Nephrogenic systemic fibrosis (NSF) is a rapidly progressive, debilitating condition that causes cutaneous and visceral fibrosis in patients with renal failure. Little is known about its prevalence or etiology. The aim of this study was to establish the prevalence of NSF and associated risk factors
Two cohorts of patients were recruited from 6 outpatient hemodialysis centers and examined for cutaneous changes of NSF, which were defined using a scoring system based on hyperpigmentation, hardening, and tethering of skin on the extremities. Demographic data were gathered, mortality was followed up prospectively for 24 months, and gadolinium exposure was ascertained for a subgroup of patients in the second cohort.
Examination reproducibility was 97% in cohort 1. In cohort 2, 25 (13%) of 186 patients demonstrated cutaneous changes of NSF. Twenty-four–month mortality following examination was 48% and 20% in patients with and those without cutaneous changes of NSF, respectively (adjusted hazard ratio 2.9, 95% confidence interval [95% CI] 1.4–5.9). Cutaneous changes of NSF were observed in 16 (30%) of 54 patients with prior exposure to gadopentetate dimeglumine contrast during imaging studies. Exposure to gadolinium-containing contrast was associated with an increased risk of developing cutaneous changes of NSF (odds ratio 14.7, 95% CI 1.9–117.0) compared with nonexposed patients.
In patients receiving hemodialysis, NSF is an underrecognized disorder that is associated with increased mortality. Exposure to gadolinium-containing contrast material appears to be a significant risk factor for the development of NSF.
Nephrogenic systemic fibrosis (NSF) is a recently described, extremely debilitating and painful condition that affects patients with renal failure. It is characterized by rapidly progressive skin hardening, tethering, and hyperpigmentation, predominantly on the extremities (1, 2). Flexion contractures develop in advanced NSF that may severely impair physical function, including ambulation (2). Cases of NSF with visceral involvement also have been described (3–6). Renal failure is a common feature in all patients with NSF, but reports of small case series have not suggested that the risk of development of NSF is associated with a patient's age, sex, race, or cause of renal disease (7). Recent reports suggest an association between exposure to gadolinium-containing contrast material during magnetic resonance imaging (MRI) studies and the subsequent development of NSF in patients with chronic kidney disease (8–12).
The clinical presentation of NSF is similar to that of other fibrosing disorders and most closely resembles scleromyxedema. However, the clinical features of NSF differ from those of scleromyxedema in that the extracutaneous visceral mucin deposition and associated IgGλ paraproteinemia of scleromyxedema are absent in NSF (1). The unique histologic appearance of NSF also distinguishes it as a discrete entity. Biopsy specimens obtained from early NSF skin lesions demonstrate thickened collagen bundles, mucin deposition, angiogenesis, and numerous dermal spindle cells that stain with antibodies to CD34 and procollagen I (7, 13). Decreased skin thickening and improved joint mobility have been reported in 3 patients treated with extracorporeal photopheresis (14) and in 2 other patients treated with pentoxiphylline (8). However, other modalities, such as topical steroids, oral steroids, immunosuppressive therapy, and plasmapheresis, have not been successful in treating NSF (2).
The limited number of patients with NSF who have been identified has constrained efforts to identify causative disease agents and to define effective treatment modalities. Indeed, case series describing NSF are often small, with 4 or fewer patients (15–18). Given the recent description, sporadic reporting, and relative obscurity of NSF, the prevalence of this condition is entirely unknown. This is the first study in which patients receiving hemodialysis in outpatient centers were systematically examined for skin changes of NSF. Our results allow us to establish the prevalence of cutaneous changes of NSF in this population, to assess associated mortality, and to identify factors potentially predisposing to the development of NSF.
PATIENTS AND METHODS
We systematically examined 2 cohorts of patients receiving hemodialysis at 6 outpatient centers in the Boston metropolitan area for skin changes of NSF. Patients in cohort 1 were recruited by approaching all patients receiving hemodialysis during 2 shifts at a single center on the same date. Results were used to test the reproducibility of the skin examination itself. Patients in cohort 2 were recruited by approaching all patients receiving hemodialysis at the 5 other centers on dates that spanned at least 2 different days per center (4–6 shifts per center). Patients were included regardless of race, sex, underlying renal disease, duration of renal disease, or any other medical parameter. Exclusion criteria were age younger than 18 years or the inability to provide verbal consent. There was no duplication of patients between the cohorts. All patients provided verbal consent to enroll in the study. In each cohort, only 6% of eligible patients declined participation. The study was conducted in accordance with conditions approved by the Institutional Review Boards of Partners Healthcare System (PHS) and Fresenius Medical Care North America.
