点击显示 收起
1 From the Division of Nephrology and Hypertension, Harbor-UCLA Medical Center and Los Angeles Biomedical Research Institute, Los Angeles, CA; The David Geffen School of Medicine at the University of California, Los Angeles (UCLA); and the UCLA School of Public Health, Los Angeles, CA
2 Presented at the 2004 American Society for Clinical Nutrition Annual Meeting, April 17–21. 2004, Washington, DC.
3 Presentation of the Robert H Herman Memorial Award in Clinical Nutrition is supported by ASCN.
4 Supported by NIH grants R01 D061389-01A1, 5 R01 DK54457-04, and M01-RR00425.
5 Address reprint requests to JD Kopple, Division of Nephrology and Hypertension, Harbor-UCLA Medical Center, 1124 West Carson Street, C1 Annex, Torrance, CA 90502.
Robert H Herman Memorial Award in Clinical Nutrition Lecture, 2004
ABSTRACT
This lecture reviews a recently described phenomenon in patients with advanced chronic renal failure who are undergoing maintenance hemodialysis or chronic peritoneal dialysis. The phenomenon is called risk factor reversal, reverse epidemiology, or altered risk factor patterns, and it has to do with altered relations between risk factors and the hazard ratio for morbidity or mortality in these persons. This risk factor reversal phenomenon has been reported for body weight-for-height measures, systolic and diastolic blood pressures, and serum total cholesterol, LDL-cholesterol, homocysteine, creatinine, and parathyroid hormone concentrations, as well as metabolic acidemia. These risk factors are often associated with cardiovascular morbidity or mortality and with total mortality. The relations between these risk factors and the hazard ratio for morbidity or mortality vary from major alterations from the relations found in the general population (eg, for systolic or diastolic hypertension versus the hazard ratio of mortality) to a complete, mirror-image reversal (eg, that for body mass index versus the hazard ratio of mortality). Several potential causes of altered risk factor patterns are discussed here, and it is suggested that the major cause is the confounding effects of protein-energy malnutrition and inflammatory disorders, which commonly occur in maintenance dialysis patients.
Key Words: Maintenance hemodialysis • chronic peritoneal dialysis • chronic kidney disease • chronic renal failure • paradoxical risk factors • reverse epidemiology • malnutrition • inflammation • oxidant stress • carbonyl stress
INTRODUCTION
It is a particular honor for me to receive the Robert H Herman Memorial Award in Clinical Nutrition, because I knew Bob Herman, and I had great respect and admiration for him. When I was a young academician, Bob was already an established, highly respected authority in the field of clinical nutrition and was the editor of the American Journal of Clinical Nutrition. I had never worked with Bob, and in fact I lived at the opposite end of the country. However, he went out of his way to help me publish several proceedings of meetings in the Journal. He treated me with humanity, friendship, and respect. I believe that Bob did this because helping people was the natural way that he lived. Thus, this award constitutes a great honor for me and also represents a nostalgic connection to a person with whom I had developed a friendship over the years and whose ethics and humanity I greatly respected.
ALTERED RISK FACTOR RELATIONS AND REVERSE EPIDEMIOLOGY AND EXAMPLES
In this lecture, I discuss a line of research in which my colleague Kamyar Kalantar-Zadeh and I are currently involved, and which we and others refer to as risk factor reversal, paradoxical risk factors, or reverse epidemiology (1-3). These terms refer to the alterations in the normal relation between risk factors and clinical outcomes; these alterations are often dramatic and occur in certain groups of people. The abnormal relations may be so severe that they can result in more or less the exact opposite or reversal of the usual association between a risk factor and outcome that is found in the general population. Such risk factor reversal is commonly observed in persons with advanced chronic renal failure (CRF) who are undergoing maintenance dialysis therapy. Risk factors for which an alteration in the relation to morbidity or mortality has been reported in maintenance dialysis patients are listed in Table 1.
