Nephrogenic systemic fibrosis (NSF) was first described by Cowper et al in 2000,1 when they reported on 15 renal dialysis patients dating back to 1997 who had developed a scleromyxedema-like condition characterized by thickening and hardening of the skin of the extremities. Nephrogenic systemic fibrosis is a rare but serious disease that affects primarily the skin and joints and less commonly can involve the internal organs (eg. heart, lungs, liver and muscles).
The first link between NSF and gadolinium-based contrast agents (GBCAs) was suggested in 2006, when Grobner noted the development of skin symptoms in 5 patients with end-stage renal disease 2 to 4 weeks after they underwent contrast-enhanced magnetic resonance angiography with Omniscan.2,3 Subsequent case analyses have demonstrated that most patients who develop NSF do so within 3 to 6 months of GBCA exposure.4
Nearly all NSF patients experience severe renal impairment (estimated glomerular filtration rate [eGFR] <30 mL/min/1.73 m2), usually requiring dialysis or renal transplantation, or acute kidney injury. Interestingly, the converse is not true: most patients with severe renal impairment who receive even multiple doses of GBCAs do not develop NSF, implying the presence of additional risk factor(s) necessitating development of the disease. Multiple cofactors for the development of NSF have been proposed including metabolic acidosis,2 vascular surgery,2 or treatment with erythropoietin,5 although no definitive or comprehensive list of such underlying etiologies is yet understood.
The European Medicines Agency (EMA)6 and the European Society of Urogenital Radiology7 defined risk categories for GBCAs with regard to NSF, classifying the macrocyclic GBCAs Gadovist (brand name in United States: Gadavist), Dotarem, and ProHance as having the lowest risk for NSF development. In May 2007, the Food and Drug Administration (FDA) issued a boxed warning in the product labeling of all GBCAs, highlighting the risk of NSF in patients with acute/chronic severe kidney disease (GFR <30 mL/min/1.73 m2) or acute kidney injury.8
One key criterion in the risk classification of the different GBCAs by the authorities was the number of single-agent (“unconfounded”) reports.6,8 Moreover, the ACR Manual for Contrast Media, which is widely used by radiologists as a basis for practice guidelines, lists single-agent reporting as their underlying rationale for eGFR testing requirements.9 Therefore, this analysis aims to provide further insights into these report numbers, which are frequently also cited for regulatory as well as promotional purposes, and to discuss methodological aspects that are of potential relevance for the appropriate interpretation of these numbers. In particular, the following aspects were considered:
- Year of market introduction and market share of specific GBCAs in the country from where the reports originated;
- Additional evaluation of multiagent reports with respect to temporal relationship to NSF onset and a critical discussion of the current practice to solely consider single-agent reports;
- Discussion of potential sources for variability in reporting standards of different companies using the Cowper/Girardi criteria as the basis for assessment of individual reports of NSF.
MATERIAL AND METHODS
The dataset comprised all reports received by Bayer between 2006 and 2016 globally. Duplicate reports and reports with no NSF diagnosis were not included.
Bayer's Pharmacovigilance Department receives reports of potential adverse drug reactions for all of the company's products globally. In general, adverse drug reaction reports are received from health care professionals, regulatory authorities, patients/consumers, as well as from searches of the scientific literature.
For NSF in particular, the company received most of its reports related to litigation procedures in the United States. Lawsuits would frequently name the manufacturers whose products were more likely to be possibly used, given the historically poor documentation of the specific brand of GBCA administered.
The pharmacovigilance department comprehensively evaluated all incoming reports for documentation pertaining to GBCA use (including brand, dose, indication, and date of administration), clinical symptoms, laboratory values, and histopathology reports. Repeated queries on all incomplete or missing information were sent to the reporting source using a targeted questionnaire, or in the case of litigation-related reports, the information was sought via subpoena of the patients' medical records.
Once all available information was gathered, each report was categorized as a “single-agent” or “unconfounded” report7 if it could be confirmed that no additional brand of GBCA was administered to the patient or as “multiagent” or “confounded” report7 if more than 1 brand of GBCA was documented.
