Cardiovascular Disease, Cancer, and Multimorbidity Interactions: Clinical Implications

With the evolution of medicine and its great developments in medical technology, life expectancy has increased (almost doubled) compared to a couple of hundred years ago. As a consequence of this increase in life expectancy, the incidence of diseases that were rare in the recent past (e.g., aortic aneurysms, atrioventricular conduction abnormalities, calcific aortic stenosis, and certain types of cancer, among others) due to the fact that they usually occur after 60 or 70 years of age has increased with alarming frequency today [1].

It also has become apparent that the rare occurrence of a patient having multiple morbid conditions in the past has become the rule rather than the exception today, particularly in the elderly. Moreover, it is also known that one disease may affect the development and progression of another disease leading to a vicious cycle (Fig. 1). Such a close interaction exists between cardiovascular disease (CVD), cancer, and other morbid conditions [2, 3]. This brief review will focus on interactions between CVD and cancer with other multiple morbid conditions and issues a practicing physician faces on a daily basis.

CVD and cancer are the two leading causes of death worldwide, and millions of cancer survivors are at an increased risk in developing CVD. It is estimated that there are more than 17 million cancer survivors (potential at risk for developing CVD) in the United States of America (USA), and this number continues to grow [4]. In addition to CVD, cancer survivors have other morbid conditions as well. This association, at least partially, is explained by both diseases sharing common risk factors and from accelerated vascular aging due to cancer and its associated therapies [4-7].

CVD, especially coronary atherosclerosis, and cancer typically occur more frequently in the elderly. Death from coronary atherosclerosis has declined substantially over the past several decades [8] (Fig. 2). With improved preventive measures and novel therapeutic modalities, it is expected that death related to CVD, particularly coronary atherosclerosis, will further decline in the future, as it has been seen with diseases of time past (e.g., rheumatic fever with rheumatic valve disease, peptic ulcer disease, and syphilis, among others) [1, 8].

As death related to CVD will further decline in the near future, death related to cancer will increase due to the aging population [1, 5, 8]. Among others, the development of cancer is partially related to the nature of the disease. Normal cells are constantly dividing and replicating their DNA; errors during this continuous process can occur resulting in mutations leading to cancer. These mutations increase with age and accumulate throughout the years. Further, environmental factors including carcinogens also facilitate the development of cancer. Thus, if one lives long enough, cancer eventually will occur. Even if a form of cancer can effectively be treated, cells will continue to divide, and new mutations may lead to another cancer. Studies have suggested that approximately 25% of patients aged 65 years or older and 11% of younger adults with a new diagnosis of cancer have a previous history of a different type of cancer [4, 6, 7, 9]. Due to its nature, prevention and management of cancer will be one of the greatest challenges in the next several decades.

Risk factors for CVD include age, gender, genetic predisposition, unhealthy diet, obesity, sedentary lifestyle, tobacco use, arterial hypertension, hyperlipidemia, and diabetes mellitus, among others. The same risk factors for CVD are considered to be risk factors for cancer. Each of these risk factors has a relatively small contribution to the development of disease, but the combination of several of these factors increases the incidence [10-12].

Aging is associated with organ and system dysfunction due to tissue fatigue and damage, as well as to various risk factors and stressors that accumulate over time. Dysfunction of the immune system that may occur with aging results in a diminished response to different pathogens and to the development of a chronic inflammatory process [3]. Inflammation, especially in the elderly, contributes to the development of multiple morbid conditions, such as CVD, cancer, and Alzheimer’s, among other chronic diseases. All of these morbid conditions that contribute to inflammation in turn further deteriorate the underlying disease leading to a vicious cycle [3].

Genetic predisposition can exist in which CVD and certain types of cancers affect several members within a family. For this reason, a detailed family history that may direct genetic studies, among others, is extremely important. Genome wide association studies have shown that both CVD and cancer have numerous genetic variances, but only a few are common in both diseases. As a general rule, each of these genetic variances has a relatively small contribution to the development of the disease, but a combination of these genetic variances increases the risk in the development of CVD, cancer, or both. Age-related clonal hematopoiesis of indeterminate potential is a common phenomenon where somatic mutations accumulate in cells of the blood or bone marrow. It is present in up to 20% of individuals 70 years of age or older, and it is associated with an increased risk of cancer, atherosclerosis, and heart failure (HF). Frequently, mutations in the epigenetic regulator TET2 are found in blood cells of individuals exhibiting clonal hematopoiesis. In certain instances, clonal hematopoiesis of indeterminate potential can be used in CVD assessment and as a biomarker for response to interventions to reduce cardiovascular risk [13, 14]. Also, mutations in one of the binding proteins of Wnt are involved in the development of coronary atherosclerosis and in certain types of cancer. This genetic overlap between CVD and cancer might represent novel biological shared pathways between these two diseases [13, 14].