View this table:
TABLE 1. Risk factors that display altered or reversed relations to morbidity or mortality in maintenance hemodialysis patients1
An example of risk factor reversal is the relation between body weight and the hazard ratio (HR) of mortality. In the general population, the mortality rate decreases as the body weight–for-height or the body mass index (BMI) is reduced (4-8). A "J" or "U" curve relation has been described in some studies of normal persons, whereby mortality rates begin to rise as the BMI (in kg/m2) drops below 19–22 (4, 6-8). In other studies of normal persons, the mortality rate continues to fall as the BMI decreases to 19 or lower (6-8). If the effects of cigarette smoking, the lifestyle of many cigarette smokers, and the weight-lowering effects of illnesses, including subclinical diseases, are excluded, leanness as a risk factor for mortality often disappears (6).
In maintenance hemodialysis (MHD) patients, in contrast to the above observations, the relation between body weight–for-height (Figure 1) or BMI (Figure 2) and mortality represents the reverse, and often the mirror image, of the normal association (1, 9-17). This reversed relation has been reported in virtually all epidemiologic studies in MHD patients of the relation between body weight-for-height and mortality (1, 9-17), although some exceptions to this general rule are found, particularly in Asian patients (13, 18). The inverse relation between weight-for-height and mortality was observed in studies from both the United States and Europe (16), and it is particularly strong when the comparison is limited to MHD patients whose body weight-for-height ranges from approximately normal to low values (1, 11). However, the trend toward greater survival with increasing weight-for-height or BMI in MHD patients continues into the obese range, perhaps particularly for unadjusted comparisons (Figures 1 and 2; 10, 11, 19). A continued decrease in the risk factor ratio for mortality has even been observed in MHD patients with BMIs 45 (19). The direct relation between obesity and survival in MHD patients persists over such a large range of body weights that unadjusted survival rates are greater in MHD patients with a BMI 45 than in those with a BMI 25–27.5 (19). A reversal of the normal weight-for-height versus mortality relation has also been observed among chronic peritoneal dialysis (CPD) patients (20-23).
FIGURE 1.. Odds ratio (OR) of death among combined adult male and female maintenance hemodialysis patients grouped by weight-for-height percentile. Unadjusted OR, ; OR adjusted for clinical characteristic variables (ie, age, sex, race, and diabetes mellitus), ; OR adjusted for clinical characteristics and laboratory variables (ie, predialysis serum albumin, creatinine, and cholesterol and the urea reduction ratio), . Significantly different from the reference group: *P < 0.001, **P = 0.11, P = 0.21, #P = 0.038. Reprinted with permission from Kopple et al (11).
FIGURE 2.. Reverse epidemiology of mortality risk factors in maintenance hemodialysis (MHD) patients. Comparison of the effects of BMI on all-cause mortality between the general population and an MHD population. *The follow-up period was 14 y for the general population and 4 y for the MHD patients. **The BMI stratifications differ between the 2 populations: the values on the x-axis are based on the original graph of the general population, and the value for the original MHD BMI subgroups are printed along the hemodialysis curve. Reprinted with permission from Kalantar-Zadeh et al (1).
An alteration in the relation between serum total cholesterol and the HR of mortality has also been described for MHD and CPD patients (Table 2) (9, 24-28). In these various studies, the serum total cholesterol concentration associated with the lowest HR of death in MHD or CPD patients is 160–200 mg/dL. This stands in contrast to the relation of low serum cholesterol to improved survival that exists in the general population (29, 30). Only one study in maintenance dialysis patients reported an association between elevated serum cholesterol and increased mortality, and that study was conducted in 190 CPD patients with a mean follow-up of 12 mo (31). Recently Kalantar et al (32) found that low serum LDL-cholesterol concentrations also were associated with an increased HR of death in MHD patients. On the other hand, increased serum lipoprotein(a) is associated with decreased survival in maintenance dialysis patients, as it is in the general population (33).
