Proteinuria as a Predictor of Total Plasma Homocysteine Levels in Type 2 Diabetic Nephropathy

A detailed description of the IDNT cohort is found elsewhere (23). In brief, between March 1996 and February 1999, 1,715 individuals aged 29–78 years with documented type 2 diabetes, hypertension, and proteinuria were recruited from over 200 clinical centers in North and South America, Europe, and Asia. All subjects gave written informed consent. The average subject had type 2 diabetes for 15 years. The IDNT cohort was representative of the population with type 2 diabetic nephropathy. Resource limitations restricted this analysis to a convenience sample of 42% (717 subjects) of the original cohort, in whom complete baseline data as well as tHcy levels were available.

Baseline samples were drawn between 21 March 1996 and 25 February 1999 and stored at −70°C. tHcy and B vitamins were measured in the Vitamin Metabolism and Aging Laboratory (United States Department of Agriculture Human Nutrition Research Center, Boston, MA). Previous data have shown that plasma tHcy levels remain very stable under these storage conditions (24). Other laboratory examinations were performed in one of four IDNT regional laboratories. tHcy levels were determined by high-performance liquid chromatography with fluorescence detection (25). Plasma pyridoxal 5′-phosphate (PLP) (the active form of B6) levels were measured by radioenzymatic (tyrosine decarboxylase) assay (26), and plasma folate and B12 levels were ascertained by radioassay (Bio-Rad Quantaphase II; Bio-Rad, Hercules, CA). Serum albumin, creatinine, HbA_1c, lipid levels, and 24-h urinary protein levels were measured using standard automated clinical chemistry techniques.

The skewed outcome variable, tHcy, was logarithmically transformed, as were the independent variables proteinuria (and albuminuria), folate, PLP, and B12, to improve normality. Unadjusted correlations between potential predictor variables were performed using the Pearson correlation matrix. Forward and backward stepwise linear modeling with ANCOVA was then performed to determine the independent association between potential predictor variables (age, sex, BMI, serum creatinine, LDL and HDL cholesterol, triglycerides, urinary protein or albumin excretion, plasma folate, B12, PLP, serum albumin, and HbA_1c) and tHcy levels. Two serum creatinine levels were censored before statistical analysis (88 and 483 mg/dl, respectively). Statistical analyses were performed using SYSTAT (Version 9) software.

To ensure that the cohort studied was representative of the larger IDNT cohort, the two groups were compared in Table 1. There were no clinically significant differences between the cohorts with regard to demographic and physical attributes (age, sex, and BMI), major markers of nephropathy (serum creatinine and 24-h proteinuria), and metabolic derangements (total cholesterol and HbA_1c).

Descriptive data of the 717 subjects included are presented in Table 2. Subjects were between the ages of 29 and 78 years. The ranges of serum creatinine (0.7–5.5 mg/dl), proteinuria (4 mg to 39 g), and albuminuria (1 mg to 25 g) in this cohort were very wide. B vitamin indexes indicate that a segment of the population was deficient in vitamins B6, B12, and folate. A substantial portion of the cohort had tHcy levels that were elevated above the normal upper limit (i.e., <12 μmol/l). This upper limit applies to populations exposed to folate supplementation of foodstuffs, such as in the U.S. and Canada.

As seen in Table 3, significant unadjusted correlations were found between tHcy and age (r = 0.074; P = 0.008), creatinine (r = 0.414; P < 0.001), PLP (r = −0.105; P = 0.021), B12 (r = −0.216; P < 0.001), folate (r = −0.241; P < 0.001), and HbA_1c (r = −0.119; P = 0.003), with serum albumin approaching significance (r = 0.055; P = 0.072). There was no significant univariate correlation between proteinuria or albuminuria and tHcy levels.

Stepwise multiple regression modeling in the 610 subjects with available data that included all the independent variables in Table 2 found that only serum creatinine, plasma folate and B12, serum albumin, sex, HbA_1c, and age were independent predictors of tHcy after controlling for other factors (Table 4). Proteinuria (or albuminuria) never became significant with either forward or backward elimination modeling.

An analysis of a large patient cohort found no association between proteinuria (or albuminuria) and tHcy levels after controlling for all relevant confounding variables. In fact, proteinuria was not significantly associated with tHcy in either the unadjusted or multivariable analyses when using serum creatinine as a marker of renal function. Similar analyses using 24-h creatinine clearance did not change the results (data not shown). The cohort itself comprised a representative sample of type 2 diabetic patients with nephropathy, a population that encompassed a wide range of kidney disease and proteinuria. The results did confirm previously well-documented associations between tHcy and renal function (as measured by serum creatinine), serum albumin, plasma folate and B12, sex, age, and HbA_1c.