Examination and scoring.
The extremities of patients were examined for 3 cutaneous changes characteristic of NSF: hyperpigmentation, hardening, and tethering of skin. Patients were asked to refrain from discussing the results of their examinations with the examiners. Investigators DJT, AK, and JK each independently examined the skin of all 30 patients in cohort 1 on the same day, remaining blinded to the results of the others' examinations until after completion of the study. DJT examined all 186 patients in cohort 2.
Hyperpigmentation was determined by comparing the appearance of skin on the extremities with that of skin elsewhere. Hardening was assessed by gradually applying increasing pressure while gently palpating the skin on the extremities in a circular motion. Tethering was assessed as resistance to motion while gently pinching the skin. Binary values were assigned to each characteristic, as follows: for pigmentation, 0 = no hyperpigmentation, 1 = hyperpigmented skin; for hardening, 0 = supple skin, 1 = skin that resisted applied pressure; for tethering, 0 = freely mobile skin, 1 = skin with restricted movement over the underlying fascia. Clinical evidence of NSF (“examination-positive”) was defined as a patient having any 2 or all 3 skin findings (2 or 3 points) on the extremities bilaterally. Patients were considered “examination-negative” if they had none or only 1 of the skin findings. For examination-positive patients in each cohort, the involved extremities were assessed for joint contractures.
Demographics and mortality data.
For all study patients, the following demographic data were obtained from the medical records that were present at the hemodialysis facility: date of birth, sex, race, etiology of renal disease, duration of hemodialysis, presence of diabetes mellitus, and prior renal transplant. The etiology of renal disease was categorized into that caused by diabetes mellitus, hypertension, glomerulonephritis, polycystic kidney disease, or other conditions. Mortality data were obtained from hemodialysis facility records for the 24 months following skin examination. Cause of death was that reported by the physician treating the patient at the time of death. When available, reports of skin biopsy results were obtained, after clinical changes of NSF had been determined in a patient.
To ascertain whether exposure to gadolinium-containing contrast material was associated with the development of skin changes characteristic of NSF, patients in cohort 2 were divided into 2 subgroups: those who had an active electronic medical record (EMR) within PHS and those who did not. An active EMR was defined as one that included, within the year before the skin examination, at least 2 outpatient visits or hospitalizations at 1 of the 6 PHS hospitals, and at least 2 radiographic studies of any type. For each patient with an active EMR, the EMR was then reviewed to identify any exposure to gadolinium-containing contrast material during any imaging study (MRI, computed tomography scanning, angiography, or fistulography) performed prior to the skin examination. The date of any gadolinium-enhanced imaging study was compared with the date of the first hemodialysis treatment, according to the Fresenius Medical Care North America record, to confirm that exposure of the patient to gadolinium-containing contrast media had occurred after the development of chronic kidney disease.
Examination-positive and examination-negative groups were compared across a series of variables that included mean age, sex, race, median duration of hemodialysis dependence, cause of renal disease, and presence of diabetes mellitus. Two-sided P values were calculated using t-tests for age and chi-square analyses or Fisher's exact test for the categorical exposures. Because disease duration was not normally distributed, Wilcoxon's rank sum test was used to obtain P values. Similar analyses were performed to compare the patients in cohort 2 who had died within 24 months with those who had not died; these patient groups were compared further using a 24-month Kaplan-Meier survival curve and multivariate Cox proportional hazards regression models.
Two different Cox models were used to predict mortality at 6, 12, 18, and 24 months. The first model showed the crude relationship, accounting for time to death. The second model adjusted for sex, age group, race (white versus nonwhite), duration of hemodialysis, and presence of diabetes mellitus. Both models used a statistical significance level of 0.05 and were assessed for proportional hazards assumption using a multiplicative interaction term between skin examination result and time to death. Patients with and those without records of exposure to gadolinium-containing contrast agents during imaging studies prior to the skin examination were compared using logistic regression to calculate an odds ratio (OR), a relative risk (RR), and 95% confidence intervals (95% CIs). SAS statistical package version 9.1 was used to conduct the analyses (SAS Institute, Cary, NC).
Reproducibility of the skin examination.
All 32 patients available to be in cohort 1 were eligible for the study, and 30 (94%) consented to participate. When determining whether a patient was examination-positive or examination-negative for skin changes characteristic of NSF, there was 97% agreement between DJT and JK, a senior rheumatologist with experience evaluating patients with NSF. Agreement was 90% between JK and AK, a physician who had no prior experience examining patients with NSF. According to JK, DJT, and AK, respectively, 6 (20%) of 30 patients, 7 (23%) of 30 patients, and 5 (17%) of 30 patients from cohort 1 had skin examination findings that were consistent with NSF.