View this table:
TABLE 2. Relative risk of death in maintenance hemodialysis patients according to Cox proportional hazard model for selected variables, including serum cholesterol and serum creatinine1
The relation between systolic blood pressure and the HR of death is also abnormal in MHD patients. In the general population, the data clearly indicate that hypertension is associated with an increased risk of adverse cardiovascular and cerebrovascular events and mortality due to those events (34-36). Indeed, high-normal blood pressure also increases the risk for adverse cardiovascular events (36). In maintenance dialysis patients, some studies show a similar relation between hypertension and increased mortality risk (37-39). However, hypertension, which often has been poorly controlled in MHD patients (40, 41), may not be a primary risk factor for survival in MHD patients (42-44). Several surveys in large populations of MHD patients either did not find high blood pressure to be an independent risk factor for cardiovascular disease (CVD) or for mortality or found lower blood pressures not to be protective of mortality (44-47).
Some studies found that systolic and diastolic hypertension posed little or no increased risk for CVD or for mortality or that it actually lowered the risk for these adverse events (42, 46-49). These relations were examined with blood pressures obtained immediately before a hemodialysis treatment (ie, when patients’ total body sodium, chloride, and water are increased, which tends to engender higher blood pressures) or after a hemodialysis session (ie, when patients have had excess sodium, chloride, and water removed, which tend to promote lower blood pressures). The systolic blood pressures associated with the lowest HR of death were 120–180 mm Hg before dialysis and 120–169 mm Hg after dialysis (42, 46-48, 50).
Elevated plasma total homocysteine (tHcy) concentrations are associated with an increased relative risk of death in the general population (51, 52). This is consistent with a large body of experimental evidence indicating that homocysteine is toxic to vascular endothelium, promotes atherosclerosis, and predisposes to arterial thrombosis (53-55). Several studies in CRF patients not on dialysis or in maintenance dialysis patients show a similar relation between higher plasma tHcy and the risk of CVD (56-61). On the other hand, Bostrom et al (62) found no relation between plasma homocysteine and the prevalence of CVD in maintenance dialysis patients, and several studies showed a reverse relation between plasma tHcy concentrations and the risk of CVD or cardiovascular mortality (63-66) (Figure 3). Kalantar et al (66) found that MHD patients who were in the higher quartiles of homocysteine concentrations tended to have the lowest morbidity and mortality, whereas those in the lowest quartile of serum homocysteine concentrations clearly had the highest relative risk of morbidity and mortality, including cardiovascular death. This inverse relation between plasma homocysteine and relative risk of death persisted even after multivariate adjustment for other risk factors associated with CVD such as age, diabetes mellitus, sex, race, ethnicity, and serum albumin concentrations (66). It is noteworthy that plasma tHcy is elevated in 85–95% of MHD patients, usually to 1.5–2.5 times the upper limit of normal serum concentrations.
FIGURE 3.. Cumulative proportion of surviving patients according to quartiles of plasma total homocysteine in 367 maintenance hemodialysis (MHD) patients. Reprinted with permission from Kalantar-Zadeh et al (66).
Because almost all maintenance dialysis patients have high plasma tHcy, and because these persons may also be at increased risk for CVD, the reverse risk factor findings regarding homocysteine in maintenance dialysis patients are, strictly speaking, not necessarily in conflict with the findings in the general population. The United States Veterans Administration is currently conducting a large, multicenter, randomized, prospective, double-blind clinical trial in nondialyzed CRF and maintenance dialysis patients on the effect of homocysteine-lowering therapy on morbidity and mortality. Patients randomly assigned to the treatment group receive large doses of folic acid, pyridoxine HCl, and vitamin B-12.
Similar reverse epidemiologic observations have been made for serum creatinine and parathyroid hormone (PTH). These studies show that, in MHD patients, the relation between the measure and outcome is counterintuitive. Thus, serum concentrations of creatinine, a metabolic product that increases in serum as renal function falls and that has been reported in some in vitro studies to have toxic effects (67, 68), are often 8–15 times normal in both MHD and CPD patients. People with a modest rise in serum creatinine (ie, with mild to moderate renal insufficiency) are at increased risk of morbidity and mortality, including adverse cardiovascular events (69-71). In contrast, in MHD patients, serum creatinine concentrations in blood obtained before the onset of a hemodialysis treatment or in the morning (for peritoneal dialysis patients) are inversely associated with the relative risk of mortality (24, 25, 27, 31, 72, 73). Patients with higher serum creatinine concentrations have a lower mortality rate (Table 2). These findings are particularly counterintuitive because higher serum creatinine can be caused by a reduced dialysis dose, which, below certain amounts, is considered to lead to greater uremic poisoning of the patient.
