The Concept of Atherosclerotic Risk Factors
The systematic study of risk factors for atherosclerosis emerged from a coalescence of experimental results, as well as from cross-sectional and ultimately longitudinal studies in humans. The prospective, community-based Framingham Heart Study provided rigorous support for the concept that hypercholesterolemia, hypertension, and other factors correlate with cardiovascular risk. Similar observational studies performed worldwide bolstered the concept of "risk factors" for cardiovascular disease.
From a practical viewpoint, the cardiovascular risk factors that have emerged from such studies fall into two categories: those modifiable by lifestyle and/or pharmacotherapy, and those that are immutable, such as age and sex. The weight of evidence supporting various risk factors differs. For example, hypercholesterolemia and hypertension certainly predict coronary risk, but the magnitude of the contributions of other so-called nontraditional risk factors, such as levels of homocysteine, levels of lipoprotein (a) [Lp(a)], and infection, remains controversial. Moreover, some biomarkers that predict cardiovascular risk may not participate in the causal pathway for the disease or its complications. For example, recent genetic studies suggest that C-reactive protein (CRP) does not itself mediate atherogenesis, despite its ability to predict risk. Table 241-1 lists the risk factors recognized by the current National Cholesterol Education Project Adult Treatment Panel III (ATP III). The sections below will consider some of these risk factors and approaches to their modification.
Table 241–1. Major Risk Factors (Exclusive of LDL Cholesterol) that Modify LDL Goals |Favorite Table|Download (.pdf)
Table 241–1. Major Risk Factors (Exclusive of LDL Cholesterol) that Modify LDL Goals
|Hypertension (BP ≥140/90 mmHg or on antihypertensive medication)|
|Low HDL cholesterol* [<1.0 mmol/L (<40 mg/dL)]|
|Family history of premature CHD|
|CHD in male first-degree relative <55 years|
|CHD in female first-degree relative <65 years|
|Age (men ≥45 years; women ≥55 years)|
|Lifestyle risk factors|
|Obesity (BMI ≥30 kg/m2)|
|Emerging risk factors|
|Impaired fasting glucose|
Abnormalities in plasma lipoproteins and derangements in lipid metabolism rank among the most firmly established and best understood risk factors for atherosclerosis. Chapter 356 describes the lipoprotein classes and provides a detailed discussion of lipoprotein metabolism. Current ATP III guidelines recommend lipid screening in all adults >20 years of age. The screen should include a fasting lipid profile (total cholesterol, triglycerides, LDL cholesterol, and HDL cholesterol) repeated every 5 years.
ATP III guidelines strive to match the intensity of treatment to an individual's risk. A quantitative estimate of risk places individuals in one of three treatment strata (Table 241-2). The first step in applying these guidelines involves counting an individual's risk factors (Table 241-1). Individuals with fewer than two risk factors fall into the lowest treatment intensity stratum [LDL goal <4.1 mmol/L (<160 mg/dL)]. In those with two or more risk factors, the next step involves a simple calculation that estimates the 10-year risk of developing coronary heart disease (CHD) (Table 241-2); see http://www.nhlbi.nih.gov/guidelines/cholesterol/ for the algorithm and a downloadable risk calculator. Those with a 10-year risk ≤20% fall into the intermediate stratum [LDL goal <3.4 mmol/L (<130 mg/dL)]. Those with a calculated 10-year CHD risk of >20%, any evidence of established atherosclerosis, or diabetes (now considered a CHD risk-equivalent) fall into the most intensive treatment group [LDL goal <2.6 mmol/L (<100 mg/dL)]. Members of the ATP III panel recently suggested <1.8 mmol/L (<70 mg/dL) as a goal for very high-risk patients and an optional goal for high-risk patients based on recent clinical trial data (Table 241-2). Beyond the Framingham algorithm, there are multiple risk calculators for various countries or regions. Risk calculators that incorporate family history of premature (CAD) and a marker of inflammation (CRP) have been validated for U.S. women and men.