An unhealthy diet is associated with CVD and cancer. For example, red meat consumption can increase the risk of atherosclerosis and has been associated with a higher incidence of colorectal cancer. Increased sodium intake is associated with arterial hypertension and gastric cancer. Excessive alcohol consumption may lead to dilated cardiomyopathy and esophageal, colorectal, liver, and laryngeal cancer. An unhealthy diet contributes to obesity, and it has been suggested that certain malignancies such as esophageal, liver, colorectal, prostatic, and breast cancer may be related to obesity. Proinflammatory cytokines and hormones produced within adipose tissue, such as interleukin-6 (IL-6), tumor necrosis factor, and leptin, are elevated in obese individuals suggesting that hormones and cytokines control both inflammation and metabolism. IL-6 stimulates the production of C-reactive protein (CRP), a biomarker of inflammation. Leptin exhibits proinflammatory action in several immune cells, such as macrophages and lymphocytes, among others [15].

A sedentary lifestyle is associated with CVD and cancer. Exercise has a wide range of systemic effects, and it has been found to be protective in the development of CVD and cancer. In animal models, it has been shown that exercise results in a reduction of malignant tumor progression. Further, it has been shown that exercise improves outcome in patients with cancer. The beneficial effect of exercise on cancer, among others, has been attributed to an alteration in CD8⁺ T-cell function [16]. Exercise also improves endothelial function and metabolic profile and contributes to the maintenance of a normal body weight. Further, the reduction of adipose tissue due to exercise results in a decrease in insulin, leptin, and inflammation, all of which contribute to the development of CVD and cancer [17].

Tobacco use is associated with oxidative stress, endothelial damage, thrombosis, and inflammation, all important factors in the development of CVD and cancer. It is also known that nicotine inhibits apoptosis and enhances angiogenesis, both contributing to the development and progression of cancer [10].

The role of arterial hypertension in CVD is well known. It also appears that there is some association between arterial hypertension and cancer. Angiotensin II, in addition to its role in CVD and arterial hypertension, stimulates the production of vascular endothelial growth factor (VEGF), promotes oxidative stress in the arterial wall, and has been associated with renal cell carcinoma [18].

Hyperlipidemia is related to multiple factors such as genetic, diet, body weight, and metabolic profile, among others. The association between hyperlipidemia and atherosclerosis has been well documented. Hyperlipidemia as a risk factor for cancer, however, is less well defined, though it appears to be associated with breast cancer and potentially with other malignancies [3, 10].

Diabetes mellitus, which is associated with obesity, is a risk factor for CVD and certain types of cancer including colorectal, liver, pancreatic, esophageal, breast, endometrial, kidney, and leukemia. Insulin resistance, increased insulin growth factor, and inflammation present in obesity and diabetes mellitus provide some explanation for the development of CVD and cancer [3, 19, 20].

All the above-mentioned risk factors contribute to the development of oxidative stress, endothelial dysfunction, neurohumoral activation, and immune system dysfunction resulting in inflammation that leads to CVD, cancer, and other morbid conditions. Inflammation can also alter DNA in which the frequency of random mutations increases compared to mutations that occur in normal tissue. Chronic inflammation, in addition to its contribution in the development of cancer, also constitutes a pathological basis for the development of CVD and metabolic abnormalities [3, 19, 20].

Inflammation in one organ is not limited only to that organ, but often manifests as a systemic process. Inflammation in visceral adipose tissue in obese individuals may activate and accelerate inflammation in distant organs such as the liver, pancreas, and arteries, among others. Inflammation also alters metabolism that in turn results in a further increase in inflammation leading to a vicious cycle. Inflammation may also lead to endothelial dysfunction and oxidative stress, which are important for the development of CVD and cancer (Fig. 3) [3, 11, 19, 20].