An elevated serum PTH concentration, which is a common phenomenon in nondialyzed CRF and maintenance dialysis patients, is considered to be a uremic toxin that may have pervasive adverse effects (74). Hyperphosphatemia occurs frequently in renal failure and is a cause of hyperparathyroidism (75). Increased serum phosphorus is reported to be a risk factor for higher morbidity and mortality in MHD patients (24, 31, 75-77). However, 2 epidemiologic studies in MHD and CPD patients show that those persons whose serum PTH concentrations are less elevated (but still may be greater than normal) have higher mortality rates (77, 78).
POSSIBLE CAUSES OF PARADOXICAL RISK FACTORS OR EPIDEMIOLOGY RISK
Protein-energy malnutrition-inflammation complex and oxidant and carbonyl stress
What are the causes for these reversed epidemiologic relations? We hypothesize that the combination of protein-energy malnutrition (PEM), chronic inflammatory states, and oxidant and possibly carbonyl stress may account for a substantial amount of these phenomena. Let us discuss these interrelated conditions in turn.
PEM is a common complication of advanced CRF that occurs in 40% of MHD and CPD patients (79-81). Evidence for PEM and the manifestations of this condition in CRF patients are summarized in Table 3. PEM has been epidemiologically linked to mortality in maintenance dialysis patients (82). There are many causes for PEM in CRF patients. These have been extensively reviewed and are summarized in Table 4 (79). The most important of these causes probably are low nutrient intake due to anorexia and intercurrent illnesses, nutrient losses during the dialysis procedure, the catabolic effects of coexistent acute or chronic catabolic illnesses—in the United States, prevalent maintenance dialysis patients are hospitalized, on average, 2.0 times/y (83)—and the catabolic effect of oxidative and carbonyl stress and chronic inflammation, which are common complications of CRF (79-81, 84, 85), as discussed below.
View this table:
TABLE 3. Evidence for protein-energy malnutrition in chronic renal failure1
View this table:
TABLE 4. Causes of protein-energy malnutrition in patients with chronic renal failure
CRF and maintenance dialysis patients frequently show evidence of inflammation and oxidant and carbonyl stress. This has been epidemiologically linked to atherosclerosis and cardiovascular morbidity and mortality (2, 3, 86-89). Evidence for inflammation in these patients includes increased serum concentrations of such acute phase proteins as C-reactive protein (CRP) and serum amyloid A and decreased serum concentrations of negative acute phase proteins, including albumin, transferrin, transthyretin (prealbumin), and cholesterol-carrying lipoproteins (2, 3, 86-89). Most surveys suggest that serum CRP concentrations are higher than normal values in 30–50% of American and European MHD patients and perhaps less commonly in Asian MHD patients (89). Serum concentrations of such proinflammatory cytokines as tumor necrosis factor , interleukin 1, and interleukin 6 commonly are elevated in patients with advanced CRF (89-94). In patients with advanced CRF and maintenance dialysis patients, there is also an accumulation of compounds that cause oxidant or carbonyl stress (84, 85, 89). Oxidant stress refers to cellular injury caused by exposure of the cell to compounds that oxidize chemicals in the cell (84). Carbonyl stress refers to cellular injury caused by carbon-containing compounds that chemically react with compounds in the body (85). Homocysteine is such a carbonyl-reactive compound that, as indicated above, is increased in plasma in CRF (56-66) and that, when elevated, exerts a number of adverse effects on the vascular endothelium (53-55).
Causes of inflammation and oxidant or carbonyl stress in persons with chronic kidney disease are described in Table 5; they include comorbid illnesses, which are commonly present; CRF, which itself leads to an increase in serum concentrations of a number of proinflammatory cytokines; chronic low-grade infection such as is caused by chlamydia; and reaction to vascular access protheses, the hemodialyzer itself, the peritoneal dialysis catheter (for CPD patients), and (for MHD or CPD patients) dialyzer tubing or impure dialysate (2, 3, 89-94).