Finally, Bayer pharmacovigilance personnel used the information obtained from biopsy reports, detailed history of GBCA administration, medical history, and detailed history of the emergence of clinical signs and symptoms, to determine whether the case met the established Cowper/Girardi clinico-histopathological criteria for a diagnosis of NSF (Cowper/Girardi positive reports).10 Although Broome et al4 suggested that most NSF occurs within 3 months of the last GBCA administration, Bayer conservatively considered an 18-month onset latency as positive for a temporal association.
In the case of unclear or incomplete information, the most conservative assumption was applied to give the original reporter the benefit of the doubt and to not underestimate the number of cases. It is important to note that this methodology developed and implemented by Bayer is not necessarily the industry standard; there might be legitimate but slightly different processes or criteria applied by other institutions or companies.
Because of privacy regulations and internal Bayer policies, the company is not allowed to disclose individualized pharmacovigilance data to noncompany personnel. Therefore, an independent internal audit of all cases referenced in this manuscript was conducted for data verification.
To estimate global use, annual market shares of GBCAs were obtained from Arlington Medical Resources, USA,11 and complemented with additional sales data from Bayer and IMS.
A simulation was performed according to Michel and Blenk12 based on the market shares of the 5 available GBCAs for the years 1997 to 2006. Data for 100,000 patients was simulated such that the number of exposures per patient followed a Poisson distribution with a mean of 4 exposures. Time points of exposures reflected the increase in total numbers of exposures across the 10 years by means of a multinomial distribution. Days in the respective years were drawn randomly from a uniform distribution. The treatments were determined using a multinomial distribution reflecting the market shares of the products for the respective year. In cases of a second exposure within 365 days, the previous treatment had an increased chance of 80% to be drawn again, following the approach by Michel and Blenk.12
The dataset analyzed consisted of 779 reports of potential NSF cases received from 2006 to 2016. The earliest disease onset dated back to 1996, based on undocumented patient recollection. A total of 325 reports involved 1 Bayer product only, 208 reports involved only noncompany or unknown products, and 246 reports involved both Bayer products as well as GBCAs manufactured by other companies. Most of Bayer's reports involved the linear ionic GBCA Magnevist (n = 563), followed by the non-Bayer linear nonionic GBCA Omniscan (n = 322) (Fig. 1).
After more than 145 million administrations of Magnevist (December 31, 2017, >150.9 million), 563 reports were received in which Magnevist administration was reported. Of these, 343 were multiagent reports and 220 were single-agent reports. Seventy-four of the single-agent reports met the Cowper/Girardi criteria as being “consistent with” or “diagnostic of” NSF (Fig. 2).
After more than 32 million administrations of Gadovist (December 31, 2017, >47.3 million), 13 reports were received with reported administration of Gadovist. Eight of these were multiagent reports and 5 were single-agent reports. Three of the single-agent reports were diagnostic of/consistent with NSF based on the Cowper/Girardi criteria (Fig. 3).
After more than 3.6 million administrations of Primovist (Eovist in the United States) (December 31, 2017, >4.6 million), no report of NSF or NSF-like symptoms has been received from any source.
Most of Magnevist NSF reports came from 3 countries: United States, Germany, and Denmark. Most of all reports (485/563 [86%]), of single-agent reports (191/220 [87%]), and of Cowper/Girardi positive reports (68/74 [92%]) were from the United States. More than half of all US reports were litigation related.
In comparison with the United States, other countries with large market shares of Magnevist during this same time period had very few reports. Notably, only 6 single-agent Cowper/Girardi positive reports arose from Europe and no reports originated from Canada, Japan, or other regions with large market shares of Magnevist (Table 1).
The number of Magnevist NSF reports increased steadily from 1996 to 2006 and declined sharply after the FDA box warning in 2007. In each year, most of all reports, single-agent reports, and Cowper/Girardi positive reports came from the United States. The last global disease onset was in 2012, the last in the United States, in 2010 (Fig. 4).