Several factors promote inflammation such as smoking, obesity, tobacco use, and unhealthy diet, among others, and eliminating these factors could decrease inflammation. A healthy diet, especially a Mediterranean diet, improves the metabolic profile decreasing inflammation, and thus can be beneficial in the prevention of CVD and cancer. Likewise, exercise at any intensity has a range of systemic effects including anti-inflammatory [3, 15-17, 21].

The incidence of cancer is higher in patients with CVD compared to the general population [4-7, 12, 22]. Recent studies suggest that approximately 70% of patients with coronary atherosclerosis die from other causes mostly malignancy [1]. Radiation related to diagnostic studies in patients with known or suspected CVD, and to therapeutic interventions, may increase the incidence of cancer. The magnitude of this risk is not well defined, but it should be taken into consideration together with the other risks related to these procedures [8].

In a study by Rinde et al. [23], 28,763 individuals without a history of myocardial infarction (MI) or cancer were followed for 15.7 years (median time). During follow-up, 1,747 subjects developed MI and at the same time 146 had cancer. Patients with MI had a 46% higher risk of developing cancer compared to those without MI. The incidence of cancer was highest the first 6 months after MI, though was not higher the following 2 years; however, the incidence of cancer increased again 3 years after MI [23]. Tissue ischemia and necrosis due to MI activate the innate immune system resulting in inflammation, angiogenesis, and an increase in tumor necrosis factor. All these biological changes following myocardial ischemia and necrosis contribute to the development of cancer. Studies suggest that MI may accelerate the growth of intestinal tumors [24-26].

A study from Germany suggested that HF is associated with almost all types of cancer. The strongest association was observed with lip, oral cavity, and pharynx followed by respiratory, genital organs in female, skin, lymphoid and hematopoietic tissues, digestive tract, and male genital organs [27]. The incidence of cancer increases approximately 1.5 years after the diagnosis of HF [12]. In HF, several proteins are secreted from the heart and affect tissues such as tumor necrosis factor, IL-6, IL-1, and VEGF that in turn affect cancer development and progression. Further, cardiac production of certain biomarkers (e.g., brain natriuretic peptide) may affect tumor growth. This is another indication that supports the association between HF and cancer. The incidence of cancer in HF has been estimated to range between 18.9 and 33.7 per 1,000 person years. Despite the poor prognosis in HF, noncardiovascular mortality in those patients ranges between 20% and 50%, and the larger proportion of noncardiovascular deaths is related to cancer [24, 25].

Though not yet well defined, pharmacologic agents used to treat CVD may promote or inhibit the development of cancer. Hypertension is one of the most common cardiovascular risk factors that requires lifelong treatment; such long-term therapy, in certain instances, may promote the development and progression of cancer. Recent evidence has focused on risks of cancer that may be associated with long-term antihypertensive therapy. However, it is difficult to prove cause and effect relationship between cardiovascular drugs and cancer since there are no targeted intervention trials. To illustrate this issue, few selective publications are presented.

A large meta-analysis that evaluated all classes of antihypertensive drugs (angiotensin receptor blockers, angiotensin converting enzyme inhibitors, β-blockers, diuretics, and calcium channel blockers) reported an increase in cancer or cancer-related deaths. Other meta-analyses, however, did not confirm these findings. Furthermore, another meta-analysis reported that the use of losartan decreased the incidence of cancer in diabetics, but use of candersartan and telmisartan increased the incidence of cancer in diabetics [24]. In 2017, the Food and Drug Administration (FDA) of the USA withdrew some widely used antihypertensive drugs such as losartan, valsartan, and inbesartan due to the presence of potentially carcinogenic bioproducts of the pharmaceutical process [28].

In Caucasian patients from a UK-based cohort, there was an association between lung cancer and angiotensin-converting enzyme inhibitors after 5 years of use that increased with longer duration of at least up to 10 years [29]. Battistoni and Volpe [28] concluded in a meta-analysis that the incidence of cancer or cancer-related deaths was not increased with angiotensin receptor blockers or angiotensin-converting enzyme inhibitors, β-blockers, and calcium channel blockers therapy, though the increased risk of cancer with the combination of angiotensin-converting enzyme inhibitors plus angiotensin receptor blockers cannot be excluded. Studies, however, included in this meta-analysis subsequently were retracted, and thus the information provided in this meta-analysis, at best, is questionable [28].

