Emilio D. Poggio, MD

Chronic kidney disease is most often discovered and diagnosed by primary care providers. The equations for estimating the glomerular filtration rate (GFR) facilitate earlier detection of this disease. However, the estimated GFR must be interpreted in the context of the individual patient. The diagnostic criteria and staging of chronic kidney disease must be understood so that it can be recognized and managed at the earliest possible stage. In this way, primary care physicians and nephrologists can better coordinate the care of these patients.

THE STAGES OF RENAL DISEASE AND THE GFR

Before 2002, an organized approach to the clinical management of patients with renal dysfunction was hampered by a lack of a standardized way to define this condition. This changed when the National Kidney Foundation, through the Kidney Disease Outcomes Quality Initiative (K/DOQI),¹ defined the stages of chronic kidney disease based on the GFR as estimated by the Modification of Diet in Renal Disease (MDRD) equation.^2,3

See related editorial

This system has increased the recognition of chronic kidney disease by the health care community and the general public. But the entire system hinges on the utility, accuracy, and reliability of the equations used to estimate the GFR.

In this article, we review the concepts of renal clearance and how to interpret the GFR in healthy patients and in those with chronic kidney disease. The following cases illustrate the interpretation of GFR in the context of patient care.

CASE 1: A 60-YEAR-OLD WOMAN WITH A ‘NORMAL’ CREATININE LEVEL

A 60-year-old white woman with no significant medical history has routine laboratory tests done as part of her annual physical examination. She weighs 135 pounds (61.2 kg) and is 64 inches (163 cm) tall. Her serum creatinine level is 1.1 mg/dL; her estimated GFR is 53 mL/min/1.73 m². A urine dipstick test for protein and blood is normal.

CASE 2: PROTEINURIA WITH A PRESERVED GFR

A 20-year-old African American man with no medical history is undergoing routine blood testing. His serum creatinine level is 1.1 mg/dL; his estimated GFR is reported as “> 60 mL/min/1.73 m²” (calculated at 109 mL/min/1.73 m²). He is 72 inches (183 cm) tall and weighs 180 pounds (83.0 kg); he lifts weights four times a week. Urine dipstick testing reveals 3+ proteinuria.

SERUM CREATININE: AN IMPERFECT MARKER OF KIDNEY FUNCTION

Of the various functions of the kidney, the ability of the glomeruli to filter the blood, as assessed by the GFR, is considered the best index of overall kidney function.^4,5 The GFR can be thought of as the clearance of a substance from the plasma by the kidney in a period of time. This is useful because no method is available to routinely and directly measure filtration across the glomerular basement membrane.

Substances that are cleared by the kidney are used to estimate the GFR. The ideal substance for this estimate is one that is cleared only by filtration and not through metabolism or excretion by other means.

The urinary clearance of the exogenous substance inulin is considered the gold standard method, but radioisotopes such as iothalamate and other markers have replaced inulin in clinical laboratories. Because these methods are expensive, time-consuming, and not widely available, alternative methods that use endogenous markers such as creatinine have been developed for clinical practice.

The serum creatinine concentration possesses many of the qualities of an ideal marker for estimating kidney function. Creatinine is produced by the body at a relatively constant rate under normal conditions and is easy and inexpensive to measure. However, it has several limitations:

• 

  Data presented in Rolin HA III, et al. Evaluation of glomerular filtration rate and renal plasma flow. In: Jacobson HR, et al, eds. The Principles and Practice of Nephrology. St. Louis: Mosby-Year Book 1995:8-13.

  Its clearance does not solely reflect glomerular filtration because the renal tubules also excrete it into the urine.⁶ As a result, creatinine clearance (see below) will tend to overestimate the GFR (Figure 1).
• The serum creatinine concentration is directly dependent on muscle mass, which varies with sex (women tend to have less muscle mass as a percent of body weight than men), age (muscle mass decreases with age), and race (African Americans have a higher serum creatinine level for the same GFR than other Americans).⁶ Thus, there is no “normal” value for serum creatinine that applies to all patients.
• 
  Other factors can alter the creatinine level without changing the GFR, such as changes in dietary protein intake, exercise, and drugs such as cimetidine⁷ and fibrates⁸ (Table 1).

Another important point is that the relationship between the serum creatinine concentration and the GFR is parabolic.⁹ At high kidney function, large changes in the GFR are reflected by very small changes in serum creatinine—the GFR must fall quite a bit before the serum creatinine level rises very much (points A to B in Figure 1). At lower kidney function, small changes in GFR are reflected by large changes in serum creatinine (points C to D in Figure 1). This phenomenon can cause physicians to view small changes in creatinine as unimportant in patients with creatinine levels in the normal or near-normal range. Conversely, small changes may be due to random error inherent in the methods of measuring creatinine rather than to changes in kidney function.

Because the serum creatinine concentration by itself may be misleading when estimating GFR, the National Kidney Foundation and the National Kidney Disease Education Program recommend that it not be used on its own to estimate kidney function.

 

ESTIMATING THE GFR

Measuring 24-hour creatinine clearance

Measuring 24-hour creatinine clearance involves measuring the concentrations of creatinine in the serum and the urine and the volume of urine excreted in 24 hours.

The 24-hour creatinine clearance was long considered the best alternative to the serum creatinine concentration for assessing kidney function, as it adjusts for changes in the creatinine concentration by taking into account creatinine’s excretion in the urine. However, 24-hour urine collection is burdensome for the patient, and the results are not always reliable because of variations in collection technique. Also, using the creatinine clearance does not resolve problems with using the serum creatinine concentration, such as tubular secretion and overestimation of GFR.

In an effort to more easily estimate GFR from blood tests alone, efforts to develop mathematical equations that more closely estimate GFR began over 40 years ago. These equations take into account factors such as age, sex, and ethnicity. The best known of these are the Cockcroft and Gault¹⁰ and the MDRD equations.⁵

The Cockcroft-Gault equation

The Cockcroft-Gault equation is fairly simple, using serum creatinine, ideal body weight, and an adjustment factor for sex. Its main drawbacks are that it was developed to model creatinine clearance, itself an imperfect estimation of GFR, and it depends heavily on the accuracy of the value for “lean” body weight used in the equation.

The MDRD equation

The MDRD equation has now largely replaced the Cockcroft-Gault equation. Developed using iothalamate GFR measurements, it therefore estimates GFR rather than the less-accurate creatinine clearance. Also, it is normalized to a standard body surface area (1.73 m²), obviating the need to determine ideal body weight.

Since the estimated GFR can often be calculated using data available in most electronic medical record systems, it can be reported directly with any laboratory report that includes a serum creatinine value.

The main drawback of the MDRD equation is that it tends to underestimate GFR at higher ranges of kidney function, ie, higher than 60 mL/min/1.73 m²).^3,11

The CKD-EPI equation

The Chronic Kidney Disease Epidemiology Collaboration study (CKD-EPI) equation,¹² published in 2009, is expected to eventually replace the currently used MDRD equation, as it performs better at higher ranges of GFR.

Although the CKD-EPI equation still lacks precision and accuracy, it underestimates GFR to a lesser degree than the MDRD equation in patients with preserved renal function. Also, it was developed with the objective of reporting a specific value even when the estimated GFR is greater than 60 mL/min/1.73 m². (In contrast, when laboratories use the MDRD equation, the recommendation is to report any value above this level as “greater than 60 mL/min/1.73 m²”).

A limitation of all equations that use the serum creatinine concentration to assess kidney function is the assumption that creatinine production is both stable over time and similar among patients. As a result, these equations should not be used in situations in which renal function is changing rapidly, such as in acute kidney injury. Also, they should be used with caution in patients at the extremes of body mass, since they underestimate GFR in very muscular patients (eg, as in case 2) and overestimate GFR in very small patients (eg, as in case 1).

Calculators for estimating the GFR using these equations are available on many Web sites (see www.kidney.org/professionals/kdoqi/gfr_calculator.cfm).

SCREEN EVERYONE AT RISK

As most patients with established medical problems have blood drawn periodically for routine chemistry panels, the diagnosis of chronic kidney disease often occurs through routine testing. For patients who do not yet carry this diagnosis, it is important to recognize the risk factors for chronic kidney disease (Table 2) and to determine who should be screened.

In general, anyone at higher risk of chronic kidney disease should be screened for it. This group includes US minorities and patients with hypertension, cardiovascular disease, and diabetes mellitus, among others.¹³ Screening includes an assessment of estimated GFR and urinalysis for proteinuria or hematuria.

CHRONIC KIDNEY DISEASE DEFINED: DAMAGE AND DURATION

The definition of chronic kidney disease contains two components—kidney damage and duration (Table 3).¹

The kidney damage can be either parenchymal renal damage independent of GFR (for example, cystic disease, glomerular hematuria, or proteinuria) or depressed GFR independent of evidence of parenchymal renal disease (an estimated GFR of less than 60 mL/min/1.73 m²).

The duration component requires that the abnormality be present for at least 3 months (ie, chronic).

 

 

Concerns about the definition

This definition has not been without controversy.

An unintended consequence of not reporting estimated GFR values above 60 mL/min/1.73 m² in absolute numbers is that providers may ignore changes in serum creatinine at estimated GFRs in this range, as they assume that the kidney function is “normal.” This may change in the future if the CKD-EPI equation is used, which produces less bias at slightly higher GFRs.

Providers may also tend to focus solely on the estimated GFR criterion and ignore other evidence of chronic kidney disease, such as abnormalities in urinalysis or imaging studies. For example, proteinuria has been shown to be more important than absolute GFR values in predicting progression of renal dysfunction and cardiovascular risk.¹⁴ Proteinuria, especially in the setting of an estimated GFR above 60 mL/min/1.73 m², can be missed if not screened for and underappreciated once found.

Moreover, in elderly patients, the current GFR equations underperform at borderline GFR values and can yield depressed values even at impressively “normal” serum creatinine levels. As a result, there is concern that chronic kidney disease is being overdiagnosed under the current system. This is especially worrisome in elderly white women without risk factors for chronic kidney disease (eg, as in case 1).

In addition, the question arises whether the arbitrary cutoff for chronic kidney disease— 60 mL/min/1.73 m²—applies to all populations.^15,16 The utility of classifying someone as having chronic kidney disease who has an estimated GFR of 55 mL/min/1.73 m² and no risk factors for chronic kidney disease (Table 2) should be questioned if the risk of progressing to end-stage renal disease or suffering a cardiovascular event is only minimally higher than in patients with a higher estimated GFR.^6,17 If the true purpose of developing the chronic kidney disease classification system is to improve patient care and outcomes, then it is of no benefit to overclassify such patients. Indeed, the stress induced by the diagnosis and the negative implications on insurance coverage and health care costs may outweigh any benefits.¹⁸

Nevertheless, these concerns do not invalidate the entire chronic kidney disease definition system, but have stimulated current efforts to improve it based on outcomes research.¹⁹

CASES REVISITED

Case 1: Problems with estimating GFR in a small woman

Case 1 has several points to note.

The patient’s small body size reflects low-level creatinine production. It is not atypical to find serum creatinine levels of 0.5 mg/dL in such patients. Thus, her serum creatinine level of 1.1 mg/dL may be abnormal. The fact that the MDRD equation “normalizes” the result to 1.73 m² of body surface area in patients with very low muscle mass will lead to an overestimation of GFR. However, she has no risk factors for chronic kidney disease.

Additionally, in up to two-thirds of patients kidney function declines with age.²⁰ Whether or not this is “normal aging” of the kidney, it is not clear that this decline in GFR reflects an underlying pathologic process.