View this table:
TABLE 5. Potential causes of inflammation and oxidant or carbonyl stress in chronic renal failure1
As implied in the foregoing discussion, there is an overlap between the manifestations of PEM and of inflammation. Thus, both PEM and inflammation may reduce serum concentrations of such negative acute phase proteins as albumin, transferrin, prealbumin, and cholesterol-carrying lipoproteins (79-81, 86, 89, 92). The lowest serum concentrations of these proteins are generally found with inflammation rather than with PEM. Inflammation may cause PEM by inducing anorexia—for example, tumor necrosis factor and interleukin 6 are anorexigens—and also by engendering a hypercatabolic state (89, 92, 95-97).
Because oxidant and carbonyl stress may cause or be associated with inflammation, there has been a tendency to lump their clinical manifestations together under the rubric of inflammation. That policy will be followed here for purposes of discussion. The occurrence of measures indicating the presence of either PEM or inflammation in CRF or maintenance dialysis patients has become a matter of great concern to nephrology researchers and clinicians because of the strong association between PEM or inflammation and an increased HR of morbidity and mortality (27, 82, 86, 89, 98). This issue is of particular importance because the adjusted mortality rate for prevalent maintenance dialysis patients in the United States is very high—24%/y (83). Hence, in a population that is already at high risk of morbidity and mortality, the identification of clinical characteristics indicating a subgroup of these persons who are at even greater risk for these adverse outcomes is a cause for alarm, and, at the same time, it represents a potential opportunity to develop interventions that may improve such poor prognoses.
As indicated above, the relative contributions of PEM and inflammation to the high morbidity and mortality of CRF patients are controversial, particularly because the syndrome of PEM shares many clinical manifestations with inflammation. Because inflammatory processes may cause endothelial injury and predispose a person to atherosclerosis and vascular thrombosis, it is easy to perceive why there would be a causal connection between inflammation and morbidity and mortality from vascular disease.
These considerations have led some investigators to question whether PEM by itself is hazardous or whether it is only an important risk factor for morbidity and mortality when it occurs in association with inflammation (99). Intuitively, it seems that a nutrient intake that is inadequate to maintain a healthy quantity of body protein mass or that does not provide adequate energy must eventually place the patient at risk of increased morbidity and mortality. I suspect that PEM may, among its other adverse consequences, predispose persons to inflammation and vascular disease, possibly by increasing the risk of infection or enhancing the inflammatory response to other stimuli. This is a question that demands further investigation. In any case, PEM and inflammation occur together so commonly in end-stage renal disease patients that some investigators have described them as components of a single syndrome referred to as the malnutrition-inflammation syndrome (2, 28, 98).
How may PEM and inflammation contribute to the abnormal epidemiologic relations observed in CRF patients? We hypothesize that at least part of the mechanism is that people who have PEM or an inflammatory state are much more likely to have superimposed illnesses, which themselves may increase the risk for morbid events or mortality. Also, as indicated above, PEM and inflammation not only may be manifestations of life-threatening illnesses but also may, in themselves, predispose a person to morbidity and fatality. Thus, MHD patients with decreased appetite; reduced dietary protein intake; low weight-for-height; low serum albumin, prealbumin, homocysteine, and cholesterol concentrations; increased serum CRP; or other manifestations of PEM or inflammation are more likely to die.
In summary, markers of PEM or inflammation may indicate a high HR of death in maintenance dialysis patients, and some of these measures may be quite different or even the reverse of those markers that indicate an increased HR of mortality in the general population. Indeed, the measures indicating an increased HR of mortality in the general population are often associated with excessive nutrient intake, obesity, and elements of the metabolic syndrome (100, 101) rather than with PEM. These markers include hypercholesterolemia, hypertension, and hyperhomocysteinemia. Although obesity itself may promote an inflammatory state (102, 103), people who have excess body fat are less likely than are people without excess body fat to have PEM and, possibly, other diseases that may engender PEM or inflammation. In CRF and maintenance dialysis patients, lower blood pressure might also represent a mortality risk factor because low blood pressure may be associated with cardiac or vascular disease.