Table 241–2. LDL Cholesterol Goals and Cutpoints for Therapeutic Lifestyle Changes (TLC) and Drug Therapy in Different Risk Categories |Favorite Table|Download (.pdf)
Table 241–2. LDL Cholesterol Goals and Cutpoints for Therapeutic Lifestyle Changes (TLC) and Drug Therapy in Different Risk Categories
|LDL Level, mmol/L (mg/dL)|
|Risk Category||Goal||Initiate TLC||Consider Drug Therapy|
|Very high||<1.8 (<70)||≥1.8 (≥70)||≥1.8 (≥70)|
|ACS, or CHD w/DM, or multiple CRFs|
|High||<2.6 (<100)||≥2.6 (≥100)||≥2.6 (≥100) [<2.6 (<100): consider drug Rx]|
|CHD or CHD risk equivalents (10-year risk >20%) ||[optional goal: <1.8 (<70)]|
|If LDL <2.6 (<100)||<1.8 (<70)|
|Moderately high||<2.6 (<100)||≥3.4 (≥130)||≥3.4 (≥130) [2.6–3.3 (100–129): consider drug Rx]|
|2+ risk factors (10-year risk, 10–20%)|
|Moderate||<3.4 (<130)||≥3.4 (≥130)||≥4.1 (≥160)|
|2+ risk factors (risk <10%)|
|Lower||<4.1 (<160)||≥4.1 (≥160)||≥4.9 (≥190)|
|0–1 risk factor|
The first maneuver to achieve the LDL goal involves therapeutic lifestyle changes (TLC), including specific diet and exercise recommendations established by the guidelines. According to ATP III criteria, those with LDL levels exceeding goal for their risk group by >0.8 mmol/L (>30 mg/dL) merit consideration for drug therapy. In patients with triglycerides >2.6 mmol/L (>200 mg/dL), ATP III guidelines specify a secondary goal for therapy: "non-HDL cholesterol" (simply, the HDL cholesterol level subtracted from the total cholesterol). Cutpoints for the therapeutic decision for non-HDL cholesterol are 0.8 mmol/L (30 mg/dL) more than those for LDL.
An extensive and growing body of rigorous evidence now supports the effectiveness of aggressive management of LDL. Addition of drug therapy to dietary and other nonpharmacologic measures reduces cardiovascular risk in patients with established coronary atherosclerosis and also in individuals who have not previously experienced CHD events (Fig. 241-3). As guidelines often lag the emerging clinical trial evidence base, the practitioner may elect to exercise clinical judgment in making therapeutic decisions in individual patients.
Lipid lowering reduces coronary events, as reflected on this graph showing the reduction in major cardiovascular events as a function of low-density lipoprotein level in a compendium of clinical trials with statins. (Adapted from CTT Collaborators, Lancet 366:1267, 2005.) The Management of Elevated Cholesterol in the Primary Prevention Group of Adult Japanese (MEGA), Treating to New Targets (TNT), and Incremental Decrease in Endpoints through Aggressive Lipid Lowering (IDEAL) studies have been added.
LDL-lowering therapies do not appear to exert their beneficial effect on cardiovascular events by causing a marked "regression" of stenoses. Angiographically monitored studies of lipid lowering have shown at best a modest reduction in coronary artery stenoses over the duration of study, despite abundant evidence of event reduction. These results suggest that the beneficial mechanism of lipid lowering does not require a substantial reduction in the fixed stenoses. Rather, the benefit may derive from "stabilization" of atherosclerotic lesions without decreased stenosis. Such stabilization of atherosclerotic lesions and the attendant decrease in coronary events may result from the egress of lipids or from favorably influencing aspects of the biology of atherogenesis discussed above. In addition, as sizable lesions may protrude abluminally rather than into the lumen due to complementary enlargement, shrinkage of such plaques may not be apparent on angiograms. The consistent benefit of LDL lowering by 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase inhibitors (statins) observed in many risk groups may depend not only on their salutary effects on the lipid profile but also on direct modulation of plaque biology independent of lipid lowering.
A new class of LDL-lowering medications reduces cholesterol absorption from the proximal small bowel by targeting an enterocyte cholesterol transporter denoted Niemann-Pick C1-like 1 protein (NPC1L1). The NPC1L1 inhibitor ezetimibe provides a useful adjunct to current therapies to achieve LDL goals; however, no clinical trial evidence has yet demonstrated that ezetimibe improves CHD outcomes.