Finally, since the patient is an older white woman, the estimated GFR tends to underestimate the true GFR. So while her body size may predispose to an overestimation of GFR, her age, race, and sex predispose to an underestimation of GFR. Many nephrologists would simply order urinalysis and ultrasonography to rule out other evidence of renal dysfunction, then recommend routine monitoring of kidney function in this case.

Case 2: Proteinuria is not normal

In case 2, because the patient is African American, young, and male, his creatinine level yields a higher estimated GFR than in case 1, despite having the same value. However, his estimated GFR still underestimates his true GFR because of his greater creatinine production due to his muscular physique.

This patient subsequently underwent iothalamate GFR testing, which yielded a GFR of 115 mL/min/1.73 m². However, he has dipstick-positive proteinuria, which, if confirmed on further testing, would meet the criteria for chronic kidney disease and put him at a higher risk of cardiovascular events and progression to lower kidney function than the patient in case 1. He also needs to be screened for undiagnosed hypertension and underlying glomerular disease.

REFERRAL TO (AND COLLABORATION WITH) A NEPHROLOGIST

Effective co-management with a nephrologist is essential for the overall health of the patient with chronic kidney disease, as well as slowing the progression to end-stage renal disease. Exactly when and to what extent the care of a patient with chronic kidney disease should be transferred to a nephrologist depends largely on the individual nephrologist and the comfort level of the primary care provider. When the referral does occur, effective communication between providers and a mutual understanding of the goals of care (eg, the blood pressure target) are essential to optimize patient care.

Chronic kidney disease is most often discovered and diagnosed by primary care providers. The equations for estimating the glomerular filtration rate (GFR) facilitate earlier detection of this disease. However, the estimated GFR must be interpreted in the context of the individual patient. The diagnostic criteria and staging of chronic kidney disease must be understood so that it can be recognized and managed at the earliest possible stage. In this way, primary care physicians and nephrologists can better coordinate the care of these patients.

THE STAGES OF RENAL DISEASE AND THE GFR

Before 2002, an organized approach to the clinical management of patients with renal dysfunction was hampered by a lack of a standardized way to define this condition. This changed when the National Kidney Foundation, through the Kidney Disease Outcomes Quality Initiative (K/DOQI),¹ defined the stages of chronic kidney disease based on the GFR as estimated by the Modification of Diet in Renal Disease (MDRD) equation.^2,3

See related editorial

This system has increased the recognition of chronic kidney disease by the health care community and the general public. But the entire system hinges on the utility, accuracy, and reliability of the equations used to estimate the GFR.

In this article, we review the concepts of renal clearance and how to interpret the GFR in healthy patients and in those with chronic kidney disease. The following cases illustrate the interpretation of GFR in the context of patient care.

CASE 1: A 60-YEAR-OLD WOMAN WITH A ‘NORMAL’ CREATININE LEVEL

A 60-year-old white woman with no significant medical history has routine laboratory tests done as part of her annual physical examination. She weighs 135 pounds (61.2 kg) and is 64 inches (163 cm) tall. Her serum creatinine level is 1.1 mg/dL; her estimated GFR is 53 mL/min/1.73 m². A urine dipstick test for protein and blood is normal.

CASE 2: PROTEINURIA WITH A PRESERVED GFR

A 20-year-old African American man with no medical history is undergoing routine blood testing. His serum creatinine level is 1.1 mg/dL; his estimated GFR is reported as “> 60 mL/min/1.73 m²” (calculated at 109 mL/min/1.73 m²). He is 72 inches (183 cm) tall and weighs 180 pounds (83.0 kg); he lifts weights four times a week. Urine dipstick testing reveals 3+ proteinuria.

SERUM CREATININE: AN IMPERFECT MARKER OF KIDNEY FUNCTION

Of the various functions of the kidney, the ability of the glomeruli to filter the blood, as assessed by the GFR, is considered the best index of overall kidney function.^4,5 The GFR can be thought of as the clearance of a substance from the plasma by the kidney in a period of time. This is useful because no method is available to routinely and directly measure filtration across the glomerular basement membrane.

Substances that are cleared by the kidney are used to estimate the GFR. The ideal substance for this estimate is one that is cleared only by filtration and not through metabolism or excretion by other means.

The urinary clearance of the exogenous substance inulin is considered the gold standard method, but radioisotopes such as iothalamate and other markers have replaced inulin in clinical laboratories. Because these methods are expensive, time-consuming, and not widely available, alternative methods that use endogenous markers such as creatinine have been developed for clinical practice.

The serum creatinine concentration possesses many of the qualities of an ideal marker for estimating kidney function. Creatinine is produced by the body at a relatively constant rate under normal conditions and is easy and inexpensive to measure. However, it has several limitations:

• 

  Data presented in Rolin HA III, et al. Evaluation of glomerular filtration rate and renal plasma flow. In: Jacobson HR, et al, eds. The Principles and Practice of Nephrology. St. Louis: Mosby-Year Book 1995:8-13.

  Its clearance does not solely reflect glomerular filtration because the renal tubules also excrete it into the urine.⁶ As a result, creatinine clearance (see below) will tend to overestimate the GFR (Figure 1).
• The serum creatinine concentration is directly dependent on muscle mass, which varies with sex (women tend to have less muscle mass as a percent of body weight than men), age (muscle mass decreases with age), and race (African Americans have a higher serum creatinine level for the same GFR than other Americans).⁶ Thus, there is no “normal” value for serum creatinine that applies to all patients.
• 
  Other factors can alter the creatinine level without changing the GFR, such as changes in dietary protein intake, exercise, and drugs such as cimetidine⁷ and fibrates⁸ (Table 1).

Another important point is that the relationship between the serum creatinine concentration and the GFR is parabolic.⁹ At high kidney function, large changes in the GFR are reflected by very small changes in serum creatinine—the GFR must fall quite a bit before the serum creatinine level rises very much (points A to B in Figure 1). At lower kidney function, small changes in GFR are reflected by large changes in serum creatinine (points C to D in Figure 1). This phenomenon can cause physicians to view small changes in creatinine as unimportant in patients with creatinine levels in the normal or near-normal range. Conversely, small changes may be due to random error inherent in the methods of measuring creatinine rather than to changes in kidney function.

Because the serum creatinine concentration by itself may be misleading when estimating GFR, the National Kidney Foundation and the National Kidney Disease Education Program recommend that it not be used on its own to estimate kidney function.

 

ESTIMATING THE GFR

Measuring 24-hour creatinine clearance

Measuring 24-hour creatinine clearance involves measuring the concentrations of creatinine in the serum and the urine and the volume of urine excreted in 24 hours.

The 24-hour creatinine clearance was long considered the best alternative to the serum creatinine concentration for assessing kidney function, as it adjusts for changes in the creatinine concentration by taking into account creatinine’s excretion in the urine. However, 24-hour urine collection is burdensome for the patient, and the results are not always reliable because of variations in collection technique. Also, using the creatinine clearance does not resolve problems with using the serum creatinine concentration, such as tubular secretion and overestimation of GFR.

In an effort to more easily estimate GFR from blood tests alone, efforts to develop mathematical equations that more closely estimate GFR began over 40 years ago. These equations take into account factors such as age, sex, and ethnicity. The best known of these are the Cockcroft and Gault¹⁰ and the MDRD equations.⁵

The Cockcroft-Gault equation

The Cockcroft-Gault equation is fairly simple, using serum creatinine, ideal body weight, and an adjustment factor for sex. Its main drawbacks are that it was developed to model creatinine clearance, itself an imperfect estimation of GFR, and it depends heavily on the accuracy of the value for “lean” body weight used in the equation.

The MDRD equation

The MDRD equation has now largely replaced the Cockcroft-Gault equation. Developed using iothalamate GFR measurements, it therefore estimates GFR rather than the less-accurate creatinine clearance. Also, it is normalized to a standard body surface area (1.73 m²), obviating the need to determine ideal body weight.

Since the estimated GFR can often be calculated using data available in most electronic medical record systems, it can be reported directly with any laboratory report that includes a serum creatinine value.

The main drawback of the MDRD equation is that it tends to underestimate GFR at higher ranges of kidney function, ie, higher than 60 mL/min/1.73 m²).^3,11

The CKD-EPI equation

The Chronic Kidney Disease Epidemiology Collaboration study (CKD-EPI) equation,¹² published in 2009, is expected to eventually replace the currently used MDRD equation, as it performs better at higher ranges of GFR.

Although the CKD-EPI equation still lacks precision and accuracy, it underestimates GFR to a lesser degree than the MDRD equation in patients with preserved renal function. Also, it was developed with the objective of reporting a specific value even when the estimated GFR is greater than 60 mL/min/1.73 m². (In contrast, when laboratories use the MDRD equation, the recommendation is to report any value above this level as “greater than 60 mL/min/1.73 m²”).

A limitation of all equations that use the serum creatinine concentration to assess kidney function is the assumption that creatinine production is both stable over time and similar among patients. As a result, these equations should not be used in situations in which renal function is changing rapidly, such as in acute kidney injury. Also, they should be used with caution in patients at the extremes of body mass, since they underestimate GFR in very muscular patients (eg, as in case 2) and overestimate GFR in very small patients (eg, as in case 1).

Calculators for estimating the GFR using these equations are available on many Web sites (see www.kidney.org/professionals/kdoqi/gfr_calculator.cfm).

SCREEN EVERYONE AT RISK

As most patients with established medical problems have blood drawn periodically for routine chemistry panels, the diagnosis of chronic kidney disease often occurs through routine testing. For patients who do not yet carry this diagnosis, it is important to recognize the risk factors for chronic kidney disease (Table 2) and to determine who should be screened.

In general, anyone at higher risk of chronic kidney disease should be screened for it. This group includes US minorities and patients with hypertension, cardiovascular disease, and diabetes mellitus, among others.¹³ Screening includes an assessment of estimated GFR and urinalysis for proteinuria or hematuria.

CHRONIC KIDNEY DISEASE DEFINED: DAMAGE AND DURATION

The definition of chronic kidney disease contains two components—kidney damage and duration (Table 3).¹

The kidney damage can be either parenchymal renal damage independent of GFR (for example, cystic disease, glomerular hematuria, or proteinuria) or depressed GFR independent of evidence of parenchymal renal disease (an estimated GFR of less than 60 mL/min/1.73 m²).

The duration component requires that the abnormality be present for at least 3 months (ie, chronic).

 

 

Concerns about the definition

This definition has not been without controversy.

An unintended consequence of not reporting estimated GFR values above 60 mL/min/1.73 m² in absolute numbers is that providers may ignore changes in serum creatinine at estimated GFRs in this range, as they assume that the kidney function is “normal.” This may change in the future if the CKD-EPI equation is used, which produces less bias at slightly higher GFRs.

Providers may also tend to focus solely on the estimated GFR criterion and ignore other evidence of chronic kidney disease, such as abnormalities in urinalysis or imaging studies. For example, proteinuria has been shown to be more important than absolute GFR values in predicting progression of renal dysfunction and cardiovascular risk.¹⁴ Proteinuria, especially in the setting of an estimated GFR above 60 mL/min/1.73 m², can be missed if not screened for and underappreciated once found.

Moreover, in elderly patients, the current GFR equations underperform at borderline GFR values and can yield depressed values even at impressively “normal” serum creatinine levels. As a result, there is concern that chronic kidney disease is being overdiagnosed under the current system. This is especially worrisome in elderly white women without risk factors for chronic kidney disease (eg, as in case 1).