From 1996 to the end of 2006, Magnevist and Omniscan dominated the US market, with a combined market share of greater than 80%. In 2006, Magnevist had a market share of greater than 50%, and Omniscan, greater than 30%. The other US-approved GBCAs each had US market shares of 10% or lower, that is, ProHance, less than 6%; OptiMARK, less than 2%; and MultiHance, less than 1%. Neither Primovist, Gadovist, or Dotarem was marketed in the United States at that time (approved in 2008, 2011, and 2013, respectively). When looking at NSF disease onset, symptom onset peaked in 2006, 1 year before the introduction of the boxed warning, with 17 single-agent Cowper/Girardi positive reports recorded. Case reports with new onset fell drastically after awareness based on the Grobner publication2 and implementation of the FDA warning8 (Fig. 5).
Of the 563 reports with Magnevist administration, 343 were multiple-agent reports. In this group of multiagent reports, at least 150 reports were identified in which a noncompany product showed the closest temporal association to NSF onset. These reports were forwarded as a courtesy to the other manufacturers. In 93% of these reports, the product used in closest temporal relationship to the development of NSF was a different linear agent: Omniscan (n = 118), OptiMARK (n = 15), or MultiHance (n = 6). In 11 reports, a macrocyclic GBCA was administered in closest temporal association to the onset of NSF-like symptoms (ProHance, n = 6; Gadovist, n = 2; Dotarem, n = 3). Most reports were from the United States (Table 2). In some reports, the product administered in closest temporal association to NSF onset remained unknown.
Tables 3 to 5 show Magnevist cases where another GBCA was administered in closest temporal association to onset of symptoms in detail. Most patients developed NSF within 6 months or less after administration of this other GBCA: For Omniscan, these were 94 of 118 (79.7%); for OptiMARK, 9 of 15 (60%) (Table 3); and for MultiHance, 3 of 6 (50%) (Tables 3 and 4). For the 11 cases where a macrocyclic agent was administered in closest temporal association, in 5 (45%) cases, the NSF onset occurred within 6 months of macrocyclic administration (Tables 3 and 5). In comparison, 41 of 74, or 55%, of Magnevist single-agent Cowper/Girardi positive patients experienced NSF within 3 months of the last Magnevist administration.
Of 100,000 simulated patients, 73,426 (73.4%) patients were observed in a confounded setting, whereas 26,574 (26.6%) were observed in an unconfounded setting. Of 84,752 simulated patients who were treated with Magnevist, 80.2% were in a confounded setting. Products with single digit market shares showed more than 94% of their NSF reports in a confounded setting (Table 6).
After the accumulation of 10 years of knowledge on NSF, Bayer performed an in-depth analysis of all reported NSF cases and established a contextual framework through which these cases could be discussed.
The goal of this analysis was to incorporate less widely discussed aspects for NSF risk assessment, such as (1) the year of market introduction and the market share of the GBCAs, (2) the interpretation of multiagent reports with special focus on the sequence of previous administration of other GBCAs, and (3) potential differences in evaluation and reporting standards by different companies.
First and most basically, NSF report numbers need to be seen in the context of utilization rate and dosing: Magnevist was the first and for a time the only GBCA available worldwide. Magnevist had the highest market share (>50%) in the United States until May 2007 and also some years beyond. In the 74 Cowper/Girardi positive reports (Fig. 2), the severely renally impaired patients received multiple (up to 19) and/or high doses (with individual doses as high as 0.62 mmol/kg) of the contrast medium, sometimes for unapproved indications.
Until 2007, other GBCAs than Magnevist and Omniscan had a comparably low US market shares (ProHance, Optimark, and MultiHance—all below 10%) or were not approved in the United States at that time (Primovist, Gadovist, and Dotarem) (Fig. 5).
As is confirmed by a simulation model, the different numbers of NSF cases for the different GBCAs are not unexpected, given the very low prevalence of NSF. Because of the consistently high market share of higher than 50% over the simulated 10 years, Magnevist was the agent that was observed most often in an unconfounded setting (63.1% of all unconfounded cases). Especially agents with a late introduction into the market (MultiHance, Primovist, Gadovist, and Dotarem) and/or a low market share (Optimark and ProHance, both <10%) had a markedly lower chance to be observed in an unconfounded setting of 2 or more exposures.
Therefore, any comparison of report numbers of GBCAs or “ratios of reports/administrations” over time since market introduction is significantly biased and hence misleading, as this would not take into account the market presence in different countries before and after May 2007. Therefore, those numbers need to be used with appropriate and utmost caution and are not suited for reliable differential risk assessments between GBCAs.