As the mechanism by which elevated LDL levels promote atherogenesis probably involves oxidative modification, several trials have tested the possibility that antioxidant vitamin therapy might reduce CHD events. Rigorous and well-controlled clinical trials have failed to demonstrate that antioxidant vitamin therapy improves CHD outcomes. Therefore, the current evidence base does not support the use of antioxidant vitamins for this indication.
The clinical use of effective pharmacologic strategies for lowering LDL has reduced cardiovascular events markedly, but even their optimal utilization in clinical trials prevents only a minority of these endpoints. Hence, other aspects of the lipid profile have become tempting targets for addressing the residual burden of cardiovascular disease that persists despite aggressive LDL lowering. Indeed, in the "poststatin" era, patients with LDL levels at or below target not infrequently present with acute coronary syndromes. Low levels of HDL present a growing problem in patients with CAD as the prevalence of metabolic syndrome and diabetes increases. Blood HDL levels vary inversely with those of triglycerides, and the independent role of triglycerides as a cardiovascular risk factor remains unsettled. For these reasons, approaches to raising HDL have emerged as a prominent next hurdle in the management of dyslipidemia. Weight loss and physical activity can raise HDL. Nicotinic acid, particularly in combination with statins, can robustly raise HDL. Some clinical trial data support the effectiveness of nicotinic acid in cardiovascular risk reduction. However, flushing and pruritus remain a challenge to patient acceptance, even with improved dosage forms of nicotinic acid. A combination of nicotinic acid with an inhibitor of prostaglandin D receptor, a mediator of flushing, may limit this unwanted effect of nicotinic acid and is currently in clinical trials, but it has not received regulatory approval.
Agonists of nuclear receptors provide another potential avenue for raising HDL levels. Yet patients treated with peroxisome proliferator–activated receptors alpha and gamma (PPAR-α and -γ) agonists have not consistently shown improved cardiovascular outcomes, and at least some PPAR-agonists have been associated with worsened cardiovascular outcomes. Other agents in clinical development raise HDL levels by inhibiting cholesteryl ester transfer protein (CETP). The first of these agents to undergo large-scale clinical evaluation showed increased adverse events, leading to cessation of its development. Clinical studies currently underway will assess the effectiveness of other CETP inhibitors that lack some of the adverse off-target actions encountered with the first agent.
(See also Chap. 247) A wealth of epidemiologic data support a relationship between hypertension and atherosclerotic risk, and extensive clinical trial evidence has established that pharmacologic treatment of hypertension can reduce the risk of stroke, heart failure, and CHD events.
Diabetes Mellitus, Insulin Resistance, and the Metabolic Syndrome
(See also Chap. 344) Most patients with diabetes mellitus die of atherosclerosis and its complications. Aging and rampant obesity underlie a current epidemic of type 2 diabetes mellitus. The abnormal lipoprotein profile associated with insulin resistance, known as diabetic dyslipidemia, accounts for part of the elevated cardiovascular risk in patients with type 2 diabetes. Although diabetic individuals often have LDL cholesterol levels near the average, the LDL particles tend to be smaller and denser and, therefore, more atherogenic. Other features of diabetic dyslipidemia include low HDL and elevated triglyceride levels. Hypertension also frequently accompanies obesity, insulin resistance, and dyslipidemia. Indeed, the ATP III guidelines now recognize this cluster of risk factors and provide criteria for diagnosis of the "metabolic syndrome" (Table 241-3). Despite legitimate concerns about whether clustered components confer more risk than an individual component, the metabolic syndrome concept may offer clinical utility.