Prevalence of cutaneous changes of NSF.
Of the 204 patients available to be in cohort 2, only 7 (3%) were ineligible based on exclusion criteria. Of the remaining 197 patients, 186 (94%) consented to participate (23–60 patients per site). The demographic characteristics for all patients are listed in Table 1. Cohorts 1 and 2 differed only in that cohort 1 had more patients who were African American, a slightly younger mean age, and more patients with glomerulonephritis as the etiology of their renal disease, as compared with cohort 2.
Table 1. Characteristics of the patients with and those without cutaneous changes of nephrogenic systemic fibrosis*
The extremities of each patient in cohort 2 were examined for 3 cutaneous changes characteristic of NSF: hyperpigmentation, hardening, and tethering of skin. Of the 186 patients, 108 had no evidence of any of these changes. Hyperpigmentation was present in 43 patients (23%), and it was the only skin change in 24 of these patients. Hardening was present in 36 patients (19%), and it was the only skin change in 15 of these patients. Tethering was present in 36 patients (19%), and it was the only skin change in 14 of these patients.
Of the 186 patients in cohort 2, 25 (13%) were examination-positive, with 2 or 3 cutaneous changes consistent with NSF. Of these patients, 3 had hyperpigmentation and hardening alone, 4 had hyperpigmentation and tethering alone, 6 had hardening and tethering alone, and 12 demonstrated all 3 skin changes. None of the demographic variables tested between examination-positive and examination-negative patients differed significantly (Table 1). Skin changes were not associated with the hemodialysis center or the shift during which hemodialysis was performed (data not shown). Detailed data for examination-positive patients are listed in Table 2.
Table 2. Patients with cutaneous changes of nephrogenic systemic fibrosis*
We obtained results of skin biopsies that were available for 1 and 4 of the examination-positive patients from cohort 1 and cohort 2, respectively. All skin biopsy specimens were consistent with NSF, thus confirming the clinical diagnosis. Three patients with biopsy-proven NSF had all 3 skin findings, whereas the other 2 patients had skin hardening and tethering. The extremity of a patient with histologically proven NSF and all 3 skin changes is shown in Figure 1.
Association of increased mortality with cutaneous changes of NSF.
Of the 186 patients in cohort 2, 45 (24%) died within 24 months of skin examination. Of all demographic data, only diabetes mellitus was significantly associated with mortality (Table 3). Twenty-four–month mortality was 48% and 20% in patients with and those without cutaneous changes of NSF, respectively (Table 4 and Figure 2). Factoring for time to death, the crude hazard ratio (HR) of 24-month mortality was 3.1 (95% CI 1.6–6.0) (Table 4). After adjusting for age, sex, race, duration of hemodialysis, and presence of diabetes mellitus, the adjusted HR was 2.9 (95% CI 1.4–5.9). The significantly increased crude and adjusted HRs were apparent as early as 6 months following skin examination. Mortality was not associated with the hemodialysis center or the shift during which hemodialysis was performed (data not shown).
Table 3. Association of increased 24-month mortality with cutaneous changes of NSF*
Alive (n = 141)
Dead (n = 45)
OR (95% CI)
Except where indicated otherwise, values are the number (%). NSF = nephrogenic systemic fibrosis; OR = odds ratio; 95% CI = 95% confidence interval; HD = hemodialysis.
By 2-tailed t-test for independent samples.
By chi-square test of independence.
By Fisher's exact test.
By Wilcoxon's rank sum test.
Categories are not mutually exclusive; some patients had more than 1 cause of renal failure.
Except where indicated otherwise, values are the number (%). 95% CI = 95% confidence interval.
At indicated time intervals following skin examination, examination-positive patients versus examination-negative patients.
Adjusted for sex, age group, duration of hemodialysis (4 groups), race (white versus nonwhite), and presence of diabetes mellitus.
Cardiovascular causes accounted for 7 (58%) of 12 deaths and 16 (48%) of 33 deaths in patients with and those without cutaneous changes of NSF, respectively. No other single cause of death predominated in either group. Causes of death for patients with cutaneous changes of NSF included cardiac arrest (n = 3), myocardial infarction (n = 2), cardiac dysrhythmia (n = 1), acute pericarditis (n = 1), pulmonary actinomycosis (n = 1), septicemia (n = 2), bladder cancer (n = 1), and unspecified debility (n = 1). Causes of death for patients without cutaneous changes of NSF included cardiac arrest (n = 9), myocardial infarction (n = 4), cardiac dysrhythmia (n = 1), heart failure (n = 1), intracerebral hemorrhage (n = 1), respiratory arrest (n = 4), aspiration pneumonitis (n = 1), bacterial pneumonia (n = 1), septicemia (n = 1), endocarditis (n = 1), lung cancer (n = 3), oral neoplasm (n = 1), cerebral edema (n = 1), spinal cord disease (n = 1), bleeding peptic ulcer (n = 1), unspecified debility (n = 1), and unattended (n = 1).