In addition, the question arises whether the arbitrary cutoff for chronic kidney disease— 60 mL/min/1.73 m²—applies to all populations.^15,16 The utility of classifying someone as having chronic kidney disease who has an estimated GFR of 55 mL/min/1.73 m² and no risk factors for chronic kidney disease (Table 2) should be questioned if the risk of progressing to end-stage renal disease or suffering a cardiovascular event is only minimally higher than in patients with a higher estimated GFR.^6,17 If the true purpose of developing the chronic kidney disease classification system is to improve patient care and outcomes, then it is of no benefit to overclassify such patients. Indeed, the stress induced by the diagnosis and the negative implications on insurance coverage and health care costs may outweigh any benefits.¹⁸

Nevertheless, these concerns do not invalidate the entire chronic kidney disease definition system, but have stimulated current efforts to improve it based on outcomes research.¹⁹

CASES REVISITED

Case 1: Problems with estimating GFR in a small woman

Case 1 has several points to note.

The patient’s small body size reflects low-level creatinine production. It is not atypical to find serum creatinine levels of 0.5 mg/dL in such patients. Thus, her serum creatinine level of 1.1 mg/dL may be abnormal. The fact that the MDRD equation “normalizes” the result to 1.73 m² of body surface area in patients with very low muscle mass will lead to an overestimation of GFR. However, she has no risk factors for chronic kidney disease.

Additionally, in up to two-thirds of patients kidney function declines with age.²⁰ Whether or not this is “normal aging” of the kidney, it is not clear that this decline in GFR reflects an underlying pathologic process.

Finally, since the patient is an older white woman, the estimated GFR tends to underestimate the true GFR. So while her body size may predispose to an overestimation of GFR, her age, race, and sex predispose to an underestimation of GFR. Many nephrologists would simply order urinalysis and ultrasonography to rule out other evidence of renal dysfunction, then recommend routine monitoring of kidney function in this case.

Case 2: Proteinuria is not normal

In case 2, because the patient is African American, young, and male, his creatinine level yields a higher estimated GFR than in case 1, despite having the same value. However, his estimated GFR still underestimates his true GFR because of his greater creatinine production due to his muscular physique.

This patient subsequently underwent iothalamate GFR testing, which yielded a GFR of 115 mL/min/1.73 m². However, he has dipstick-positive proteinuria, which, if confirmed on further testing, would meet the criteria for chronic kidney disease and put him at a higher risk of cardiovascular events and progression to lower kidney function than the patient in case 1. He also needs to be screened for undiagnosed hypertension and underlying glomerular disease.

REFERRAL TO (AND COLLABORATION WITH) A NEPHROLOGIST

Effective co-management with a nephrologist is essential for the overall health of the patient with chronic kidney disease, as well as slowing the progression to end-stage renal disease. Exactly when and to what extent the care of a patient with chronic kidney disease should be transferred to a nephrologist depends largely on the individual nephrologist and the comfort level of the primary care provider. When the referral does occur, effective communication between providers and a mutual understanding of the goals of care (eg, the blood pressure target) are essential to optimize patient care.

Chronic kidney disease is most often discovered and diagnosed by primary care providers. The equations for estimating the glomerular filtration rate (GFR) facilitate earlier detection of this disease. However, the estimated GFR must be interpreted in the context of the individual patient. The diagnostic criteria and staging of chronic kidney disease must be understood so that it can be recognized and managed at the earliest possible stage. In this way, primary care physicians and nephrologists can better coordinate the care of these patients.

THE STAGES OF RENAL DISEASE AND THE GFR

Before 2002, an organized approach to the clinical management of patients with renal dysfunction was hampered by a lack of a standardized way to define this condition. This changed when the National Kidney Foundation, through the Kidney Disease Outcomes Quality Initiative (K/DOQI),¹ defined the stages of chronic kidney disease based on the GFR as estimated by the Modification of Diet in Renal Disease (MDRD) equation.^2,3

See related editorial

This system has increased the recognition of chronic kidney disease by the health care community and the general public. But the entire system hinges on the utility, accuracy, and reliability of the equations used to estimate the GFR.

In this article, we review the concepts of renal clearance and how to interpret the GFR in healthy patients and in those with chronic kidney disease. The following cases illustrate the interpretation of GFR in the context of patient care.

CASE 1: A 60-YEAR-OLD WOMAN WITH A ‘NORMAL’ CREATININE LEVEL

A 60-year-old white woman with no significant medical history has routine laboratory tests done as part of her annual physical examination. She weighs 135 pounds (61.2 kg) and is 64 inches (163 cm) tall. Her serum creatinine level is 1.1 mg/dL; her estimated GFR is 53 mL/min/1.73 m². A urine dipstick test for protein and blood is normal.

CASE 2: PROTEINURIA WITH A PRESERVED GFR

A 20-year-old African American man with no medical history is undergoing routine blood testing. His serum creatinine level is 1.1 mg/dL; his estimated GFR is reported as “> 60 mL/min/1.73 m²” (calculated at 109 mL/min/1.73 m²). He is 72 inches (183 cm) tall and weighs 180 pounds (83.0 kg); he lifts weights four times a week. Urine dipstick testing reveals 3+ proteinuria.

SERUM CREATININE: AN IMPERFECT MARKER OF KIDNEY FUNCTION

Of the various functions of the kidney, the ability of the glomeruli to filter the blood, as assessed by the GFR, is considered the best index of overall kidney function.^4,5 The GFR can be thought of as the clearance of a substance from the plasma by the kidney in a period of time. This is useful because no method is available to routinely and directly measure filtration across the glomerular basement membrane.

Substances that are cleared by the kidney are used to estimate the GFR. The ideal substance for this estimate is one that is cleared only by filtration and not through metabolism or excretion by other means.

The urinary clearance of the exogenous substance inulin is considered the gold standard method, but radioisotopes such as iothalamate and other markers have replaced inulin in clinical laboratories. Because these methods are expensive, time-consuming, and not widely available, alternative methods that use endogenous markers such as creatinine have been developed for clinical practice.

The serum creatinine concentration possesses many of the qualities of an ideal marker for estimating kidney function. Creatinine is produced by the body at a relatively constant rate under normal conditions and is easy and inexpensive to measure. However, it has several limitations:

• 

  Data presented in Rolin HA III, et al. Evaluation of glomerular filtration rate and renal plasma flow. In: Jacobson HR, et al, eds. The Principles and Practice of Nephrology. St. Louis: Mosby-Year Book 1995:8-13.

  Its clearance does not solely reflect glomerular filtration because the renal tubules also excrete it into the urine.⁶ As a result, creatinine clearance (see below) will tend to overestimate the GFR (Figure 1).
• The serum creatinine concentration is directly dependent on muscle mass, which varies with sex (women tend to have less muscle mass as a percent of body weight than men), age (muscle mass decreases with age), and race (African Americans have a higher serum creatinine level for the same GFR than other Americans).⁶ Thus, there is no “normal” value for serum creatinine that applies to all patients.
• 
  Other factors can alter the creatinine level without changing the GFR, such as changes in dietary protein intake, exercise, and drugs such as cimetidine⁷ and fibrates⁸ (Table 1).

Another important point is that the relationship between the serum creatinine concentration and the GFR is parabolic.⁹ At high kidney function, large changes in the GFR are reflected by very small changes in serum creatinine—the GFR must fall quite a bit before the serum creatinine level rises very much (points A to B in Figure 1). At lower kidney function, small changes in GFR are reflected by large changes in serum creatinine (points C to D in Figure 1). This phenomenon can cause physicians to view small changes in creatinine as unimportant in patients with creatinine levels in the normal or near-normal range. Conversely, small changes may be due to random error inherent in the methods of measuring creatinine rather than to changes in kidney function.

Because the serum creatinine concentration by itself may be misleading when estimating GFR, the National Kidney Foundation and the National Kidney Disease Education Program recommend that it not be used on its own to estimate kidney function.

 

ESTIMATING THE GFR

Measuring 24-hour creatinine clearance

Measuring 24-hour creatinine clearance involves measuring the concentrations of creatinine in the serum and the urine and the volume of urine excreted in 24 hours.

The 24-hour creatinine clearance was long considered the best alternative to the serum creatinine concentration for assessing kidney function, as it adjusts for changes in the creatinine concentration by taking into account creatinine’s excretion in the urine. However, 24-hour urine collection is burdensome for the patient, and the results are not always reliable because of variations in collection technique. Also, using the creatinine clearance does not resolve problems with using the serum creatinine concentration, such as tubular secretion and overestimation of GFR.

In an effort to more easily estimate GFR from blood tests alone, efforts to develop mathematical equations that more closely estimate GFR began over 40 years ago. These equations take into account factors such as age, sex, and ethnicity. The best known of these are the Cockcroft and Gault¹⁰ and the MDRD equations.⁵

The Cockcroft-Gault equation

The Cockcroft-Gault equation is fairly simple, using serum creatinine, ideal body weight, and an adjustment factor for sex. Its main drawbacks are that it was developed to model creatinine clearance, itself an imperfect estimation of GFR, and it depends heavily on the accuracy of the value for “lean” body weight used in the equation.

The MDRD equation

The MDRD equation has now largely replaced the Cockcroft-Gault equation. Developed using iothalamate GFR measurements, it therefore estimates GFR rather than the less-accurate creatinine clearance. Also, it is normalized to a standard body surface area (1.73 m²), obviating the need to determine ideal body weight.

Since the estimated GFR can often be calculated using data available in most electronic medical record systems, it can be reported directly with any laboratory report that includes a serum creatinine value.

The main drawback of the MDRD equation is that it tends to underestimate GFR at higher ranges of kidney function, ie, higher than 60 mL/min/1.73 m²).^3,11

The CKD-EPI equation

The Chronic Kidney Disease Epidemiology Collaboration study (CKD-EPI) equation,¹² published in 2009, is expected to eventually replace the currently used MDRD equation, as it performs better at higher ranges of GFR.

Although the CKD-EPI equation still lacks precision and accuracy, it underestimates GFR to a lesser degree than the MDRD equation in patients with preserved renal function. Also, it was developed with the objective of reporting a specific value even when the estimated GFR is greater than 60 mL/min/1.73 m². (In contrast, when laboratories use the MDRD equation, the recommendation is to report any value above this level as “greater than 60 mL/min/1.73 m²”).

A limitation of all equations that use the serum creatinine concentration to assess kidney function is the assumption that creatinine production is both stable over time and similar among patients. As a result, these equations should not be used in situations in which renal function is changing rapidly, such as in acute kidney injury. Also, they should be used with caution in patients at the extremes of body mass, since they underestimate GFR in very muscular patients (eg, as in case 2) and overestimate GFR in very small patients (eg, as in case 1).

Calculators for estimating the GFR using these equations are available on many Web sites (see www.kidney.org/professionals/kdoqi/gfr_calculator.cfm).

SCREEN EVERYONE AT RISK

As most patients with established medical problems have blood drawn periodically for routine chemistry panels, the diagnosis of chronic kidney disease often occurs through routine testing. For patients who do not yet carry this diagnosis, it is important to recognize the risk factors for chronic kidney disease (Table 2) and to determine who should be screened.

In general, anyone at higher risk of chronic kidney disease should be screened for it. This group includes US minorities and patients with hypertension, cardiovascular disease, and diabetes mellitus, among others.¹³ Screening includes an assessment of estimated GFR and urinalysis for proteinuria or hematuria.

CHRONIC KIDNEY DISEASE DEFINED: DAMAGE AND DURATION

The definition of chronic kidney disease contains two components—kidney damage and duration (Table 3).¹

The kidney damage can be either parenchymal renal damage independent of GFR (for example, cystic disease, glomerular hematuria, or proteinuria) or depressed GFR independent of evidence of parenchymal renal disease (an estimated GFR of less than 60 mL/min/1.73 m²).

