [26-29]. Between 8 and 16 percent will evolve into aortic dissection [27,29,30].

Penetrating aortic ulcer — Penetrating aortic ulcer refers to a region of the aorta (ulcer-like projection) where the aortic intima is denuded with the lesion progressing through a variable amount of the aortic wall, over which there may or may not be overlying thrombus [31,32]. Penetrating aortic ulcers are typically associated with atherosclerotic changes of the adjacent aortic wall [32].

Penetrating aortic ulcer may be associated with hematoma within the media and may progress to perforation or aortic dissection [2,32-34]. Penetrating ulcer is the initiating lesion in <5 percent of all aortic dissections (image 1) [33,35].

Periaortic hematoma — Periaortic hematoma represents a contained aortic rupture due to slow oozing from the damaged aorta at or near the site of aortic injury (figure 1). Periaortic hematoma is more common in aortic intramural hematoma compared with acute aortic dissection.

Not surprisingly, patients with periaortic hematoma had higher rates of shock, cardiac tamponade, and altered consciousness/coma and had a significantly higher mortality rate compared with those without a periaortic hematoma (33 versus 20 percent).

Type B aortic dissection is considered complicated if there is evidence of malperfusion (figure 4 and figure 5), rapid expansion or aneurysmal degeneration of the aortic wall, impending or frank rupture, uncontrolled pain, or refractory hypertension (persisting despite three or more classes of antihypertensives at max doses) [53]. Patients with type B aortic dissection are otherwise considered uncomplicated.

Approximately 25 percent of patients presenting with type B aortic dissection are acutely complicated [ 54,55].

The overall incidence of acute aortic syndromes ranges from two to four cases per 100,000 individuals.

Risk factors associated with acute aortic dissection include hypertension (including mediated by cocaine or other mechanisms [61,62]), atherosclerosis, prior cardiac surgery, aortic aneurysm, connective tissue disorder (eg, Marfan syndrome (table 1), Loeys-Dietz syndrome), bicuspid aortic valve, and prior aortic surgery [41,60].

In a separate review, in patients under age 40, only 34 percent had a history of hypertension, and only 1 percent had a history of atherosclerosis [64].

In the authors' experience, acute aortic syndromes are rarely identified as incidental findings with minimal symptoms on advanced imaging studies, although this has been reported [66-69]. Although very similar, there are some differences in the clinical manifestations of acute aortic dissection and those of other acute aortic syndromes.

The acute onset of severe chest or back pain occurs in 80 to 90 percent of patients with acute aortic dissection [63]. The pain is usually described as severe, sharp, or "tearing" and is located in the anterior chest pain for type A aortic dissection and in the posterior chest or back pain for type B aortic dissection.

Other symptoms or signs can be related to progression of the dissection and end-organ malperfusion (eg, shock, syncope, acute heart failure, myocardial ischemia, stroke, paraplegia, extremity ischemia, mesenteric ischemia) [ 41,58,60,63,70]

Pain is also common with intramural hematoma [26,27,30], but since patients with intramural hematoma are more likely to have type B aortic lesions compared with acute aortic dissection (eg, 60 versus 35 percent in the International Registry of Acute Aortic Dissections review), patients are more likely to present with upper or lower back pain [17,30,71].

Other manifestations seen with type A aortic dissections, such as myocardial infarction, stroke, aortic regurgitation, and syncope, are relatively infrequent with type A intramural hematomas [ 17,30,43,72].

Pericardial effusion occurs in 60 to 70 percent of type A aortic intramural hematoma and is much more common than with acute aortic dissection [26,27,30,39]. Malperfusion and aortic valve regurgitation is less common with aortic intramural hematoma.

No biomarkers are available to provide a diagnosis of acute aortic syndrome; however, for acute aortic dissection, a low D-dimer (less than 500 mcg/L) may help exclude the diagnosis among those who present with chest pain [73].

Although certain radiographic features, such as a widened mediastinum and pleural effusions, may raise suspicion for aortic disease, chest radiographs are not sensitive for a diagnosis of acute aortic syndromes but are also often not completely normal [30,74].

Computed tomographic (CT) angiography (eg, 3 to 4 mL/second peripheral injection with a 35-second imaging delay) is the diagnostic imaging modality of choice in hemodynamically stable patients. It is highly sensitive and specific for imaging aortic pathology.

Other imaging modalities such as transthoracic echocardiography (TTE) or transesophageal echocardiography (TEE) may be useful in hemodynamically unstable patients.