Diuretics, as a general rule, do not affect the incidence of cancer, but has been suggested that therapy with hydrochlorothiazide may increase the incidence of skin cancer due to photosensitivity [28]. The FDA in a recent session raised questions about the benefits of aspirin on cancer prevention and indicated that further research is needed in this area [30]. Likewise, the use of metformin for the treatment of diabetes mellitus has been associated with a decrease in the frequency of cancer; this effect is attributed to a decrease in insulin and insulin growth factor levels [31].

In general, observational studies lack randomization, while post hoc meta-analysis of randomized clinical trials may include important differences in control populations. These differences make it very difficult to use data generated to establish firm conclusions about the risk of cancer in relation to therapy. For this reason, more data needs to be collected from registries; processing of huge amounts of data using artificial intelligence most likely will help in this issue [32]. At this time, recommendations to interrupt therapies are not justified.

Finally, CVD, and especially left ventricular dysfunction, in certain cases may prohibit cancer treatment due to the cardiotoxicity effects of cancer therapies. In other instances, CVD may prohibit surgery or other interventional therapies in cancer patients [33].

The effects of cancer on CVD are basically related to the direct effects of cancer on the human body and on the cardiovascular system, and to the effects of cancer therapy on the cardiovascular system. Cancer cells affect the function of the immune system resulting in immunosuppression and inflammation that promotes the development and progression of CVD. Substances secreted from tumor cells promote thrombosis leading to a thromboembolic phenomenon. It has been reported that the incidence of stroke and myocardial infarction, among others, is increased even before the diagnosis of cancer [34].

Cardiac abnormalities occur in 60–70% of patients with carcinoid, but clinically apparent disease is less common. Carcinoid predominantly involves right heart valves, while left heart valves are much less often affected [35]. Other cardiac tumors primary or metastatic are rare. Pericarditis with or without pericardial effusion also can be seen in patients with cancer. In other instances, tumors may compress the arterial or venous system, most common being the superior vena cava due to chest tumors.

The rapid advancement in cancer therapies has resulted in an increase in cancer survivors who are known to be at a higher risk for developing CVD. Cardiotoxicity related to cancer therapy, especially left ventricular dysfunction, among others, is one of the main reasons that cancer survivors have an increased incidence of CVD [36]. In a recent study, 160,000 cancer patients were evaluated; among adolescent and young adult 5-year cancer survivors, it was shown that the cumulative mortality from CVD was 1.4 times greater (up to 40 years of follow-up) compared to the general population. In addition to CVD, cancer survivors can have other morbid conditions that may affect the outcome of CVD. Among 10,397 cancer survivors, the incidence of chronic morbid conditions was higher compared to their siblings [6, 7, 9].

The major cardiotoxic effects related to cancer therapies are briefly outlined. Anthracyclines (e.g., doxorubicin) are effective in the treatment of various types of cancers, but are associated with cardiotoxic effects resulting in left ventricular dysfunction, HF, acute myocarditis, and cardiac arrhythmias [36]. Therapy with platinum (e.g., oxaliplatin) may induce arterial hypertension and myocardial ischemia. Antimetabolites (e.g., fluorouracil) can induce myocardial ischemia and cardiac arrhythmias, while therapy with antimicrotubular agents (e.g., paclitaxel) is associated with cardiac arrhythmias (atrial and ventricular) and atrioventricular conduction abnormalities. Therapies that target human epidermal growth factor receptor 2-positive breast cancer (e.g., trastuzumab) are associated with left ventricular dysfunction and HF. VEGF signaling pathway inhibitors may induce myocardial edema, left ventricular hypertrophy, acute increase in arterial pressure, and venous or arterial thrombosis. Cardiac side effects of tyrosine kinase inhibitors include pulmonary hypertension, QT prolongation, and thromboembolic phenomena resulting in myocardial infarction, stroke, and peripheral vascular thromboembolic events. Immunomodulatory agents (e.g., lenalidomide) may result in venous or arterial thromboembolic events. Immune checkpoint inhibitors (e.g., nivolumab) activate the immune system by inhibiting cancer-mediated downregulation of T cells; this therapy, in certain cases, is associated with fulminant myocarditis. With the expanded use of immune checkpoint inhibitors in cancer patients, these complications will constitute a new clinical entity in the near future [36-38].