Association of gadolinium exposure with cutaneous changes of NSF.
The association between exposure to gadolinium-containing contrast agents and cutaneous changes of NSF was assessed in the subgroup of 90 (49%) of the 186 patients in cohort 2 for whom data were available in an active EMR. These 90 patients included 17 (68%) of 25 examination-positive patients and 73 (45%) of 161 examination-negative patients. All identified imaging studies had been performed at PHS imaging facilities, where gadopentetate dimeglumine was used as the gadolinium-containing contrast agent.
Patients who had undergone gadolinium-enhanced imaging studies had an OR of 14.7 (95% CI 1.9–117.0) for developing cutaneous changes of NSF. Sixteen (30%) of 54 patients who previously had undergone gadopentetate dimeglumine–enhanced imaging studies developed cutaneous changes of NSF, compared with only 1 (3%) of 36 patients who did not have gadopentetate dimeglumine exposure (RR 10.7, 95% CI 1.5–76.9) (Table 5). In contrast, 16 (94%) of 17 patients in whom cutaneous changes of NSF developed had previously been exposed to gadopentetate dimeglumine, as compared with 38 (52%) of 73 patients in whom these skin changes did not develop (RR 1.8, 95% CI 1.4–2.3). None of the individual demographic variables differed significantly between the entirety of cohort 2 and the subgroup of cohort 2 with data available in an active EMR. Mortality 24 months after skin examination in this subgroup remained significantly increased for patients with cutaneous changes of NSF (data not shown).
Table 5. Association of exposure to gadolinium-containing contrast media during imaging studies and cutaneous changes of NSF*
Results of skin examination for NSF
Exposure (n = 54)
No exposure (n = 36)
Values are the number (%) of patients. NSF = nephrogenic systemic fibrosis.
Positive (n = 17)
Negative (n = 73)
This study is the first to establish the prevalence of cutaneous changes characteristic of NSF in patients receiving hemodialysis. We used a simple, reproducible, noninvasive skin scoring examination that assigned 1 point for each of 3 skin changes: hyperpigmentation, hardening, and tethering. This examination allowed us to screen large numbers of hemodialysis patients for cutaneous evidence of NSF. Data from cohort 1 demonstrated that the scoring system is highly reproducible. In cohort 2, we found that 13% of 186 outpatients receiving hemodialysis accrued a skin score of 2 or 3, suggesting the presence of NSF in these patients.
In this population of patients with chronic kidney disease who were receiving hemodialysis, our 3-component skin examination identified those who had a greater likelihood of dying within 6, 12, 18, and 24 months following skin examination. This cohort study also quantifies for the first time the association between cutaneous changes of NSF and exposure to gadolinium-containing contrast material in a cross-sectional cohort study; the presence of 2 or 3 skin changes correlated with a nearly 15-fold greater likelihood of having undergone a gadolinium-enhanced MRI study prior to our examination. This finding is consistent with previous studies that identified prior exposure to gadolinium-containing contrast agents in patients with NSF (8–12).
In the absence of characteristic clinical features, skin biopsy alone is insufficient to diagnose NSF, because histologic changes of NSF are difficult to distinguish from those of scleromyxedema, another fibrosing skin condition (19). In our study, skin biopsy data were available for 5 of the patients with skin scores of 2 or 3 (1 patient in cohort 1 and 4 patients in cohort 2). Each biopsy specimen demonstrated histologic evidence of NSF, lending validity to our skin scoring system. We intentionally limited our analysis for NSF to a noninvasive skin examination so as to maximize the number of patients who would participate and to avoid the less than trivial infectious complications of a full-thickness (often lower extremity) skin biopsy (20, 21), especially given that the patients in this study universally had renal failure and a high prevalence of skin fibrosis and diabetes mellitus, which are known risk factors for surgical wound infection and poor wound healing (22–24). By not requiring a skin biopsy, we were able to achieve a 94% participation rate and to minimize selection bias. Had consent to skin biopsy been required for study entry, those patients who were concerned that they had skin changes might have participated preferentially, thereby overestimating the prevalence of NSF. The paucity of available skin biopsy specimens highlights that NSF likely is underrecognized by many practicing physicians.