The presence of intraluminal thrombus is a good marker of the false lumen, especially in the chronic phase. In the majority of cases, the false lumen is larger than the true lumen.

Absence of an intimal flap is a prerequisite for the imaging diagnosis of aortic intramural hematoma [39].

Missed diagnosis — Acute aortic syndromes are frequently mistaken for other etiologies that cause chest pain, the most common being acute coronary syndromes. The incidence of initial misdiagnosis is up to 40 percent and may be more common when the ascending aorta is involved [77,78].

In one review of 68 patients, the likelihood of missed diagnosis, which occurred in 38 percent, was significantly higher in the absence of a pulse deficit or absence of widened mediastinum on chest radiography. Thus, these features cannot be used to exclude a diagnosis.

In a review of 127 patients with a final diagnosis of type A aortic dissection, an inappropriate initial diagnosis occurred in 37 percent of patients [79]. Significant factors leading to the initially incorrect diagnosis were walk-in status and presence of coronary malperfusion.

In one review of 66 patients, of whom 26 patients with acute aortic syndrome were initially misdiagnosed, acute coronary syndrome was the most common misdiagnosis and resulted in treatment with aspirin in all patients, clopidogrel in 1, heparin in 22, and fibrinolytic agents in 3 [77]. Exposure to antithrombotic agents was associated with major bleeding and a trend toward greater in-hospital mortality. In addition, there were higher rates of hemodynamic instability, hemorrhagic pericardial fluid, and hemorrhagic pleural effusion.

Triple rule-out CT (TRO-CT) is a modified coronary CT angiography protocol with extended thoracic coverage that may have advantages for identifying life-threatening causes of chest pain in the emergency department beyond coronary heart disease, such as pulmonary thromboembolism or acute aortic syndromes [81].

Other acute aortic pathologies include aortic aneurysm, chronic aortic disease with new symptoms, and complications of prior aortic repair (eg, endoleak, pseudoaneurysm) [82].

The general principles of the treatment of acute aortic syndromes are similar to acute aortic dissection (table 3). A crucial aspect of early therapy is ensuring an early and correct diagnosis so that the appropriate treatment can be instituted in a timely fashion. Approximately two-thirds of aortic dissections present as type A, and one-third are type B [1].

Acute medical management of acute aortic syndromes includes controlling pain and anti-impulse therapy by controlling the blood pressure to minimize the likelihood of rupture or progression, unless hypotension is present. These should be initiated immediately for all patients (type A and type B dissections) once the diagnosis has been made but should not interfere with the timely transfer to the operating room for those with indications for immediate aortic repair.

The treatment of type A acute aortic syndromes is surgical, with ongoing medical management of type A lesions reserved for patients who would not survive surgery. By contrast, type B acute aortic syndromes can generally be managed medically, with surgery or endovascular intervention reserved for those experiencing complications or progressive symptoms.

Acute medical management — Patients with suspected acute aortic syndromes should be admitted to an intensive care unit as rapidly as possible after confirmation of the diagnosis for pain control with morphine and anti-impulse therapy to reduce systolic blood pressure to the lowest level tolerated, typically using beta blockers (table 3) [2,80].

A toxicology screen should be considered in any patient presenting with an acute aortic syndrome, particularly if there are no other known predisposing factors. The results of toxicology screening may alter the pharmacological agents chosen for acute medical management.

Before fluid volume is administered, hypotensive patients should be evaluated to determine if the cause is hemopericardium with tamponade, valvular dysfunction, or left ventricular systolic dysfunction. In patients with cardiac tamponade, percutaneous pericardiocentesis can accelerate bleeding and shock [84].

Anti-impulse therapy aims to reduce the velocity of left ventricular contraction, thereby decreasing shear stress and minimizing lesion progression [85]. This is usually accomplished with antihypertensive therapy, typically using intravenous beta blockers (table 4). Inotropic agents should be avoided since they will increase aortic wall shear stress and may lead to progression. Systolic blood pressure should be reduced to the lowest level that is tolerated without compromising mentation or urine output, generally between 100 and 120 mmHg. Although there are no randomized trials in patients with acute aortic syndromes, but observational studies suggest that lowering blood pressure reduces the rate of progression [86,87].