The reports that Bayer has received involving noncompany GBCAs are presumed to encompass a small sample of reports for those products. But in general, data published by the health authorities are supportive of the trends seen in the Bayer internal data. An EMA report published in 20106 and by the FDA in 20118 list single-agent reports of various GBCAs in relation to total administrations. According to the FDA's table (January 21, 2011) with US data through December 2009, more than 105 million administrations of Magnevist triggered 179 single-agent reports, whereas for Omniscan, more than 49 million administrations triggered 505 single-agent reports. Two single-agent reports were each listed for MultiHance and ProHance. Gadavist and Dotarem were not on the US market at that time.8 The EMA's table shows a similar pattern.6
The second and probably most significant criticism with the current reporting standard is the limitation to single-agent cases. Nephrogenic systemic fibrosis occurs almost exclusively in multimorbid patients who received multiple doses of GBCAs spread over numerous magnetic resonance imaging (MRI) examinations. Statistically speaking, patients receive, on average, 3.5 MRI exams over their lifetime,13 and therefore, many patients who received an MRI after 2006, with high likelihood, would have had an MRI with Magnevist or Omniscan before that. Consequently, the likelihood of a single-agent NSF case with a new agent would be exceedingly rare. And of course, most of the use of these more recently approved or less used agents occurred after a potential association of GBCAs and NSF was identified, and therefore, safer clinical practices had already gone into effect (including discontinuation of high dosing, repeated use, and off-label procedures in at-risk patients).
Although the inclusion of multiagent cases certainly adds a level of variability, it is standard pharmacovigilance practice to consider temporal relationship in any causality assessment. The data from Bayer's case reports are supportive of this approach (Tables 4 and 5), given the stark contrast between the time of onset of symptoms from administration of the last GBCA compared with the time since the last Magnevist administration.
Finally, it is worth noting that Bayer has taken a very conservative approach in its reporting and counting of cases of NSF, an approach that might differ among companies. Following the rules for spontaneous reporting, Bayer reported and counted all cases that were presented to the company as NSF whether or not that diagnosis was confirmed clinically or histopathologically. It has been suggested in previous publications that this approach may diverge from that taken by other companies.14 In addition, NSF is a very difficult diagnosis to make, and Bayer applied the Cowper/Girardi criteria10 very conservatively, giving “the benefit of the doubt” in many cases despite a lack of evidence that exactly matches these criteria. For example, although 1 of the clinical criteria is “marked induration/peau d'orange,’ Bayer assigned a score of 1 for any mention of “induration.” For the histopathological finding of “CD34+ cells with tram tracking,” Bayer assigns a score of 1 for any mention of either CD34+ cells OR tram tracking. For “thick and thin collagen bundles,” any mention of collagen bundles, thick or thin, scores a 1. The result is that even the subset of cases that Bayer considers to be possibly representative of NSF are very likely to be overestimated. In conclusion, lack of standardization results in potential variability and hence concluding on product/safety risk differences between GBCAs based on minimal numerical differences in unconfounded NSF reports is unscientific and should not be used for product differentiation or even misleading promotional purposes.
Several clinical studies performed in renally impaired patients confirmed the low risk of developing NSF after administration of Magnevist (data on file), Gadovist,15,16 Primovist,17 MultiHance,18–20 Dotarem,21 and ProHance.18
In conclusion, this comprehensive analysis highlights numerous aspects that are insufficiently reflected in the existing GBCA risk classification systems. Pure report numbers are probably the weakest criterion to compare the true risk for NSF as long as the above-mentioned aspects are not factored in. The currently published report numbers should be used with caution and are certainly not suited for commercially promoted risk differences.
This analysis demonstrates the limitations of relying on case reports, particularly single-agent reports, when assessing the risk of NSF associated with any GBCA. Given the reliance by both professional society guidelines and health authorities on these numbers, the authors propose bias minimization strategies including correction for market introduction year, market share, temporal association, and reporting standards. Comparative risk assessments should be primarily based on objective product parameters such as stability, pharmacokinetics, and dose.