Previous reports found that albuminuria in patients with type 1 or type 2 diabetes (12–16), and even in individuals without diabetes (13), could predict tHcy levels. However, these studies were limited by numerous significant shortcomings. Ranges of serum creatinine and proteinuria were narrowly and arbitrarily restricted (12–14), sample sizes were often modest (15,16), and of particular importance, statistical analyses failed to control for all known major predictors of tHcy (e.g., renal function, B vitamins, albumin, age, and sex) (13–16). Finally, renal function was assessed by serum creatinine measurements (12–16). Persistently albuminuric patients have, by definition, suffered microvascular damage at the glomerular level (27). Because even early subtle declines in the glomerular filtration rate (GFR) (19,20) can raise tHcy levels, it is important to obtain accurate measurements of renal function when identifying factors that predict tHcy. Although serum creatinine is a simple and inexpensive marker of renal function, it is unfortunately not very good at detecting declines in GFR in early renal failure (28).

In contrast to these findings, other analyses performed in diabetic (20,21) and nondiabetic subjects (19,22) found no association between proteinuria and tHcy levels after controlling for all major confounding factors. Furthermore, most of these studies determined GFR by highly accurate clearance methods (19,20,22). However, external validity was compromised by their modest sample sizes and limited ranges of proteinuria (19–21) and renal function (21) measured. Of note, our analysis studied a much larger patient cohort that includes far broader ranges of serum creatinine and proteinuria. Furthermore, the study subjects were all enrolled in the well-designed multicenter IDNT cohort and were representative of the high-cardiovascular risk population of type 2 diabetic patients with nephropathy. These facts extend and improve the external validity of the previous negative studies (19–22). Therefore, it is unlikely that the positive association between proteinuria and cardiovascular events (3–6) is related to hyperhomocysteinemia.

The associations with tHcy found in this analysis have, in general, been well documented in the literature. Serum creatinine has traditionally been one of the major, if not primary, determinants of tHcy levels in populations with and without renal failure (29,30). This is related to its role as a renal function marker as well as its link to homocysteine production through creatine-creatinine synthesis via one-carbon metabolism (31). tHcy levels have been conclusively shown to be inversely proportional to GFR (19,20,22). Whether tHcy levels rise because the kidney itself is progressively unable to clear and metabolize homocysteine or whether accumulating uremic factors interfere with normal extrarenal homocysteine metabolism is not presently known (31).

By facilitating the normal metabolism of homocysteine, folate and vitamin B12 have been shown as two of the more important predictors of tHcy levels (32,33). The relatively low folate levels seen in some subjects is consistent with the fact that baseline measurements were performed in many individuals not exposed to folate supplementation of foodstuffs, as is currently done in the U.S. and Canada.

As a major plasma binder of tHcy, serum albumin is also a known determinant of tHcy (32). Advancing age is associated with increasing tHcy levels (32), possibly because of such factors as declining renal function, lower B vitamin status, or a higher prevalence of hypothyroidism (34).

For unclear reasons, women had slightly higher mean tHcy levels than men after controlling for other variables (men 13.7 μmol/l, women 14.4 μmol/l; P = 0.03). Differences in tHcy levels between men and women have previously been noted. One study of healthy subjects found that men had higher tHcy levels than women, presumably because homocysteine is linked to creatine/creatinine synthesis, which is typically greater in men (32). Another study in subjects with type 2 diabetes found tHcy levels to be higher in women who were postmenopausal and/or after a methionine load (21).

This study found that HbA_1c is inversely predictive of tHcy levels, although previous investigations have reached conflicting results with regard to glycemic markers in patients with diabetes. Some investigations have found that these markers are positively correlated with tHcy levels (35,36), others have found a negative correlation (37), and still others have found no significant relationship (15,38). The cause of the inverse relationship found here is not obvious. It is conceivable that as the complications of diabetes, such as nephropathy, manifest themselves to the patient and health care practitioner, greater attempts at glycemic control are made.

Finally, proteinuria has been found to be independently predictive of cardiovascular outcomes in a number of retrospective and prospective studies (4,7,8). Preliminary studies suggest that measuring the true GFR using highly accurate filtration markers may help to further stratify CVD risk among individuals with albuminuria (9–11).

The main funding for this research effort was provided by RO1 HL 67695-02, awarded to A.G.B. Support was also provided by the U.S. Department of Agriculture, under agreement 581950-9-001. The IDNT was supported by grants from Bristol-Myers Squibb and Sanofi-Synthelabo.

The authors wish to thank Marie Nadeau for her invaluable laboratory assistance.