The duration component requires that the abnormality be present for at least 3 months (ie, chronic).

 

 

Concerns about the definition

This definition has not been without controversy.

An unintended consequence of not reporting estimated GFR values above 60 mL/min/1.73 m² in absolute numbers is that providers may ignore changes in serum creatinine at estimated GFRs in this range, as they assume that the kidney function is “normal.” This may change in the future if the CKD-EPI equation is used, which produces less bias at slightly higher GFRs.

Providers may also tend to focus solely on the estimated GFR criterion and ignore other evidence of chronic kidney disease, such as abnormalities in urinalysis or imaging studies. For example, proteinuria has been shown to be more important than absolute GFR values in predicting progression of renal dysfunction and cardiovascular risk.¹⁴ Proteinuria, especially in the setting of an estimated GFR above 60 mL/min/1.73 m², can be missed if not screened for and underappreciated once found.

Moreover, in elderly patients, the current GFR equations underperform at borderline GFR values and can yield depressed values even at impressively “normal” serum creatinine levels. As a result, there is concern that chronic kidney disease is being overdiagnosed under the current system. This is especially worrisome in elderly white women without risk factors for chronic kidney disease (eg, as in case 1).

In addition, the question arises whether the arbitrary cutoff for chronic kidney disease— 60 mL/min/1.73 m²—applies to all populations.^15,16 The utility of classifying someone as having chronic kidney disease who has an estimated GFR of 55 mL/min/1.73 m² and no risk factors for chronic kidney disease (Table 2) should be questioned if the risk of progressing to end-stage renal disease or suffering a cardiovascular event is only minimally higher than in patients with a higher estimated GFR.^6,17 If the true purpose of developing the chronic kidney disease classification system is to improve patient care and outcomes, then it is of no benefit to overclassify such patients. Indeed, the stress induced by the diagnosis and the negative implications on insurance coverage and health care costs may outweigh any benefits.¹⁸

Nevertheless, these concerns do not invalidate the entire chronic kidney disease definition system, but have stimulated current efforts to improve it based on outcomes research.¹⁹

CASES REVISITED

Case 1: Problems with estimating GFR in a small woman

Case 1 has several points to note.

The patient’s small body size reflects low-level creatinine production. It is not atypical to find serum creatinine levels of 0.5 mg/dL in such patients. Thus, her serum creatinine level of 1.1 mg/dL may be abnormal. The fact that the MDRD equation “normalizes” the result to 1.73 m² of body surface area in patients with very low muscle mass will lead to an overestimation of GFR. However, she has no risk factors for chronic kidney disease.

Additionally, in up to two-thirds of patients kidney function declines with age.²⁰ Whether or not this is “normal aging” of the kidney, it is not clear that this decline in GFR reflects an underlying pathologic process.

Finally, since the patient is an older white woman, the estimated GFR tends to underestimate the true GFR. So while her body size may predispose to an overestimation of GFR, her age, race, and sex predispose to an underestimation of GFR. Many nephrologists would simply order urinalysis and ultrasonography to rule out other evidence of renal dysfunction, then recommend routine monitoring of kidney function in this case.

Case 2: Proteinuria is not normal

In case 2, because the patient is African American, young, and male, his creatinine level yields a higher estimated GFR than in case 1, despite having the same value. However, his estimated GFR still underestimates his true GFR because of his greater creatinine production due to his muscular physique.

This patient subsequently underwent iothalamate GFR testing, which yielded a GFR of 115 mL/min/1.73 m². However, he has dipstick-positive proteinuria, which, if confirmed on further testing, would meet the criteria for chronic kidney disease and put him at a higher risk of cardiovascular events and progression to lower kidney function than the patient in case 1. He also needs to be screened for undiagnosed hypertension and underlying glomerular disease.

REFERRAL TO (AND COLLABORATION WITH) A NEPHROLOGIST

Effective co-management with a nephrologist is essential for the overall health of the patient with chronic kidney disease, as well as slowing the progression to end-stage renal disease. Exactly when and to what extent the care of a patient with chronic kidney disease should be transferred to a nephrologist depends largely on the individual nephrologist and the comfort level of the primary care provider. When the referral does occur, effective communication between providers and a mutual understanding of the goals of care (eg, the blood pressure target) are essential to optimize patient care.

The use of gadolinium as a contrast agent in magnetic resonance imaging (MRI) in patients with impaired kidney function has come under scrutiny because of recent reports of a potential association between its use and nephrogenic systemic fibrosis (NSF).

See related editorial

This entity was first identified in the United States in 1997. Cowper et al¹ in 2000 described 15 hemodialysis patients who developed thickening and hardening of the skin with brawny hyperpigmentation, papules, and subcutaneous nodules on the extremities.

This “new disease” was initially called “nephrogenic fibrosing dermopathy,” as it was exclusively seen in patients with renal impairment and was thought to affect only the skin and subcutaneous tissue. With growing evidence of the extent and pathogenicity of the fibrosis in visceral organs, the nomenclature was changed to NSF, to better reflect the systemic nature of the disease.

PRESENTATION: MILD TO DEVASTATING

NSF has thus far been reported only in patients with renal impairment, most of whom were dialysis-dependent. It does not seem to be more common in one sex or the other, in any age range, or in any ethnic group. It can range in severity from mild to a devastating scleroderma-like systemic fibrosing disorder.

Cutaneous changes are the most predominant and impressive manifestations. NSF typically causes dermal hardening with tethering to deep dermal tissues, giving the skin the appearance of textured plaques, papules, or nodules with irregular edges and a brawny wooden texture to palpation (Figure 1). The lesions can be erythematous or brown-pigmented and can be painful and pruritic. NSF typically presents between the ankles and the thighs in a symmetric fashion and progresses proximally and distally to involve the entire lower extremities. Upper extremity involvement occurs frequently, but usually with lower extremity disease.² The trunk is involved less commonly than the legs and arms, and usually late in extensive disease. The face is typically spared (Figure 2).

NSF can cause loss of motion and contractures in multiple joints, leading to almost total loss of function and devastating debility within a short time—days to a few weeks.² These contractures are attributed to periarticular fibrosis of the overlying skin and subcutaneous tissue rather than to erosive joint disease. About 5% of patients develop a fulminant form of NSF³; these patients may become wheelchair-dependent.

The heart, lungs, skeletal muscle, and diaphragm can also be involved, sometimes leading to serious complications and death.^4–6

The disease is usually progressive and unremitting. Mendoza et al,⁷ in a review of 12 cases of NSF, reported that the disease had a progressive course in 6 patients, of whom 3 died within 2 years and 3 were ultimately confined to a wheelchair. More severe findings and rapid progression of the skin disease are associated with a poor prognosis.

Todd et al⁸ prospectively examined 186 dialysis patients to look for possible NSF. Of those with skin changes consistent with NSF, 48% died within 2 years, compared with 20% of those without these skin changes. Cardiovascular causes accounted for 58% of the deaths in patients with cutaneous changes of NSF and for 48% of the deaths in patients without these changes. Most of the excess deaths occurred within 6 months after the skin examination, suggesting an increased risk for early death in patients with skin changes suggestive of NSF.

DIAGNOSIS OF NSF IS CLINICAL

At presentation, NSF is frequently misdiagnosed and treated as cellulitis or edema. However, now that subspecialists—especially dermatologists, rheumatologists, and nephrologists—are becoming more aware of it, the correct diagnosis is being made earlier.

NSF should be suspected in any patient with underlying renal dysfunction—especially if on dialysis and if he or she has received a gadolinium contrast agent during MRI—who develops scleroderma-like cutaneous lesions affecting the distal extremities. Because most health care providers are still unfamiliar with this emerging disease, patients with renal impairment and suspected NSF should be referred to a rheumatologist or dermatologist to confirm the diagnosis, which is mainly entertained on a clinical basis. There is no laboratory biomarker for NSF.

A deep incisional skin biopsy may aid in the diagnosis. Due to the regional distribution of the disease, sampling error may occur, and repeat biopsy is warranted if the initial biopsy is nondiagnostic but the clinical picture suggests NSF.

 

Histopathologic examination typically shows lesions containing proliferation of dermal spindle cells, thick collagen bundles with surrounding clefts, and a variable amount of mucin and elastic fibers.² A characteristic and almost pathognomonic staining profile is the immunohistochemical identification of CD34 reactivity in the fibroblast-like cells (Figure 3). Cells expressing CD34 are normally found in the umbilical cord, the bone marrow (as pluripotential hematopoietic stem cells), and in the vascular endothelium. How they come to be in the skin is still speculative, but their presence suggests that circulating fibrocytes migrate from the bone marrow and deposit in the skin and other organs.^9,10

Pulmonary function testing can be done to rule out lung involvement and transthoracic two-dimensional echocardiography can be done to rule out possible cardiomyopathy if these conditions are suggested by examination at the time of diagnosis.⁷ Muscle biopsy is not necessary to determine the extent of systemic involvement, since the findings do not necessarily correlate with other systemic involvement.

 

DIFFERENTIAL DIAGNOSIS

Other disorders that can cause thickening and hardening of the skin of the extremities and trunk include systemic sclerosis or scleroderma, scleromyxedema, and eosinophilic fasciitis (Table 1). However, skin thickening, tethering, and hyperpigmentation in a patient with chronic kidney disease or end-stage renal disease after exposure to gadolinium-containing contrast agents suggests NSF.

An important diagnostic feature of NSF is that it spares the face, a finding derived from all reported and confirmed cases of NSF (Figure 2). In contrast, scleromyxedema, systemic scleroderma, and morphea often involve the face.

Scleromyxedema is often associated with monoclonal gammopathy (usually an immunoglobulin G lambda paraproteinemia) whereas NSF is not.

Scleroderma is supported by the findings of Raynaud’s phenomenon, antinuclear antibodies, and either anticentromere or anti-DNA topoisomerase I (Scl-70) antibodies, but the absence of these antibodies does not necessarily rule it out.

Eosinophilic fasciitis is diagnosed on the basis of histologic examination of a deep wedge skin biopsy specimen that includes fascia.

Other diagnoses that should be considered include amyloidosis and calciphylaxis.

ASSOCIATION WITH GADOLINIUM: WHAT IS THE EVIDENCE?

Case series

The association of gadolinium use with NSF has been described in several case reports and case series.

Grobner¹¹ reported that administration of gadodiamide (Omniscan, a gadolinium compound) for MRI was associated with NSF in five patients on chronic hemodialysis who had end-stage renal disease. Their ages ranged from 43 to 74 years, and they had been on dialysis from 10 to 58 months. The time of onset of NSF ranged from 2 to 4 weeks after exposure to gadodiamide.

Marckmann et al¹² reported that NSF developed in 13 (3.5%) of 370 patients with severe kidney disease who received gadodiamide. Five of the 13 patients had stage 5 (advanced) chronic kidney disease and were not yet on renal replacement therapy, 7 were on hemodialysis, and 1 was on peritoneal dialysis. The time of onset ranged from 2 to 75 days (median 25 days) after exposure.

Kuo et al¹³ similarly estimated the incidence of NSF at approximately 3% in patients with severe renal failure who receive intravenous gadolinium-based contrast material for MRI.

Broome et al¹⁴ reported that 12 patients developed NSF within 2 to 11 weeks after receiving gadodiamide. Eight of the 12 patients had end-stage renal disease and were on hemodialysis; the other 4 patients had acute kidney injury attributed to hepatorenal syndrome, and 3 of these 4 patients were on hemodialysis.