It is possible that the standard risk factors for morbidity and mortality in the general population—such as obesity, increased serum LDL-cholesterol and homocysteine concentrations, and hypertension—may still be operative in maintenance dialysis patients. However, the risk of adverse consequences from these measures, in the short term, might be superceded by pathophysiological conditions that cause morbidity and mortality more rapidly and possibly more frequently. Moreover, these more rapidly acting pathophysiological conditions are more likely to be present when measures of PEM and inflammation are detected.
The data of Liu et al (28) are relevant in this regard. These authors described, in MHD patients, the same increased association between low serum total cholesterol and mortality that we and others have reported. However, Liu et al also restricted their analysis to MHD patients who had no evidence of the malnutrition-inflammation syndrome (28). These persons had a serum albumin concentration of 3.6 g/dL and serum CRP and interleukin 6 concentrations of <10 mg/dL and 3.09 pg/mL, respectively, and presumably were more likely to be healthier. In these latter MHD patients, the relation between serum total cholesterol and mortality was similar to that in the general population.
It may be helpful to distinguish between 2 types of risk factors for mortality. Some risk factors do not themselves appear to affect the likelihood of fatality but reflect the presence of factors that do increase mortality in MHD patients (eg, low serum albumin or prealbumin concentrations and high serum creatinine concentrations). Some other risk factors appear to directly increase the risk of mortality. Examples of this latter type of risk factor include elevated serum LDL-cholesterol (104), homocysteine (51-55), and CRP (105) concentrations and high BMI (4, 6-8). As indicated above, these risk factors are associated with higher mortality in the general population and, except for serum CRP, with lower mortality in MHD patients. However, it would appear unlikely that the biochemistry or metabolism of MHD patients has changed so much that the adverse effects of these latter risk factors are no longer operative. Hence, an advantage of the hypothesis proposed above is that it is consistent with the thesis that the risk factors for mortality—particularly cardiovascular mortality—in the general population are still operative in maintenance dialysis patients. However, it suggests that the more typical risk factors tend to be superseded, at least in the short term, by the risk factors associated with the malnutrition-inflammation complex syndrome.
The cause of the unanticipated relation between lower serum PTH and increased mortality might be explained by the malnutrition-inflammation syndrome. Low intakes of calcium, phosphorus, and protein and low serum phosphorus, which may all be associated with malnutrition or an inflammatory state (or both), may account for this relation (79-81). Reduced intakes of these substances might lead to lower serum PTH concentrations and, directly or as a result of associated diseases, might induce higher mortality.
Other potential causes for altered risk factor relations
Another potential explanation for the altered risk factor relations in MHD patients is that persons with chronic kidney disease who are not yet in need of chronic dialysis therapy are at increased risk of mortality, particularly that due to CVD (69-71). It has been suggested that the deaths of many of these persons with chronic kidney disease led to a selected population of survivors receiving maintenance dialysis treatment who do not respond normally in terms of risk factors. For example, they may be less vulnerable to the injurious effects on the vascular endothelium of elevated serum LDL-cholesterol and homocysteine concentrations or of obesity.
Another hypothesis is that a risk factor that is usually hazardous may also be protective in some circumstances. For example, it has been postulated that higher serum lipoprotein concentrations, as reflected by higher serum cholesterol concentrations, might bind circulating lipopolysaccharides (eg, endotoxin) more effectively (106). Such an explanation might account for the observation that morbidly obese MHD patients have greater probability of survival than do overweight MHD patients, such as those with BMIs 25–28 (19, 107). This observation cannot be readily explained by PEM or greater inflammation in the less obese group. Neither the overweight nor the morbidly obese MHD patients should have PEM, and the morbidly obese patients would, if anything, be expected to have a greater inflammatory disorder because of their greater obesity. A possible explanation for this paradox is that the inflammatory response to obesity may also have protective effects. Some of the cytokines elaborated in obese tissue may protect against inflammation and CVD (108, 109). This reversal of risk from obesity may be particularly likely to occur in persons with CRF who have not undergone dialysis and in MHD and CPD patients, because those groups tend to have both increased serum concentrations of proinflammatory cytokines that directly correlate with mortality and increased serum inhibitors of these cytokines and cytokines that correlate directly with survival (90, 91).