Table 241–3. Clinical Identification of the Metabolic Syndrome—Any Three Risk Factors |Favorite Table|Download (.pdf)
Table 241–3. Clinical Identification of the Metabolic Syndrome—Any Three Risk Factors
|Risk Factor||Defining Level|
|Men (waist circumference)b||>102 cm (>40 in.)|
|Women||>88 cm (>35 in.)|
|Triglycerides||>1.7 mmol/L (>150 mg/dL)|
|Men||<1 mmol/L (<40 mg/dL)|
|Women||<1.3 mmol/L (<50 mg/dL)|
|Blood pressure||≥130/≥85 mmHg|
|Fasting glucose||>6.1 mmol/L (>110 mg/dL)|
Therapeutic objectives for intervention in these patients include addressing the underlying causes, including obesity and low physical activity, by initiating TLC. The ATP III guidelines provide an explicit step-by-step plan for implementing TLC, and treatment of the component risk factors should accompany TLC. Establishing that strict glycemic control reduces the risk of macrovascular complications of diabetes has proved much more elusive than the established beneficial effects on microvascular complications such as retinopathy and renal disease. Indeed, "tight" glycemic control may increase adverse events in patients with type 2 diabetes, lending even greater importance to aggressive control of other aspects of risk in this patient population. In this regard, multiple clinical trials, including the Collaborative Atorvastatin Diabetes Study (CARDS) that addressed specifically the diabetic population, have demonstrated unequivocal benefit of HMG-CoA reductase inhibitor therapy in diabetic patients over all ranges of LDL cholesterol levels (but not those with end-stage renal disease). In view of the consistent benefit of statin treatment for diabetic populations and the thus far equivocal results with PPAR agonists, the current stance of the American Diabetic Association that statins be considered for persons with diabetes older than age 40 who have a total cholesterol level ≥135 appears amply justified. Among the oral hypoglycemic agents, metformin possesses the best evidence base for cardiovascular event reduction.
Diabetic populations appear to derive particular benefit from antihypertensive strategies that block the action of angiotensin II. Thus, the antihypertensive regimen for patients with the metabolic syndrome should include angiotensin converting-enzyme inhibitors or angiotensin receptor blockers when possible. Most of these individuals will require more than one antihypertensive agent to achieve the recently updated American Diabetes Association blood pressure goal of 130/80 mmHg.
Male Sex/Postmenopausal State
Decades of observational studies have verified excess coronary risk in men compared with premenopausal women. After menopause, however, coronary risk accelerates in women. At least part of the apparent protection against CHD in premenopausal women derives from their relatively higher HDL levels compared with those of men. After menopause, HDL values fall in concert with increased coronary risk. Estrogen therapy lowers LDL cholesterol and raises HDL cholesterol, changes that should decrease coronary risk.
Multiple observational and experimental studies have suggested that estrogen therapy reduces coronary risk. However, a spate of clinical trials has failed to demonstrate a net benefit of estrogen with or without progestins on CHD outcomes. In the Heart and Estrogen/Progestin Replacement Study (HERS), postmenopausal female survivors of acute MI were randomized to an estrogen/progestin combination or to placebo. This study showed no overall reduction in recurrent coronary events in the active treatment arm. Indeed, early in the 5-year course of this trial, there was a trend toward an actual increase in vascular events in the treated women. Extended follow-up of this cohort did not disclose an accrual of benefit in the treatment group. The Women's Health Initiative (WHI) study arm, using a similar estrogen plus progesterone regimen, was halted due to a small but significant hazard of cardiovascular events, stroke, and breast cancer. The estrogen without progestin arm of WHI (conducted in women without a uterus) was stopped early due to an increase in strokes, and failed to afford protection from MI or CHD death during observation over 7 years. The excess cardiovascular events in these trials may result from an increase in thromboembolism (Chap. 348). Physicians should work with women to provide information and help weigh the small but evident CHD risk of estrogen ± progestin versus the benefits for postmenopausal symptoms and osteoporosis, taking personal preferences into account. Post hoc analyses of observational studies suggest that estrogen therapy in women younger than or closer to menopause than the women enrolled in WHI might confer cardiovascular benefit. Thus, the timing in relation to menopause or the age at which estrogen therapy begins may influence its risk/benefit balance.
The lack of efficacy of estrogen therapy in cardiovascular risk reduction highlights the need for redoubled attention to known modifiable risk factors in women. The recent JUPITER trials randomized over 6000 women over age 65 without known cardiovascular disease with LDL <130 mg/dL and high-sensitivity (hs) CRP >2 mg/L to a statin or placebo. The statin-treated women had a striking reduction in cardiovascular events, as did the men. This trial, which included more women than any prior statin study, provides strong evidence supporting the efficacy of statins in women who meet those entry criteria.
Dysregulated Coagulation or Fibrinolysis
Thrombosis ultimately causes the gravest complications of atherosclerosis. The propensity to form thrombi and/or lyse clots once they form clearly influences the manifestations of atherosclerosis. Thrombosis provoked by atheroma rupture and subsequent healing may promote plaque growth. Certain individual characteristics can influence thrombosis or fibrinolysis and have received attention as potential coronary risk factors. For example, fibrinogen levels correlate with coronary risk and provide information about coronary risk independent of the lipoprotein profile.