The effect of radiation on the heart can be classified as acute and chronic. Acutely myocardial ischemia, myocarditis, pericarditis, and cardiac arrhythmias may occur. When the endocardium is involved, fibrosis and fibroelastosis may affect the atrioventricular valves and produce mitral and tricuspid regurgitation; depending on the radiation window, the aortic valve can also be involved. Pericardial calcification may occur years after radiation therapy. When epicardial coronary arteries are involved, ostial stenoses are common [35]. Young women treated with radiation for left-sided breast cancer have over twice the risk of developing coronary artery disease compared to women treated for right-sided breast cancer [39]. Surgery for valve or coronary artery disease is often challenging due to the associated radiation damage to the pleura and chest tissues [35].

Cardiac arrhythmias, atrial and/or ventricular, are common in cancer patients and are at least partially related to cancer therapy. Development of atrial fibrillation and related anticoagulation therapy required to prevent thromboembolic complications can be a challenge in certain cancer patients with hematologic abnormalities [40, 41].

Risk factors for cardiotoxicity related to cancer therapy include accumulating dose of the drug, both extremes of the age spectrum (very young or very old), cardiovascular risk factors, combination therapy with anticancer drugs or radiation, and left ventricular dysfunction prior to the initiation of therapy. Certain preventive measures and close monitoring of patients with cardiovascular imaging and biomarkers may result in a decrease, but not elimination, of these often devastating toxic effects related to cancer therapy. Eliminating risk factors for CVD and moderate aerobic exercise may contribute to this effort [17, 36]. The use of β-blockers, angiotensin-converting enzyme inhibitors, or angiotensin receptor blockers to prevent cardiotoxicity still is controversial and is based on limited studies [36, 42]. More recently, it has been suggested that statins may decrease cardiotoxicity related to chemotherapy in patients with breast cancer [43]. Genetic predisposition has been postulated to be one of the reasons why certain patients may develop cardiotoxicity. Pharmocogenetic testing may be beneficial in certain cases. There are some data that have shown that circulating microRNA biomarkers may be used to predict cardiotoxicity [44].

Thrombocytopenia and thrombasthenia due to underlying disease and/or therapy is not uncommon in patients with cancer. Thrombocytopenia, among others, may prohibit invasive therapeutic interventions in patients with cancer and CVD. Further, cancer patients are at higher risk of thromboembolic and ischemic events after percutaneous coronary intervention [33, 45].

Cancer in certain cases may prohibit surgery and/or percutaneous interventions in patients with cardiovascular disease [46]. Due to the cardiotoxicity of certain cancer therapies and due to the multiple effects of cancer on the human body, treatment strategies in patients with CVD, in certain cases, should be modified and potentially discontinued, at least temporally [25, 36, 37].

Cutting the Gordian KnotAlexander the Great

As a general rule, there is no curative therapy for most chronic diseases, and with advancements in therapies, patients live longer, but with the disease. The longer an individual lives, the greater the possibility to develop another morbid condition [1, 47].

It is well appreciated today that interactions exist between the various organs. There are organ-organ interactions, organ-system interactions, and organ interactions with the entire body. These interactions are critical under normal conditions to maintain homeostasis; however, in different diseases, these interactions can become detrimental, as a disease in one organ may affect another organ and in certain instances may affect the entire body [2, 3, 48] (Fig. 1, 3).

In patients with cancer, multiple cancer-organ-system interactions exist. Cancer cells in response to hypoxia secrete VEGF increasing angiogenesis and promoting metastasis. Inflammation due to cancer further deteriorates the function of other organs, and proinflammatory cytokines secreted from cancer cells promote muscle wasting leading to cachexia [3, 47]. The heart in certain disorders and diseases (e.g., HF) also has multiple CVD-organ-system interactions that can essentially affect all organs [2, 3]. All these morbid conditions contribute to the progression of inflammation that in turn further deteriorates these underlying morbid conditions leading to a vicious cycle (Fig. 1, 3). In addition, the multiple risk factors typically present in the elderly, as well as different stresses that accumulate with age, contribute to the progression of inflammation [3, 49].