The unexpectedly high occurrence of skin changes consistent with NSF might raise concern that our skin scoring system overestimates the prevalence of NSF by detecting other fibrosing skin conditions, such as scleroderma, scleromyxedema, or eosinophilic fasciitis. It is unlikely that these rare conditions are prevalent enough to significantly affect our results. Chronic venous insufficiency (CVI), which almost exclusively affects the lower extremities, can lead to pitting edema, hyperpigmentation, and other skin changes (25). However, pitting edema causes neither hardening nor tethering of the skin and thus does not alter the score in our 3-component skin examination. Severe CVI can evolve into lipodermatosclerosis, a condition characterized by fibrosis of the dermis and subcutaneous tissue of the legs (25). However, lipodermatosclerosis does not adequately explain the skin changes of many patients in this study. Of the 25 patients with skin changes of NSF, 12 (48%) had skin involvement of both the upper and lower extremities or joint contractures on affected extremities, neither of which can be explained solely on the basis of CVI or lipodermatosclerosis.
Recent case–control studies have implicated gadolinium in the pathogenesis of NSF (8–12). Our study, however, is the first to address this association by using a cross-sectional population-based approach. In our subgroup analysis of the 90 patients in cohort 2 with an active PHS EMR, we may have underestimated the number of patients with previous gadolinium exposure by assuming that the 36 patients with no record of a gadolinium-enhanced imaging study had not received prior gadolinium-containing contrast agents. However, it is exceedingly unlikely that all but 1 of these 36 patients would cluster into the examination-negative group; only 1 patient with cutaneous changes of NSF lacked documented gadolinium exposure. If indeed this patient had never received gadolinium-containing contrast material for any imaging study, then other factors besides gadolinium might also be sufficient to cause NSF. Gadolinium has been identified in skin biopsy specimens from patients with NSF who previously underwent gadolinium-enhanced MRI (26). However, the mechanisms by which gadolinium might lead to fibrosis in NSF have not been elucidated. Reported data conflict as to whether the serum bicarbonate level or acidemia plays a role (8, 9). Additional studies are needed to understand why in many patients receiving hemodialysis, NSF develops after exposure to gadolinium-containing contrast material. Until then, gadolinium-containing contrast media should be used only with extreme caution in patients with chronic kidney disease.
This is the first cross-sectional population-based study to suggest an increased mortality risk for patients with NSF. Twenty-four–month mortality was significantly greater in patients with cutaneous changes of NSF, as compared with those without such changes (48% and 20%, respectively). The majority of the increased mortality for patients with cutaneous changes of NSF occurred in the first 6 months following skin examination, suggesting that established NSF is associated with increased early mortality. In a case series of 12 patients with NSF, 3 patients (25%) died within 2 years of symptom onset (27). One might postulate that indication bias explains the association between NSF and increased mortality; patients with high cardiovascular risk might be more likely to undergo a gadolinium-enhanced vascular study and therefore have an increased frequency of NSF compared with patients with low cardiovascular risk. However, cardiovascular causes of death in our study were not overly represented in patients with cutaneous changes of NSF. Further study of a larger cohort of patients with NSF will be necessary to identify factors that contribute to the higher rate of early death among affected individuals and to determine whether NSF is directly implicated as the cause of death.
Our study suggests that NSF, a complication of renal failure that has emerged over the past decade, is prevalent. Although seemingly underrecognized, NSF appears to be an emerging epidemic in patients receiving hemodialysis. Most notably, the presence of at least 2 abnormal findings in our skin examination suggestive of NSF identifies patients with an adjusted 4-fold increased risk of dying within 6 months. To assess for NSF, the routine evaluation of hemodialysis patients should include a careful skin examination. In those patients in whom clinical evidence of NSF is equivocal, a deep skin biopsy may be performed to support this diagnosis. The identification of larger numbers of patients with NSF will allow further investigations into the pathogenesis, treatment, and prevention of this recently described debilitating, and potentially fatal, condition.
Dr. Kay had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study design. Todd, Kagan, Chibnik, Kay.
Acquisition of data. Todd, Kagan, Kay.
Analysis and interpretation of data. Todd, Kagan, Chibnik, Kay.
We acknowledge Norma Ofsthun, PhD, Ming Teng, MD, MS, Eduardo Lacson, Jr., MD, MPH, and J. Michael Lazarus, MD, of Fresenius Medical Care North America for their cooperation. We also thank Jeffrey N. Katz, MD, MS, for methodologic advice and Morton N. Schwartz, MD, for helpful suggestions in the preparation of this manuscript.