The stability of an arterial thrombus depends on the balance between fibrinolytic factors such as plasmin, and inhibitors of the fibrinolytic system such as plasminogen activator inhibitor 1 (PAI-1). Individuals with diabetes mellitus or the metabolic syndrome have elevated levels of PAI-1 in plasma, and this probably contributes to the increased risk of thrombotic events. Lp(a) (Chap. 356) may modulate fibrinolysis, and individuals with elevated Lp(a) levels have increased CHD risk.
Aspirin reduces CHD events in several contexts. Chapter 243 discusses aspirin therapy in stable ischemic heart disease, Chap. 244 reviews recommendations for aspirin treatment in acute coronary syndromes, and Chap. 370 describes aspirin's role in preventing recurrent ischemic stroke. In primary prevention, pooled trial data show that low-dose aspirin treatment (81 mg/d to 325 mg on alternate days) can reduce the risk of a first MI in men. Although the recent Women's Health Study (WHS) showed that aspirin (100 mg on alternate days) reduced strokes by 17%, it did not prevent MI in women. Current American Heart Association (AHA) guidelines recommend the use of low-dose aspirin (75–160 mg/d) for women with high cardiovascular risk (≥20% 10-year risk), for men with a ≥10% 10-year risk of CHD, and for all aspirin-tolerant patients with established cardiovascular disease who lack contraindications.
A large body of literature suggests a relationship between hyperhomocysteinemia and coronary events. Several mutations in the enzymes involved in homocysteine accumulation correlate with thrombosis and, in some studies, with coronary risk. Prospective studies have not shown a robust utility of hyperhomocysteinemia in CHD risk stratification. Clinical trials have not shown that intervention to lower homocysteine levels reduces CHD events. Fortification of the U.S. diet with folic acid to reduce neural tube defects has lowered homocysteine levels in the population at large. Measurement of homocysteine levels should be reserved for individuals with atherosclerosis at a young age or out of proportion to established risk factors. Physicians who advise consumption of supplements containing folic acid should consider that this treatment may mask pernicious anemia.
An accumulation of clinical evidence shows that markers of inflammation correlate with coronary risk. For example, plasma levels of CRP, as measured by a high-sensitivity assay (hsCRP), prospectively predict the risk of MI. CRP levels also correlate with the outcome in patients with acute coronary syndromes. In contrast to several other novel risk factors, CRP adds predictive information to that derived from established risk factors, such as those included in the Framingham score (Fig. 241-4). Recent Mendelian randomization studies do not support a causal role for CRP in cardiovascular disease. Thus, CRP serves as a validated biomarker of risk but probably not as a direct contributor to pathogenesis.
C-reactive protein (CRP) level adds to the predictive value of the Framingham score. hsCRP, high-sensitivity measurement of CRP. (Adapted from PM Ridker et al: Circulation 109:2818, 2004.)
Elevations in acute-phase reactants such as fibrinogen and CRP reflect the overall inflammatory burden, not just vascular foci of inflammation. Visceral adipose tissue releases proinflammatory cytokines that drive CRP production and may represent a major extravascular stimulus to elevation of inflammatory markers in obese and overweight individuals. Indeed, CRP levels rise with body mass index (BMI), and weight reduction lowers CRP levels. Infectious agents might also furnish inflammatory stimuli related to cardiovascular risk. To date, randomized clinical trials have not supported the use of antibiotics to reduce CHD risk.
Intriguing evidence suggests that lipid-lowering therapy reduces coronary events in part by muting the inflammatory aspects of the pathogenesis of atherosclerosis. For example, in the JUPITER trial, a prespecified analysis showed that those who achieved lower levels of both LDL and CRP had better clinical outcomes than did those who only reached the lower level of either the inflammatory marker or the atherogenic lipoprotein (Fig. 241-5). Similar analyses of studies of statin treatment in patients after acute coronary syndromes showed the same pattern. The anti-inflammatory effect of statins appears independent of LDL lowering, as these two variables correlated very poorly in individual subjects in multiple clinical trials.
Evidence from the JUPITER study that both LDL-lowering and anti-inflammatory actions contribute to the benefit of statin therapy in primary prevention. See text for explanation. hsCRP, high-sensitivity measurement of C-reactive protein (CRP). (Adapted from PM Ridker et al: Lancet 373:1175, 2009.)