Khurana et al¹⁵ reported that 6 patients on hemodialysis developed NSF from 2 weeks to 2 months after receiving a dose of gadodiamide of between 0.11 and 0.36 mmol/kg. These doses are high, and the findings suggest an association between the gadolinium dose and NSF. The dose approved by the US Food and Drug Administration (FDA) is only 0.1 mmol/kg, and the use of gadolinium is approved only in MRI. However, higher doses (0.3–0.4 mmol/kg) are widely used in practice for better imaging quality in magnetic resonance angiography (MRA).

Deo et al¹⁶ reported 3 cases of NSF in 87 patients with end-stage renal disease who underwent 123 radiologic studies with gadolinium. No patient with end-stage renal disease who was not exposed to gadolinium developed NSF, and the association between exposure to gadolinium and the subsequent development of NSF was statistically significant (P = .006). The authors concluded that each gadolinium study presented a 2.4% risk of NSF in end-stage renal disease patients.

This retrospective study is flawed by not having been cross-sectional or case-controlled, since the other 84 patients who received gadolinium were not examined at all to establish the absence of NSF.

Case-control studies

More evidence of association of NSF with gadolinium exposure comes from other reports.

Physicians in St. Louis, MO,¹⁷ identified 33 cases of NSF and performed a case-control study, matching each of 19 of the patients (for whom data were available and who met their entry criteria) with 3 controls. They found that exposure to gadolinium was independently associated with the development of NSF.

Sadowski et al¹⁸ reported that 13 patients with biopsy-confirmed NSF all had been exposed to gadodiamide and one had been exposed to gadobenate (MultiHANCE) in addition to gadodiamide. All 13 patients had renal insufficiency, with an estimated glomerular filtration rate (GFR) less than 60 mL/minute/1.73 m². The investigators compared this group with a control group of patients with renal insufficiency who did not develop NSF. The NSF group had more proinflammatory events (P < .001) and more gadolinium-contrast-enhanced MRI examinations per patient (P = .002) than the control group.

Marckmann et al¹⁹ compared 19 patients who had histologically proven cases of NSF and 19 sex- and age-matched controls; all 38 patients had chronic kidney disease and had been exposed to gadolinium. Patients with NSF had received higher cumulative doses of gadodiamide and higher doses of erythropoietin and had higher serum concentrations of ionized calcium and phosphate than did their controls, as did patients with severe NSF compared with those with nonsevere NSF.

Comment. All the above reports are limited by their study design and suffer from recognition bias because not all of the patients with severe renal insufficiency who were exposed to gadolinium were examined for possible asymptomatic skin changes that might be characteristic of NSF. Therefore, it is impossible to be certain that all of the patients classified as not having NSF truly did not have it or did not subsequently develop it. Furthermore, the reports lacked standardized diagnostic criteria. Hence, the real prevalence and incidence of NSF are difficult to determine.

 

 

A cross-sectional study

As mentioned above, Todd et al⁸ examined 186 dialysis patients for cutaneous changes of NSF (using a scoring system based on hyper-pigmentation, hardening, and tethering of skin on the extremities). Patients who had been exposed to gadolinium had a higher risk of developing these skin changes than did nonexposed patients (odds ratio 14.7, 95% confidence interval 1.9–117.0). More importantly, the investigators found cutaneous changes of NSF in 25 (13%) of the 186 patients, 4 of whom had prior skin biopsies available for review, each revealing the histologic changes of NSF. This study suggests that NSF may be more prevalent than previously thought.

Is kidney dysfunction always present?

All the reported patients with NSF had underlying renal impairment. The renal dysfunction ranged from acute kidney injury to advanced chronic kidney disease (estimated GFR < 30 mL/minute/1.73 m²) and end-stage renal disease on renal replacement therapy, ie, hemodialysis or peritoneal dialysis. The incidence of NSF does not seem to be related to the cause of the underlying kidney disease.

What other diseases or comorbidities can be associated with NSF?

It is still unclear why not every patient with advanced renal failure develops NSF after exposure to gadolinium.

A variety of complex diseases and conditions have been reported to be associated with NSF, with no clear-cut evidence of causality or trigger. These include hypercoagulability states, thrombotic events, surgical procedures (especially those with reconstructive vascular components), calciphylaxis, kidney transplantation, hepatic disease (hepatorenal syndrome, liver transplantation, and hepatitis B and C), idiopathic pulmonary fibrosis, systemic lupus erythematosus, hypothyroidism, elevated serum ionized calcium or serum phosphate, hyperparathyroidism, and metabolic acidosis. A possible explanation is that most of these conditions are associated with an increased use of MRI or MRA testing (eg, in the workup for kidney or liver transplantation).

Many drugs have also been reported to be associated with NSF, including high-dose erythropoietin,²⁰ sevelamer (Renagel),²¹ and, conversely, lack of angiotensin-converting enzyme inhibitor therapy,²² but none of these findings has been reproduced to date.

GADOLINIUM CHARACTERISTICS AND PHARMACOKINETICS

Gadolinium is a rare-earth lanthanide metallic element (atomic number 64) that is used in MRI and MRA because of its paramagnetic properties that enhance the quality of imaging. Its ionic form (Gd³⁺) is highly toxic if injected intravenously, so it is typically bound to a “chelate” to decrease its toxicity.²³ The chelate stabilizes Gd³⁺ and thereby prevents its dissociation in vivo. These Gd-chelates can be classified (Table 2) according to their charge (ionic vs nonionic) and their structure (linear vs cyclic).

Most of the reported cases of NSF have been in patients who received gadodiamide, a nonionic, linear agent. Why gadodiamide has the highest rates of association with NSF is still unclear; perhaps it is simply the most widely used agent. Also, linear Gd compounds may be less stable and more likely to dissociate in vivo. The updated FDA Public Health Advisory in May 2007 warned against the use of all gadolinium-containing contrast agents for MRI, not just gadodiamide.

After intravenous injection, Gd-chelate equilibrates rapidly (within 2 hours) in the extracellular space. Very little of it enters into cells or binds to proteins. It is eliminated unchanged in the glomerular filtrate with no tubular secretion. In a study by Joffe et al,²⁴ the elimination half-life of gadodiamide in patients with severely reduced renal function was considerably longer than in healthy volunteers (34.3 hours ± 22.9 vs 1.3 hours ± 0.25).

Since gadolinium compounds are not protein-bound and have a limited volume of distribution, they are typically removed by hemodialysis. Joffe et al found that an average of 65% of the gadodiamide was removed in a single hemodialysis session. However, they did not describe the specific features of the hemodialysis session, and it took four hemodialysis treatments to remove 99% of a single dose of gadolinium.²⁴ A dialysis membrane with high permeability (large pores) seems to increase the clearance of the Gd-chelate during hemodialysis.²⁵

Peritoneal dialysis may not remove gadolinium as effectively: Joffe et al²⁴ reported that after 22 days of continuous ambulatory peritoneal dialysis, only 69% of the total amount of gadodiamide had been excreted, suggesting a very low peritoneal clearance.

SPECULATIVE PATHOGENESIS

Although a causal relationship between gadolinium use in patients with renal dysfunction and NSF has not been definitively established, the data derived from case reports assuredly raise this suspicion. Furthermore, on biopsy, gadolinium can be found in the skin of patients with NSF, adding evidence of causality.^26–28

The mechanism by which Gd³⁺ might trigger NSF is still not understood. A plausible speculation is that if renal function is reduced, the half-life of the Gd-chelate molecule is significantly increased, as is the chance of Gd³⁺ dissociating from its chelate, leading to increased tissue exposure. Vascular trauma and endothelial dysfunction may allow free Gd³⁺ to enter tissues more easily, where macrophages phagocytose the metal, produce local profibrotic cytokines, and send out signals that recruit circulating fibrocytes to the tissues. Once in tissues, circulating fibrocytes induce a fibrosing process that is indistinguishable from normal scar formation.²⁹

 

TREATMENTS LACK DATA

There is no consistently successful treatment for NSF.

In isolated reports, successful kidney transplantation slowed the skin fibrosis, but these findings need to be confirmed.^30,31 Data from case reports should be interpreted very cautiously, as they are by nature sporadic and anecdotal. Moreover most of the reports of NSF were published on Web sites or as editorials and did not undergo exhaustive peer review. Because the evidence is weak, kidney transplantation should not be recommended as a treatment for NSF.

Oral steroids, plasmapheresis, extracorporeal photopheresis, thalidomide, topical ultraviolet-A therapy, and other treatments have yielded very conflicting results, with only anecdotal improvement of symptoms. In a recent case report,³² the use of intravenous sodium thiosulfate in addition to aggressive physical therapy provided some benefit by reducing the pain and improving the skin lesions.

Because of the lack of strong evidence of efficacy, we cannot advocate the use of any of these treatments until larger clinical trial results are available. Aggressive physical therapy along with appropriate pain control may have benefits and should be offered to all patients suffering from NSF.

Avoid gadolinium exposure in patients with renal insufficiency

The FDA³³ recently asked manufacturers to include a new boxed warning on the product labeling of all gadolinium-based contrast agents (Magnevist, MultiHance, Omniscan, Opti-MARK, ProHance), due to risk of NSF in patients with acute or chronic severe renal insufficiency (GFR < 30 mL/minute/1.73 m²) and in patients with acute renal insufficiency of any severity due to hepatorenal syndrome or in the perioperative liver transplantation period.

For the time being, gadolinium should be contraindicated in patients with acute kidney injury and chronic kidney disease stages 4 and 5 and in those who are on renal replacement therapy (either hemodialysis or peritoneal dialysis). If an MRI study with gadolinium-based contrast is absolutely required in a patient with end-stage renal disease or advanced chronic kidney disease, an agent other than gadodiamide should be used in the lowest possible dose.

Will hemodialysis prevent NSF?

In a patient who is already on hemodialysis, it seems prudent to perform hemodialysis soon after gadolinium exposure and again the day after exposure to increase gadolinium elimination. However, to date, there are no data to support the theory that doing this will prevent NSF.

Because peritoneal dialysis has been reported to clear gadolinium poorly, use of gadolinium is contraindicated. If gadolinium is absolutely needed, either more-aggressive peritoneal dialysis (keeping the abdomen “wet”) or temporary hemodialysis may be considered.

For patients with advanced chronic kidney disease who are not yet on renal replacement therapy, the use of gadolinium is contraindicated, and hemodialysis should not be empirically recommended after gadolinium exposure because we have no evidence to support its utility and because hemodialysis may cause harm.

Nephrology consultation should be considered before any gadolinium use in a patient with impaired renal function, whether acute or chronic.

The use of gadolinium as a contrast agent in magnetic resonance imaging (MRI) in patients with impaired kidney function has come under scrutiny because of recent reports of a potential association between its use and nephrogenic systemic fibrosis (NSF).

See related editorial

This entity was first identified in the United States in 1997. Cowper et al¹ in 2000 described 15 hemodialysis patients who developed thickening and hardening of the skin with brawny hyperpigmentation, papules, and subcutaneous nodules on the extremities.

This “new disease” was initially called “nephrogenic fibrosing dermopathy,” as it was exclusively seen in patients with renal impairment and was thought to affect only the skin and subcutaneous tissue. With growing evidence of the extent and pathogenicity of the fibrosis in visceral organs, the nomenclature was changed to NSF, to better reflect the systemic nature of the disease.

PRESENTATION: MILD TO DEVASTATING

NSF has thus far been reported only in patients with renal impairment, most of whom were dialysis-dependent. It does not seem to be more common in one sex or the other, in any age range, or in any ethnic group. It can range in severity from mild to a devastating scleroderma-like systemic fibrosing disorder.