If after beta blockade the systolic blood pressure remains elevated, nitroprusside can be added, if needed, to achieve a systolic blood pressure of 100 to 120 mmHg. The initial dose of nitroprusside is 0.25 to 0.5 mcg/kg per minute to a maximum of 10 mcg/kg per minute. Nitroprusside should not be used without first controlling heart rate with beta blockade since vasodilation alone induces reflex activation of the sympathetic nervous system, leading to enhanced ventricular contraction and increased aortic wall shear stress.

While nitroprusside is the preferred second-line agent, intravenous nicardipine, clevidipine, angiotensin converting enzyme (ACE) inhibitors, verapamil, or diltiazem may also be effective in lowering blood pressure [2,84,88,89]. Other direct vasodilators, such as hydralazine, should be avoided since they increase aortic wall shear stress and provide less accurate and reversible control of the blood pressure.

Beta blockers are generally avoided in patients with acute cocaine intoxication [90,91], but selective beta blockers such as esmolol or mixed agents such as labetalol may be reasonable for treating patients with acute aortic dissection or other acute aortic syndromes, when needed (table 4). The use of non-selective beta blockers alone may lead to unopposed alpha stimulation worsening hypertension [92-94].

Means to provide pain control other than antihypertensive therapies are warranted. This may include intravenous opioids as needed or patient-controlled analgesia (PCA) as would be administered to a postoperative patient, in the absence of contraindications.

For patients who are not undergoing immediate intervention, follow-up clinical examination and vascular imaging (computed tomographic [CT] or magnetic resonance [MR] angiography) are performed at 1, 3, 6, and 12 months and annually thereafter to detect malperfusion or aneurysm formation [110].

Type A acute aortic syndromes – For type A acute aortic syndromes, the definitive treatment is surgical. Surgery may not be feasible in patients of advanced age or other comorbid conditions, and these patients are managed medically. Compared with patients with classic ascending aortic dissection who are managed medically, patients with type A aortic intramural hematoma have a better prognosis. (See 'Ascending aorta (type A)' above.)

Type B acute aortic syndromes – For type B acute aortic syndromes, the treatment is generally medical, with surgery or endovascular intervention reserved for those experiencing complications or refractory pain or progressive symptoms. For most patients with complicated type B lesions, we suggest an initial endovascular approach, rather than open surgery, provided the patient's anatomy is suitable for endograft placement (Grade 2C). Perioperative morbidity and mortality is lower for an endovascular compared with open surgical approach. However, patients with genetically mediated thoracic aortic aneurysm/dissection affecting the descending thoracic aorta who have complications should be managed using open surgical techniques. (See 'Descending aorta (type B)' above.)

Aortic dissection is relatively uncommon, but it often presents acutely as a catastrophic illness with severe chest or back pain and acute hemodynamic compromise. Early and accurate diagnosis and treatment are crucial for survival.

Seasonal variation in the incidence of aortic dissection has been described, with winter months associated with higher admission rates for aortic dissection [21,22].

The incidence of acute aortic dissection in the general population is estimated to range from 2.6 to 3.5 per 100,000 person-years [17-20].

Patients with acute aortic dissection tend to be 60- to 80-year-old males [3-5,23-25].

Hypertension – The most important predisposing factor of acute aortic dissection is systemic hypertension [4,5,23,33,34]. In the IRAD review, 76.6 percent had a history of hypertension [4]. Hypertension was more common in those with a distal (type B) dissection compared with a type A dissection (70 versus 36 percent) [3,35].

An abrupt, transient, severe increase in blood pressure has been associated with acute aortic dissection through various mechanisms. Crack cocaine, which may cause transient hypertension due to catecholamine release, accounted for 37 percent of dissections in a report of an urban population [36]. The mean duration from last cocaine use to the onset of symptoms was 12 hours.

Genetically mediated connective tissue disorders (eg, Marfan syndrome, Ehlers-Danlos syndrome) – In an IRAD review, Marfan syndrome was present in 50 percent of those under age 40, compared with only 2 percent of older patients [28]. Most patients with Marfan syndrome (image 2) and aortic dissection have a family history of dissection.

Preexisting aortic aneurysm – In an IRAD review, 13 percent of patients had a known aortic aneurysm prior to dissection [28]. The ascending aorta was more often the site of origin of the dissection than the aortic arch or descending aorta. Such a history was more common in patients under age 40 (19 percent).

Bicuspid aortic valve – In an IRAD review, 9 percent of patients under age 40 with aortic dissection had a bicuspid valve, compared with 1 percent of those over age 40 [28] and 1 percent in the general population.