The association of these pathophysiologic abnormalities with CVD and cancer, and the strong connection among systems, provides an opportunity for novel targeted therapies (Fig. 4). Thus, development of new pharmacologic agents and other therapeutic modalities should be addressed to focus on multiple organs and systems and not just to one disease or one particular organ. Clinical practice guidelines, however, as a general rule, are focused only on one disease mostly ignoring multimorbidity [3, 50].

In the CANTOS trial, the anti-inflammatory medication canakinumab that inhibits IL-1β was used for the treatment of atherosclerotic disease and was shown to reduce the incidence of cardiovascular events in post-MI patients with a high C-reactive protein level of 2 mg/L or greater, as well as the incidence of lung cancer [51]. The effect of canakinumab in lung cancer mortality was more obvious in smokers. In this trial, total cancer mortality was significantly lower in the canakinumab group compared to placebo (p = 0.0007 trend across groups). It should be mentioned, however, that the CANTOS trial was not designed for cancer detection or treatment.

Therapies targeted at the regulation of systems have proven to be effective as well. Inhibition of the renin-angiotensin-aldosterone system that inhibits various pathways and inflammation has proven to be beneficial in the management of several diseases. Likewise, blockade of β-adrenergic receptors involving the sympathetic nervous system inhibits the effects of various stressors not only on the cardiovascular system, but also other organs. Therapy with β-adrenergic receptor blockers has proven to be effective in the treatment of different diseases such as arterial hypertension, HF, and cardiac arrhythmias, among others [2, 8]. The sodium glucose cotransporter 2 inhibitors (e.g., empagliflozin, canaglifizin, and dapagliflozin), in addition to their effects on blood glucose, have pleiotropic effects that can be used for the treatment of other diseases as well (e.g., HF) [52, 53].

Exercise has a wide range of systemic effects including anti-inflammatory, improved metabolic profile and endothelial function, maintenance of a normal body weight, decrease in the incidence of CVD and cancer, and decrease in the incidence and improvement in the outcomes of diabetes mellitus, among others. The reduction of adipose tissue through physical activity results in a reduction of insulin, leptin, and inflammation [17]. A healthy diet, particularly the Mediterranean diet, also contributes to the improvement of the metabolic profile, decrease in inflammation, and has been shown to be beneficial in the prevention of cardiovascular and other diseases as well [21].

It is unlikely that the problem of multimorbidity can be solved with randomized clinical trials, the cornerstone of guidelines, as patients with multiple disorders have been systematically excluded from these trials. Collecting information from registries and processing large amounts of data using artificial intelligence may assist the clinician when treating an individual patient with multimorbidity in the future [32].

Our understanding of CVD and cancer is continuously evolving. Diagnostic and therapeutic modalities related to these diseases are constantly changing [1, 8]. In patients with CVD and cancer, studies typically were conducted in patients with only one disease and not in patients with both of these diseases. Further, as a general rule, the multiple morbid conditions often present in patients with CVD plus cancer were not taken into consideration. Currently, guidelines do not exist for patients with CVD plus cancer who also have multiple other morbid conditions. In these cases, therapy should be individualized, as one size does not fit all [2, 50, 54]. For the management of these patients, team work is necessary in most of the instances. The physician should understand the pathophysiologic mechanisms of the underlying diseases, pharmacokinetics and pharmacodynamics of the pharmacologic agents, and drug-drug interactions. Patient characteristics and preferences, and physician’s experience for each therapeutic modality or procedure in each particular hospital, should be taken into consideration. This is a rational and practical approach for the individual patient with cardiovascular disease, cancer, and other morbid conditions (Fig. 5) [50, 54].

The physician who is taking care of patients with CVD plus cancer that also have other multimorbidity, in addition to knowledge, should have excellent clinical skills, clinical wisdom, clinical experience, and common sense in order to be able to apply the fast-accumulating knowledge to the individual patient. Clinical wisdom and experience are acquired only by following patients and solving clinical problems over a long period of time on a daily basis; there is no substitute for this [1, 8, 50, 54].

H. Boudoulas is an editorial board member of “Cardiology.”

The authors have no funding sources to declare.

K.D. Boudoulas and H. Boudoulas were involved in conception, design, obtaining and interpretation of data, and writing including the original draft of the manuscript. F. Triposkiadis, R. Gumina, D. Addison, and C. Iliescu were involved in conception, interpretation of data, and critical revision of the manuscript. All the authors completed, read, and approved the final manuscript.