Cutaneous changes are the most predominant and impressive manifestations. NSF typically causes dermal hardening with tethering to deep dermal tissues, giving the skin the appearance of textured plaques, papules, or nodules with irregular edges and a brawny wooden texture to palpation (Figure 1). The lesions can be erythematous or brown-pigmented and can be painful and pruritic. NSF typically presents between the ankles and the thighs in a symmetric fashion and progresses proximally and distally to involve the entire lower extremities. Upper extremity involvement occurs frequently, but usually with lower extremity disease.² The trunk is involved less commonly than the legs and arms, and usually late in extensive disease. The face is typically spared (Figure 2).

NSF can cause loss of motion and contractures in multiple joints, leading to almost total loss of function and devastating debility within a short time—days to a few weeks.² These contractures are attributed to periarticular fibrosis of the overlying skin and subcutaneous tissue rather than to erosive joint disease. About 5% of patients develop a fulminant form of NSF³; these patients may become wheelchair-dependent.

The heart, lungs, skeletal muscle, and diaphragm can also be involved, sometimes leading to serious complications and death.^4–6

The disease is usually progressive and unremitting. Mendoza et al,⁷ in a review of 12 cases of NSF, reported that the disease had a progressive course in 6 patients, of whom 3 died within 2 years and 3 were ultimately confined to a wheelchair. More severe findings and rapid progression of the skin disease are associated with a poor prognosis.

Todd et al⁸ prospectively examined 186 dialysis patients to look for possible NSF. Of those with skin changes consistent with NSF, 48% died within 2 years, compared with 20% of those without these skin changes. Cardiovascular causes accounted for 58% of the deaths in patients with cutaneous changes of NSF and for 48% of the deaths in patients without these changes. Most of the excess deaths occurred within 6 months after the skin examination, suggesting an increased risk for early death in patients with skin changes suggestive of NSF.

DIAGNOSIS OF NSF IS CLINICAL

At presentation, NSF is frequently misdiagnosed and treated as cellulitis or edema. However, now that subspecialists—especially dermatologists, rheumatologists, and nephrologists—are becoming more aware of it, the correct diagnosis is being made earlier.

NSF should be suspected in any patient with underlying renal dysfunction—especially if on dialysis and if he or she has received a gadolinium contrast agent during MRI—who develops scleroderma-like cutaneous lesions affecting the distal extremities. Because most health care providers are still unfamiliar with this emerging disease, patients with renal impairment and suspected NSF should be referred to a rheumatologist or dermatologist to confirm the diagnosis, which is mainly entertained on a clinical basis. There is no laboratory biomarker for NSF.

A deep incisional skin biopsy may aid in the diagnosis. Due to the regional distribution of the disease, sampling error may occur, and repeat biopsy is warranted if the initial biopsy is nondiagnostic but the clinical picture suggests NSF.

 

Histopathologic examination typically shows lesions containing proliferation of dermal spindle cells, thick collagen bundles with surrounding clefts, and a variable amount of mucin and elastic fibers.² A characteristic and almost pathognomonic staining profile is the immunohistochemical identification of CD34 reactivity in the fibroblast-like cells (Figure 3). Cells expressing CD34 are normally found in the umbilical cord, the bone marrow (as pluripotential hematopoietic stem cells), and in the vascular endothelium. How they come to be in the skin is still speculative, but their presence suggests that circulating fibrocytes migrate from the bone marrow and deposit in the skin and other organs.^9,10

Pulmonary function testing can be done to rule out lung involvement and transthoracic two-dimensional echocardiography can be done to rule out possible cardiomyopathy if these conditions are suggested by examination at the time of diagnosis.⁷ Muscle biopsy is not necessary to determine the extent of systemic involvement, since the findings do not necessarily correlate with other systemic involvement.

 

DIFFERENTIAL DIAGNOSIS

Other disorders that can cause thickening and hardening of the skin of the extremities and trunk include systemic sclerosis or scleroderma, scleromyxedema, and eosinophilic fasciitis (Table 1). However, skin thickening, tethering, and hyperpigmentation in a patient with chronic kidney disease or end-stage renal disease after exposure to gadolinium-containing contrast agents suggests NSF.

An important diagnostic feature of NSF is that it spares the face, a finding derived from all reported and confirmed cases of NSF (Figure 2). In contrast, scleromyxedema, systemic scleroderma, and morphea often involve the face.

Scleromyxedema is often associated with monoclonal gammopathy (usually an immunoglobulin G lambda paraproteinemia) whereas NSF is not.

Scleroderma is supported by the findings of Raynaud’s phenomenon, antinuclear antibodies, and either anticentromere or anti-DNA topoisomerase I (Scl-70) antibodies, but the absence of these antibodies does not necessarily rule it out.

Eosinophilic fasciitis is diagnosed on the basis of histologic examination of a deep wedge skin biopsy specimen that includes fascia.

Other diagnoses that should be considered include amyloidosis and calciphylaxis.

ASSOCIATION WITH GADOLINIUM: WHAT IS THE EVIDENCE?

Case series

The association of gadolinium use with NSF has been described in several case reports and case series.

Grobner¹¹ reported that administration of gadodiamide (Omniscan, a gadolinium compound) for MRI was associated with NSF in five patients on chronic hemodialysis who had end-stage renal disease. Their ages ranged from 43 to 74 years, and they had been on dialysis from 10 to 58 months. The time of onset of NSF ranged from 2 to 4 weeks after exposure to gadodiamide.

Marckmann et al¹² reported that NSF developed in 13 (3.5%) of 370 patients with severe kidney disease who received gadodiamide. Five of the 13 patients had stage 5 (advanced) chronic kidney disease and were not yet on renal replacement therapy, 7 were on hemodialysis, and 1 was on peritoneal dialysis. The time of onset ranged from 2 to 75 days (median 25 days) after exposure.

Kuo et al¹³ similarly estimated the incidence of NSF at approximately 3% in patients with severe renal failure who receive intravenous gadolinium-based contrast material for MRI.

Broome et al¹⁴ reported that 12 patients developed NSF within 2 to 11 weeks after receiving gadodiamide. Eight of the 12 patients had end-stage renal disease and were on hemodialysis; the other 4 patients had acute kidney injury attributed to hepatorenal syndrome, and 3 of these 4 patients were on hemodialysis.

Khurana et al¹⁵ reported that 6 patients on hemodialysis developed NSF from 2 weeks to 2 months after receiving a dose of gadodiamide of between 0.11 and 0.36 mmol/kg. These doses are high, and the findings suggest an association between the gadolinium dose and NSF. The dose approved by the US Food and Drug Administration (FDA) is only 0.1 mmol/kg, and the use of gadolinium is approved only in MRI. However, higher doses (0.3–0.4 mmol/kg) are widely used in practice for better imaging quality in magnetic resonance angiography (MRA).

Deo et al¹⁶ reported 3 cases of NSF in 87 patients with end-stage renal disease who underwent 123 radiologic studies with gadolinium. No patient with end-stage renal disease who was not exposed to gadolinium developed NSF, and the association between exposure to gadolinium and the subsequent development of NSF was statistically significant (P = .006). The authors concluded that each gadolinium study presented a 2.4% risk of NSF in end-stage renal disease patients.

This retrospective study is flawed by not having been cross-sectional or case-controlled, since the other 84 patients who received gadolinium were not examined at all to establish the absence of NSF.

Case-control studies

More evidence of association of NSF with gadolinium exposure comes from other reports.

Physicians in St. Louis, MO,¹⁷ identified 33 cases of NSF and performed a case-control study, matching each of 19 of the patients (for whom data were available and who met their entry criteria) with 3 controls. They found that exposure to gadolinium was independently associated with the development of NSF.

Sadowski et al¹⁸ reported that 13 patients with biopsy-confirmed NSF all had been exposed to gadodiamide and one had been exposed to gadobenate (MultiHANCE) in addition to gadodiamide. All 13 patients had renal insufficiency, with an estimated glomerular filtration rate (GFR) less than 60 mL/minute/1.73 m². The investigators compared this group with a control group of patients with renal insufficiency who did not develop NSF. The NSF group had more proinflammatory events (P < .001) and more gadolinium-contrast-enhanced MRI examinations per patient (P = .002) than the control group.

Marckmann et al¹⁹ compared 19 patients who had histologically proven cases of NSF and 19 sex- and age-matched controls; all 38 patients had chronic kidney disease and had been exposed to gadolinium. Patients with NSF had received higher cumulative doses of gadodiamide and higher doses of erythropoietin and had higher serum concentrations of ionized calcium and phosphate than did their controls, as did patients with severe NSF compared with those with nonsevere NSF.

Comment. All the above reports are limited by their study design and suffer from recognition bias because not all of the patients with severe renal insufficiency who were exposed to gadolinium were examined for possible asymptomatic skin changes that might be characteristic of NSF. Therefore, it is impossible to be certain that all of the patients classified as not having NSF truly did not have it or did not subsequently develop it. Furthermore, the reports lacked standardized diagnostic criteria. Hence, the real prevalence and incidence of NSF are difficult to determine.

 

 

A cross-sectional study

As mentioned above, Todd et al⁸ examined 186 dialysis patients for cutaneous changes of NSF (using a scoring system based on hyper-pigmentation, hardening, and tethering of skin on the extremities). Patients who had been exposed to gadolinium had a higher risk of developing these skin changes than did nonexposed patients (odds ratio 14.7, 95% confidence interval 1.9–117.0). More importantly, the investigators found cutaneous changes of NSF in 25 (13%) of the 186 patients, 4 of whom had prior skin biopsies available for review, each revealing the histologic changes of NSF. This study suggests that NSF may be more prevalent than previously thought.

Is kidney dysfunction always present?

All the reported patients with NSF had underlying renal impairment. The renal dysfunction ranged from acute kidney injury to advanced chronic kidney disease (estimated GFR < 30 mL/minute/1.73 m²) and end-stage renal disease on renal replacement therapy, ie, hemodialysis or peritoneal dialysis. The incidence of NSF does not seem to be related to the cause of the underlying kidney disease.

What other diseases or comorbidities can be associated with NSF?

It is still unclear why not every patient with advanced renal failure develops NSF after exposure to gadolinium.

A variety of complex diseases and conditions have been reported to be associated with NSF, with no clear-cut evidence of causality or trigger. These include hypercoagulability states, thrombotic events, surgical procedures (especially those with reconstructive vascular components), calciphylaxis, kidney transplantation, hepatic disease (hepatorenal syndrome, liver transplantation, and hepatitis B and C), idiopathic pulmonary fibrosis, systemic lupus erythematosus, hypothyroidism, elevated serum ionized calcium or serum phosphate, hyperparathyroidism, and metabolic acidosis. A possible explanation is that most of these conditions are associated with an increased use of MRI or MRA testing (eg, in the workup for kidney or liver transplantation).

Many drugs have also been reported to be associated with NSF, including high-dose erythropoietin,²⁰ sevelamer (Renagel),²¹ and, conversely, lack of angiotensin-converting enzyme inhibitor therapy,²² but none of these findings has been reproduced to date.

GADOLINIUM CHARACTERISTICS AND PHARMACOKINETICS

Gadolinium is a rare-earth lanthanide metallic element (atomic number 64) that is used in MRI and MRA because of its paramagnetic properties that enhance the quality of imaging. Its ionic form (Gd³⁺) is highly toxic if injected intravenously, so it is typically bound to a “chelate” to decrease its toxicity.²³ The chelate stabilizes Gd³⁺ and thereby prevents its dissociation in vivo. These Gd-chelates can be classified (Table 2) according to their charge (ionic vs nonionic) and their structure (linear vs cyclic).