The prevention of atherosclerosis presents a long-term challenge to all health care professionals and for public health policy. Both individual practitioners and organizations providing health care should strive to help patients optimize their risk factor profiles long before atherosclerotic disease becomes manifest. The current accumulation of cardiovascular risk in youth and in certain minority populations presents a particularly vexing concern from a public health perspective.
The care plan for all patients seen by internists should include measures to assess and minimize cardiovascular risk. Physicians must counsel patients about the health risks of tobacco use and provide guidance and resources regarding smoking cessation. Similarly, physicians should advise all patients about prudent dietary and physical activity habits for maintaining ideal body weight. Both National Institutes of Health (NIH) and AHA statements recommend at least 30 minutes of moderate-intensity physical activity per day. Obesity, particularly the male pattern of centripetal or visceral fat accumulation, can contribute to the elements of the metabolic syndrome (Table 241-3). Physicians should encourage their patients to take personal responsibility for behavior related to modifiable risk factors for the development of premature atherosclerotic disease. Conscientious counseling and patient education may forestall the need for pharmacologic measures intended to reduce coronary risk.
Issues in Risk Assessment
A growing panel of markers of coronary risk presents a perplexing array to the practitioner. Markers measured in peripheral blood include size fractions of LDL particles and concentrations of homocysteine, Lp(a), fibrinogen, CRP, PAI-1, myeloperoxidase, and lipoprotein-associated phospholipase A2, among many others. In general, such specialized tests add little to the information available from a careful history and physical examination combined with measurement of a plasma lipoprotein profile and fasting blood glucose. The high-sensitivity CRP measurement may well prove an exception in view of its robustness in risk prediction, ease of reproducible and standardized measurement, relative stability in individuals over time, and, most important, ability to add to the risk information disclosed by standard measurements such as the components of the Framingham risk score (Fig. 241-4). The addition of information regarding a family history of premature atherosclerosis in parents (a simply obtained indicator of genetic susceptibility), together with the marker of inflammation hsCRP, permits correct reclassification of risk in individuals—especially those whose Framingham scores place them at intermediate risk. Current advisories, however, recommend the use of the hsCRP test only in individuals in this CHD event risk group (10–20%, 10-year risk).
Available data do not support the use of imaging studies to screen for subclinical disease (e.g., measurement of carotid-intima/media thickness, coronary artery calcification, and use of computed tomographic coronary angiograms). Inappropriate use of such imaging modalities may promote excessive alarm in asymptomatic individuals and prompt invasive diagnostic and therapeutic procedures of unproven value. Widespread application of such modalities for screening should await proof that clinical benefit derives from their application.
Progress in human genetics holds considerable promise for risk prediction and for individualization of cardiovascular therapy. Many reports have identified single-nucleotide polymorphisms (SNPs) in candidate genes as predictors of cardiovascular risk. To date, the validation of such genetic markers of risk and drug responsiveness in multiple populations has often proved disappointing. The advent of technology that permits relatively rapid and inexpensive genome-wide screens, in contrast to most SNP studies, has led to identification of sites of genetic variation that do reproducibly indicate heightened cardiovascular risk (e.g., chromosome 9p21). The results of genetic studies should identify new potential therapeutic targets (e.g., the enzyme mutated in autosomal dominant hypercholesterolemia, abbreviated PCSK9) and may lead to genetic tests that help refine cardiovascular risk assessment in the future.
The Challenge of Implementation: Changing Physician and Patient Behavior
Despite declining age-adjusted rates of coronary death, cardiovascular mortality worldwide is rising due to the aging of the population, and the subsiding of communicable diseases and increased prevalence of risk factors in developing countries. Enormous challenges remain regarding translation of the current evidence base into practice. Physicians must learn how to help individuals adopt a healthy lifestyle in a culturally appropriate manner and to deploy their increasingly powerful pharmacologic tools most economically and effectively. The obstacles to implementation of current evidence-based prevention and treatment of atherosclerosis involve economics, education, physician awareness, and patient adherence to recommended regimens. Future goals in the treatment of atherosclerosis should include more widespread implementation of the current evidence-based guidelines regarding risk factor management and, when appropriate, drug therapy.