A given risk factor might have opposite and competing effects that can, at times, alter the relation between the risk factor and outcomes. An example may be the paradoxical relation between serum creatinine and survival in MHD patients. The observation that high serum creatinine concentrations, measured immediately before a hemodialysis treatment, predict increased survival might be explained as follows. Three risk factors, besides a low dialysis dose, may increase serum creatinine: they are a reduction in glomerular filtration rate, an increase in muscle mass, and an increase in the intake of meat (ie, striated muscle) (110). As indicated above, a drop in glomerular filtration rate, which in persons who have not undergone dialysis will increase the serum creatinine concentration, places a person at increased risk of CVD and cardiovascular mortality (69-71). However, MHD patients have such low glomerular filtration rates that variations in renal function among most MHD patients will have little effect on serum creatinine. On the other hand, an increase in the other 2 factors, muscle mass and protein (meat) intake, will each promote a higher serum creatinine. Moreover, higher muscle mass or protein intake will usually reflect better nutritional status, less inflammation, and a healthier clinical state and hence greater probability of survival.
For some risk factors, different epidemiologic studies have yielded discrepant results. For example, some studies show a positive correlation between plasma tHcy and CVD, and other studies describe a negative relation between these 2 variables. The explanation for these varied results is not clear. Publication bias might play a role; investigators whose data suggested a negative relation between tHcy and CVD may have been reluctant to publish their results because of the counterintuitive nature of their findings. In this regard, the studies showing elevated plasma homocysteine as a negative risk factor tend to be the more recent publications.
The phenomenon of risk factor reversal or abnormal risk factor relations is not limited to maintenance dialysis patients. It has also been observed in persons who either have other life-threatening chronic illnesses or are at increased risk of developing them. These other conditions or illnesses include advanced age (111), residence of an elderly person in a nursing home (112), hospitalization, particularly when elderly (113), congestive heart failure (114), and malignancy (115). In each of these conditions, PEM or an inflammatory state (or both) is likely to be present.
CLINICAL IMPLICATIONS OF ALTERED RISK FACTOR RELATIONS OR REVERSE EPIDEMIOLOGY
Will intervention to improve nutritional status or suppress inflammation reduce morbidity and mortality in MHD patients? This key question has not yet been addressed, although applications have recently been submitted to the National Institutes of Health for grants for research concerning this question. It has been shown in several small-scale, randomized, prospective clinical trials in MHD patients that nutritional interventions can decrease or eradicate indicators of PEM, particularly hypoalbuminemia (116, 117). Studies in MHD patients of whether antiinflammatory treatment will improve outcome have not yet been published, although some reports describe a reduced risk of progression of vascular disease in these persons with vitamin E (antioxidant) therapy (118).
The conundrum of paradoxical risk factors or reverse epidemiology in CRF and maintenance dialysis patients is a serious one. It calls into question the appropriate or target levels for some of the main measures used to clinically assess MHD patients, which include blood pressure, body weight-for-height, and serum concentrations of LDL-cholesterol, homocysteine, and PTH. For example, as a result of the epidemiologic data indicating the abnormal relations between these and other measures and mortality, the optimal blood pressure, body weight, and serum concentrations of the foregoing compounds have become uncertain. Randomized controlled interventional trials almost certainly will be necessary to ascertain the optimal levels for these measures in CRF and maintenance dialysis patients.
ACKNOWLEDGMENTS
I thank Kamyar Kalantar-Zadeh for the many discussions on reverse epidemiology, protein-energy malnutrition, and inflammation that he and I have enjoyed together and for his many insightful and stimulating comments.
The author had no personal or financial conflict of interest with regard to the subject matter of this manuscript.
REFERENCES