Most of the reported cases of NSF have been in patients who received gadodiamide, a nonionic, linear agent. Why gadodiamide has the highest rates of association with NSF is still unclear; perhaps it is simply the most widely used agent. Also, linear Gd compounds may be less stable and more likely to dissociate in vivo. The updated FDA Public Health Advisory in May 2007 warned against the use of all gadolinium-containing contrast agents for MRI, not just gadodiamide.

After intravenous injection, Gd-chelate equilibrates rapidly (within 2 hours) in the extracellular space. Very little of it enters into cells or binds to proteins. It is eliminated unchanged in the glomerular filtrate with no tubular secretion. In a study by Joffe et al,²⁴ the elimination half-life of gadodiamide in patients with severely reduced renal function was considerably longer than in healthy volunteers (34.3 hours ± 22.9 vs 1.3 hours ± 0.25).

Since gadolinium compounds are not protein-bound and have a limited volume of distribution, they are typically removed by hemodialysis. Joffe et al found that an average of 65% of the gadodiamide was removed in a single hemodialysis session. However, they did not describe the specific features of the hemodialysis session, and it took four hemodialysis treatments to remove 99% of a single dose of gadolinium.²⁴ A dialysis membrane with high permeability (large pores) seems to increase the clearance of the Gd-chelate during hemodialysis.²⁵

Peritoneal dialysis may not remove gadolinium as effectively: Joffe et al²⁴ reported that after 22 days of continuous ambulatory peritoneal dialysis, only 69% of the total amount of gadodiamide had been excreted, suggesting a very low peritoneal clearance.

SPECULATIVE PATHOGENESIS

Although a causal relationship between gadolinium use in patients with renal dysfunction and NSF has not been definitively established, the data derived from case reports assuredly raise this suspicion. Furthermore, on biopsy, gadolinium can be found in the skin of patients with NSF, adding evidence of causality.^26–28

The mechanism by which Gd³⁺ might trigger NSF is still not understood. A plausible speculation is that if renal function is reduced, the half-life of the Gd-chelate molecule is significantly increased, as is the chance of Gd³⁺ dissociating from its chelate, leading to increased tissue exposure. Vascular trauma and endothelial dysfunction may allow free Gd³⁺ to enter tissues more easily, where macrophages phagocytose the metal, produce local profibrotic cytokines, and send out signals that recruit circulating fibrocytes to the tissues. Once in tissues, circulating fibrocytes induce a fibrosing process that is indistinguishable from normal scar formation.²⁹

 

TREATMENTS LACK DATA

There is no consistently successful treatment for NSF.

In isolated reports, successful kidney transplantation slowed the skin fibrosis, but these findings need to be confirmed.^30,31 Data from case reports should be interpreted very cautiously, as they are by nature sporadic and anecdotal. Moreover most of the reports of NSF were published on Web sites or as editorials and did not undergo exhaustive peer review. Because the evidence is weak, kidney transplantation should not be recommended as a treatment for NSF.

Oral steroids, plasmapheresis, extracorporeal photopheresis, thalidomide, topical ultraviolet-A therapy, and other treatments have yielded very conflicting results, with only anecdotal improvement of symptoms. In a recent case report,³² the use of intravenous sodium thiosulfate in addition to aggressive physical therapy provided some benefit by reducing the pain and improving the skin lesions.

Because of the lack of strong evidence of efficacy, we cannot advocate the use of any of these treatments until larger clinical trial results are available. Aggressive physical therapy along with appropriate pain control may have benefits and should be offered to all patients suffering from NSF.

Avoid gadolinium exposure in patients with renal insufficiency

The FDA³³ recently asked manufacturers to include a new boxed warning on the product labeling of all gadolinium-based contrast agents (Magnevist, MultiHance, Omniscan, Opti-MARK, ProHance), due to risk of NSF in patients with acute or chronic severe renal insufficiency (GFR < 30 mL/minute/1.73 m²) and in patients with acute renal insufficiency of any severity due to hepatorenal syndrome or in the perioperative liver transplantation period.

For the time being, gadolinium should be contraindicated in patients with acute kidney injury and chronic kidney disease stages 4 and 5 and in those who are on renal replacement therapy (either hemodialysis or peritoneal dialysis). If an MRI study with gadolinium-based contrast is absolutely required in a patient with end-stage renal disease or advanced chronic kidney disease, an agent other than gadodiamide should be used in the lowest possible dose.

Will hemodialysis prevent NSF?

In a patient who is already on hemodialysis, it seems prudent to perform hemodialysis soon after gadolinium exposure and again the day after exposure to increase gadolinium elimination. However, to date, there are no data to support the theory that doing this will prevent NSF.

Because peritoneal dialysis has been reported to clear gadolinium poorly, use of gadolinium is contraindicated. If gadolinium is absolutely needed, either more-aggressive peritoneal dialysis (keeping the abdomen “wet”) or temporary hemodialysis may be considered.

For patients with advanced chronic kidney disease who are not yet on renal replacement therapy, the use of gadolinium is contraindicated, and hemodialysis should not be empirically recommended after gadolinium exposure because we have no evidence to support its utility and because hemodialysis may cause harm.

Nephrology consultation should be considered before any gadolinium use in a patient with impaired renal function, whether acute or chronic.

The use of gadolinium as a contrast agent in magnetic resonance imaging (MRI) in patients with impaired kidney function has come under scrutiny because of recent reports of a potential association between its use and nephrogenic systemic fibrosis (NSF).

See related editorial

This entity was first identified in the United States in 1997. Cowper et al¹ in 2000 described 15 hemodialysis patients who developed thickening and hardening of the skin with brawny hyperpigmentation, papules, and subcutaneous nodules on the extremities.

This “new disease” was initially called “nephrogenic fibrosing dermopathy,” as it was exclusively seen in patients with renal impairment and was thought to affect only the skin and subcutaneous tissue. With growing evidence of the extent and pathogenicity of the fibrosis in visceral organs, the nomenclature was changed to NSF, to better reflect the systemic nature of the disease.

PRESENTATION: MILD TO DEVASTATING

NSF has thus far been reported only in patients with renal impairment, most of whom were dialysis-dependent. It does not seem to be more common in one sex or the other, in any age range, or in any ethnic group. It can range in severity from mild to a devastating scleroderma-like systemic fibrosing disorder.

Cutaneous changes are the most predominant and impressive manifestations. NSF typically causes dermal hardening with tethering to deep dermal tissues, giving the skin the appearance of textured plaques, papules, or nodules with irregular edges and a brawny wooden texture to palpation (Figure 1). The lesions can be erythematous or brown-pigmented and can be painful and pruritic. NSF typically presents between the ankles and the thighs in a symmetric fashion and progresses proximally and distally to involve the entire lower extremities. Upper extremity involvement occurs frequently, but usually with lower extremity disease.² The trunk is involved less commonly than the legs and arms, and usually late in extensive disease. The face is typically spared (Figure 2).

NSF can cause loss of motion and contractures in multiple joints, leading to almost total loss of function and devastating debility within a short time—days to a few weeks.² These contractures are attributed to periarticular fibrosis of the overlying skin and subcutaneous tissue rather than to erosive joint disease. About 5% of patients develop a fulminant form of NSF³; these patients may become wheelchair-dependent.

The heart, lungs, skeletal muscle, and diaphragm can also be involved, sometimes leading to serious complications and death.^4–6

The disease is usually progressive and unremitting. Mendoza et al,⁷ in a review of 12 cases of NSF, reported that the disease had a progressive course in 6 patients, of whom 3 died within 2 years and 3 were ultimately confined to a wheelchair. More severe findings and rapid progression of the skin disease are associated with a poor prognosis.

Todd et al⁸ prospectively examined 186 dialysis patients to look for possible NSF. Of those with skin changes consistent with NSF, 48% died within 2 years, compared with 20% of those without these skin changes. Cardiovascular causes accounted for 58% of the deaths in patients with cutaneous changes of NSF and for 48% of the deaths in patients without these changes. Most of the excess deaths occurred within 6 months after the skin examination, suggesting an increased risk for early death in patients with skin changes suggestive of NSF.

DIAGNOSIS OF NSF IS CLINICAL

At presentation, NSF is frequently misdiagnosed and treated as cellulitis or edema. However, now that subspecialists—especially dermatologists, rheumatologists, and nephrologists—are becoming more aware of it, the correct diagnosis is being made earlier.

NSF should be suspected in any patient with underlying renal dysfunction—especially if on dialysis and if he or she has received a gadolinium contrast agent during MRI—who develops scleroderma-like cutaneous lesions affecting the distal extremities. Because most health care providers are still unfamiliar with this emerging disease, patients with renal impairment and suspected NSF should be referred to a rheumatologist or dermatologist to confirm the diagnosis, which is mainly entertained on a clinical basis. There is no laboratory biomarker for NSF.

A deep incisional skin biopsy may aid in the diagnosis. Due to the regional distribution of the disease, sampling error may occur, and repeat biopsy is warranted if the initial biopsy is nondiagnostic but the clinical picture suggests NSF.

 

Histopathologic examination typically shows lesions containing proliferation of dermal spindle cells, thick collagen bundles with surrounding clefts, and a variable amount of mucin and elastic fibers.² A characteristic and almost pathognomonic staining profile is the immunohistochemical identification of CD34 reactivity in the fibroblast-like cells (Figure 3). Cells expressing CD34 are normally found in the umbilical cord, the bone marrow (as pluripotential hematopoietic stem cells), and in the vascular endothelium. How they come to be in the skin is still speculative, but their presence suggests that circulating fibrocytes migrate from the bone marrow and deposit in the skin and other organs.^9,10

Pulmonary function testing can be done to rule out lung involvement and transthoracic two-dimensional echocardiography can be done to rule out possible cardiomyopathy if these conditions are suggested by examination at the time of diagnosis.⁷ Muscle biopsy is not necessary to determine the extent of systemic involvement, since the findings do not necessarily correlate with other systemic involvement.

 

DIFFERENTIAL DIAGNOSIS

Other disorders that can cause thickening and hardening of the skin of the extremities and trunk include systemic sclerosis or scleroderma, scleromyxedema, and eosinophilic fasciitis (Table 1). However, skin thickening, tethering, and hyperpigmentation in a patient with chronic kidney disease or end-stage renal disease after exposure to gadolinium-containing contrast agents suggests NSF.

An important diagnostic feature of NSF is that it spares the face, a finding derived from all reported and confirmed cases of NSF (Figure 2). In contrast, scleromyxedema, systemic scleroderma, and morphea often involve the face.

Scleromyxedema is often associated with monoclonal gammopathy (usually an immunoglobulin G lambda paraproteinemia) whereas NSF is not.

Scleroderma is supported by the findings of Raynaud’s phenomenon, antinuclear antibodies, and either anticentromere or anti-DNA topoisomerase I (Scl-70) antibodies, but the absence of these antibodies does not necessarily rule it out.

Eosinophilic fasciitis is diagnosed on the basis of histologic examination of a deep wedge skin biopsy specimen that includes fascia.

Other diagnoses that should be considered include amyloidosis and calciphylaxis.

ASSOCIATION WITH GADOLINIUM: WHAT IS THE EVIDENCE?

Case series

The association of gadolinium use with NSF has been described in several case reports and case series.

Grobner¹¹ reported that administration of gadodiamide (Omniscan, a gadolinium compound) for MRI was associated with NSF in five patients on chronic hemodialysis who had end-stage renal disease. Their ages ranged from 43 to 74 years, and they had been on dialysis from 10 to 58 months. The time of onset of NSF ranged from 2 to 4 weeks after exposure to gadodiamide.

Marckmann et al¹² reported that NSF developed in 13 (3.5%) of 370 patients with severe kidney disease who received gadodiamide. Five of the 13 patients had stage 5 (advanced) chronic kidney disease and were not yet on renal replacement therapy, 7 were on hemodialysis, and 1 was on peritoneal dialysis. The time of onset ranged from 2 to 75 days (median 25 days) after exposure.

Kuo et al¹³ similarly estimated the incidence of NSF at approximately 3% in patients with severe renal failure who receive intravenous gadolinium-based contrast material for MRI.

Broome et al¹⁴ reported that 12 patients developed NSF within 2 to 11 weeks after receiving gadodiamide. Eight of the 12 patients had end-stage renal disease and were on hemodialysis; the other 4 patients had acute kidney injury attributed to hepatorenal syndrome, and 3 of these 4 patients were on hemodialysis.

Khurana et al¹⁵ reported that 6 patients on hemodialysis developed NSF from 2 weeks to 2 months after receiving a dose of gadodiamide of between 0.11 and 0.36 mmol/kg. These doses are high, and the findings suggest an association between the gadolinium dose and NSF. The dose approved by the US Food and Drug Administration (FDA) is only 0.1 mmol/kg, and the use of gadolinium is approved only in MRI. However, higher doses (0.3–0.4 mmol/kg) are widely used in practice for better imaging quality in magnetic resonance angiography (MRA).

Deo et al¹⁶ reported 3 cases of NSF in 87 patients with end-stage renal disease who underwent 123 radiologic studies with gadolinium. No patient with end-stage renal disease who was not exposed to gadolinium developed NSF, and the association between exposure to gadolinium and the subsequent development of NSF was statistically significant (P = .006). The authors concluded that each gadolinium study presented a 2.4% risk of NSF in end-stage renal disease patients.

This retrospective study is flawed by not having been cross-sectional or case-controlled, since the other 84 patients who received gadolinium were not examined at all to establish the absence of NSF.

Case-control studies

More evidence of association of NSF with gadolinium exposure comes from other reports.

Physicians in St. Louis, MO,¹⁷ identified 33 cases of NSF and performed a case-control study, matching each of 19 of the patients (for whom data were available and who met their entry criteria) with 3 controls. They found that exposure to gadolinium was independently associated with the development of NSF.

Sadowski et al¹⁸ reported that 13 patients with biopsy-confirmed NSF all had been exposed to gadodiamide and one had been exposed to gadobenate (MultiHANCE) in addition to gadodiamide. All 13 patients had renal insufficiency, with an estimated glomerular filtration rate (GFR) less than 60 mL/minute/1.73 m². The investigators compared this group with a control group of patients with renal insufficiency who did not develop NSF. The NSF group had more proinflammatory events (P < .001) and more gadolinium-contrast-enhanced MRI examinations per patient (P = .002) than the control group.

Marckmann et al¹⁹ compared 19 patients who had histologically proven cases of NSF and 19 sex- and age-matched controls; all 38 patients had chronic kidney disease and had been exposed to gadolinium. Patients with NSF had received higher cumulative doses of gadodiamide and higher doses of erythropoietin and had higher serum concentrations of ionized calcium and phosphate than did their controls, as did patients with severe NSF compared with those with nonsevere NSF.

Comment. All the above reports are limited by their study design and suffer from recognition bias because not all of the patients with severe renal insufficiency who were exposed to gadolinium were examined for possible asymptomatic skin changes that might be characteristic of NSF. Therefore, it is impossible to be certain that all of the patients classified as not having NSF truly did not have it or did not subsequently develop it. Furthermore, the reports lacked standardized diagnostic criteria. Hence, the real prevalence and incidence of NSF are difficult to determine.

 

 

A cross-sectional study

As mentioned above, Todd et al⁸ examined 186 dialysis patients for cutaneous changes of NSF (using a scoring system based on hyper-pigmentation, hardening, and tethering of skin on the extremities). Patients who had been exposed to gadolinium had a higher risk of developing these skin changes than did nonexposed patients (odds ratio 14.7, 95% confidence interval 1.9–117.0). More importantly, the investigators found cutaneous changes of NSF in 25 (13%) of the 186 patients, 4 of whom had prior skin biopsies available for review, each revealing the histologic changes of NSF. This study suggests that NSF may be more prevalent than previously thought.

Is kidney dysfunction always present?

All the reported patients with NSF had underlying renal impairment. The renal dysfunction ranged from acute kidney injury to advanced chronic kidney disease (estimated GFR < 30 mL/minute/1.73 m²) and end-stage renal disease on renal replacement therapy, ie, hemodialysis or peritoneal dialysis. The incidence of NSF does not seem to be related to the cause of the underlying kidney disease.

What other diseases or comorbidities can be associated with NSF?

It is still unclear why not every patient with advanced renal failure develops NSF after exposure to gadolinium.

A variety of complex diseases and conditions have been reported to be associated with NSF, with no clear-cut evidence of causality or trigger. These include hypercoagulability states, thrombotic events, surgical procedures (especially those with reconstructive vascular components), calciphylaxis, kidney transplantation, hepatic disease (hepatorenal syndrome, liver transplantation, and hepatitis B and C), idiopathic pulmonary fibrosis, systemic lupus erythematosus, hypothyroidism, elevated serum ionized calcium or serum phosphate, hyperparathyroidism, and metabolic acidosis. A possible explanation is that most of these conditions are associated with an increased use of MRI or MRA testing (eg, in the workup for kidney or liver transplantation).

Many drugs have also been reported to be associated with NSF, including high-dose erythropoietin,²⁰ sevelamer (Renagel),²¹ and, conversely, lack of angiotensin-converting enzyme inhibitor therapy,²² but none of these findings has been reproduced to date.

GADOLINIUM CHARACTERISTICS AND PHARMACOKINETICS

Gadolinium is a rare-earth lanthanide metallic element (atomic number 64) that is used in MRI and MRA because of its paramagnetic properties that enhance the quality of imaging. Its ionic form (Gd³⁺) is highly toxic if injected intravenously, so it is typically bound to a “chelate” to decrease its toxicity.²³ The chelate stabilizes Gd³⁺ and thereby prevents its dissociation in vivo. These Gd-chelates can be classified (Table 2) according to their charge (ionic vs nonionic) and their structure (linear vs cyclic).

Most of the reported cases of NSF have been in patients who received gadodiamide, a nonionic, linear agent. Why gadodiamide has the highest rates of association with NSF is still unclear; perhaps it is simply the most widely used agent. Also, linear Gd compounds may be less stable and more likely to dissociate in vivo. The updated FDA Public Health Advisory in May 2007 warned against the use of all gadolinium-containing contrast agents for MRI, not just gadodiamide.

After intravenous injection, Gd-chelate equilibrates rapidly (within 2 hours) in the extracellular space. Very little of it enters into cells or binds to proteins. It is eliminated unchanged in the glomerular filtrate with no tubular secretion. In a study by Joffe et al,²⁴ the elimination half-life of gadodiamide in patients with severely reduced renal function was considerably longer than in healthy volunteers (34.3 hours ± 22.9 vs 1.3 hours ± 0.25).

Since gadolinium compounds are not protein-bound and have a limited volume of distribution, they are typically removed by hemodialysis. Joffe et al found that an average of 65% of the gadodiamide was removed in a single hemodialysis session. However, they did not describe the specific features of the hemodialysis session, and it took four hemodialysis treatments to remove 99% of a single dose of gadolinium.²⁴ A dialysis membrane with high permeability (large pores) seems to increase the clearance of the Gd-chelate during hemodialysis.²⁵

Peritoneal dialysis may not remove gadolinium as effectively: Joffe et al²⁴ reported that after 22 days of continuous ambulatory peritoneal dialysis, only 69% of the total amount of gadodiamide had been excreted, suggesting a very low peritoneal clearance.

SPECULATIVE PATHOGENESIS

Although a causal relationship between gadolinium use in patients with renal dysfunction and NSF has not been definitively established, the data derived from case reports assuredly raise this suspicion. Furthermore, on biopsy, gadolinium can be found in the skin of patients with NSF, adding evidence of causality.^26–28

The mechanism by which Gd³⁺ might trigger NSF is still not understood. A plausible speculation is that if renal function is reduced, the half-life of the Gd-chelate molecule is significantly increased, as is the chance of Gd³⁺ dissociating from its chelate, leading to increased tissue exposure. Vascular trauma and endothelial dysfunction may allow free Gd³⁺ to enter tissues more easily, where macrophages phagocytose the metal, produce local profibrotic cytokines, and send out signals that recruit circulating fibrocytes to the tissues. Once in tissues, circulating fibrocytes induce a fibrosing process that is indistinguishable from normal scar formation.²⁹

 

TREATMENTS LACK DATA

There is no consistently successful treatment for NSF.

In isolated reports, successful kidney transplantation slowed the skin fibrosis, but these findings need to be confirmed.^30,31 Data from case reports should be interpreted very cautiously, as they are by nature sporadic and anecdotal. Moreover most of the reports of NSF were published on Web sites or as editorials and did not undergo exhaustive peer review. Because the evidence is weak, kidney transplantation should not be recommended as a treatment for NSF.

Oral steroids, plasmapheresis, extracorporeal photopheresis, thalidomide, topical ultraviolet-A therapy, and other treatments have yielded very conflicting results, with only anecdotal improvement of symptoms. In a recent case report,³² the use of intravenous sodium thiosulfate in addition to aggressive physical therapy provided some benefit by reducing the pain and improving the skin lesions.

Because of the lack of strong evidence of efficacy, we cannot advocate the use of any of these treatments until larger clinical trial results are available. Aggressive physical therapy along with appropriate pain control may have benefits and should be offered to all patients suffering from NSF.

Avoid gadolinium exposure in patients with renal insufficiency

The FDA³³ recently asked manufacturers to include a new boxed warning on the product labeling of all gadolinium-based contrast agents (Magnevist, MultiHance, Omniscan, Opti-MARK, ProHance), due to risk of NSF in patients with acute or chronic severe renal insufficiency (GFR < 30 mL/minute/1.73 m²) and in patients with acute renal insufficiency of any severity due to hepatorenal syndrome or in the perioperative liver transplantation period.

For the time being, gadolinium should be contraindicated in patients with acute kidney injury and chronic kidney disease stages 4 and 5 and in those who are on renal replacement therapy (either hemodialysis or peritoneal dialysis). If an MRI study with gadolinium-based contrast is absolutely required in a patient with end-stage renal disease or advanced chronic kidney disease, an agent other than gadodiamide should be used in the lowest possible dose.

Will hemodialysis prevent NSF?

In a patient who is already on hemodialysis, it seems prudent to perform hemodialysis soon after gadolinium exposure and again the day after exposure to increase gadolinium elimination. However, to date, there are no data to support the theory that doing this will prevent NSF.

Because peritoneal dialysis has been reported to clear gadolinium poorly, use of gadolinium is contraindicated. If gadolinium is absolutely needed, either more-aggressive peritoneal dialysis (keeping the abdomen “wet”) or temporary hemodialysis may be considered.

For patients with advanced chronic kidney disease who are not yet on renal replacement therapy, the use of gadolinium is contraindicated, and hemodialysis should not be empirically recommended after gadolinium exposure because we have no evidence to support its utility and because hemodialysis may cause harm.

Nephrology consultation should be considered before any gadolinium use in a patient with impaired renal function, whether acute or chronic.