The Five-Factor Score Revisited

Assessment of Prognoses of Systemic Necrotizing Vasculitides Based on the French Vasculitis Study Group (FVSG) Cohort

Abstract

The 1996 Five-Factor Score (FFS) for systemic necrotizing vasculitides (polyarteritis nodosa [PAN], microscopic polyangiitis [MPA], and Churg-Strauss syndrome [CSS]) is used to evaluate prognosis at diagnosis. In the current study we revisited the FFS, this time including Wegener granulomatosis (WG).

We analyzed clinical, laboratory, and immunologic manifestations present at diagnosis of systemic necrotizing vasculitides for 1108 consecutive patients registered in the French Vasculitis Study Group database. All patients met the American College of Rheumatology and Chapel Hill nomenclature criteria. Univariable and multivariable analyses yielded the 2009 FFS for the 4 systemic necrotizing vasculitides.

Overall mortality was 19.8% (219/1108); mortality for each of the SNV is listed in descending order: MPA (60/218, 27.5%), PAN (86/349, 24.6%), CSS (32/230, 13.9%), and WG (41/311, 13.2%) (p < 0.001). The following factors were significantly associated with higher 5-year mortality: age >65 years, cardiac symptoms, gastrointestinal involvement, and renal insufficiency (stabilized peak creatinine ≥150 μmol/L). All were disease-specific (p < 0.001); the presence of each was accorded +1 point. Ear, nose, and throat (ENT) symptoms, affecting patients with WG and CSS, were associated with a lower relative risk of death, and their absence was scored +1 point (p < 0.001). Only renal insufficiency was retained (not proteinuria or microscopic hematuria) as impinging on outcome. According to the 2009 FFS, 5-year mortality rates for scores of 0, 1, and ≥2 were 9%, 21% (p < 0.005), and 40% (p < 0.0001), respectively.

The revised FFS for the 4 systemic necrotizing vasculitides now comprises 4 factors associated with poorer prognosis and 1 with better outcome. The retained items demonstrate that visceral involvement weighs heavily on outcome. The better WG prognosis for patients with ENT manifestations, even for patients with other visceral involvement, compared with the prognosis for those without ENT manifestations, probably reflects WG phenotype heterogeneity.

Abbreviations ANCA = antineutrophil cytoplasmic antibodies, BVAS = Birmingham Vasculitis Activity Score, CNS = central nervous system, CSS = Churg-Strauss syndrome, ENT = ear, nose, and throat, FFS = Five-Factor Score, FVSG = French Vasculitis Study Group, HBV = hepatitis B virus, MPA = microscopic polyangiitis, NS = not significant, PAN = polyarteritis nodosa, SNV = systemic necrotizing vasculitides, WG = Wegener granulomatosis.

INTRODUCTION

The Five-Factor Score (FFS), a tool designed to assess prognosis at diagnosis of polyarteritis nodosa (PAN), Churg-Strauss syndrome (CSS), and microscopic polyangiitis (MPA), was based on 342 patients entered in the French Vasculitis Study Group (FVSG) database as reported in 1996.¹¹ It is now routinely used by clinicians to evaluate the severity of these systemic necrotizing vasculitides (SNV). The FFS can also be applied in some circumstances to help choose first-line therapy, and, more specifically, to determine the need for immunosuppressants when factors of poor prognosis are present.³⁵ The 1996 FFS included the following parameters that were associated with higher risk of death: proteinuria >1 g/d, renal insufficiency (stabilized peak creatinine 140 μmol/L), cardiomyopathy, severe gastrointestinal manifestations, and central nervous system (CNS) involvement. For FFS = 0, 1, and ≥2, the respective 5-year mortality rates were: 12%, 26%, and 46%, respectively.

Other scores have been devised to assess vasculitis evolution or damage, such as the Birmingham Vasculitis Activity Score (BVAS),²⁷ the Disease Extent Index,⁵ or the Vasculitis Damage Index.⁷ In contrast, the sole primary objective of the FFS was to predict survival of patients according to clinical and biologic parameters observed at the time of diagnosis, independent of treatments prescribed, flares, or events that could occur during the course of the disease; it was subsequently extended to adapt treatment to disease severity.

Ten years after the initial publication, our cohort has grown and we decided to reevaluate the FFS and again determine the parameters predictive of poor outcome for the same 3 SNV, and, for the first time, to include Wegener granulomatosis (WG), which had not been evaluated previously. To do so, we analyzed the characteristics of the 1108 SNV patients entered into the FVSG database and managed by FVSG members between 1957 and 2005.

PATIENTS AND METHODS

The 1108 patients had SNV satisfying the American College of Rheumatology criteria, for PAN,²⁵ CSS,²⁹ and WG,²⁴ and also the Chapel Hill nomenclature criteria for those vasculitides and MPA.¹⁷ Diagnoses were biopsy-proven for 849 patients, based on clinical features and renal or abdominal angiography for 125 patients, and relied only on clinical manifestations considered characteristic of vasculitis for 134 patients.

Patients included in this analysis were treated according to good clinical practice at the time of their diagnoses or were included in prospective therapeutic trials organized by the FVSG or the European Vasculitis Study Group in which French patients had been included. Because the study objective was to identify prognostic factors at the time of diagnosis, only symptoms present at SNV onset were considered.

The revised, 2009 FFS was tested for the 4 SNV together, then for each vasculitis alone, to isolate poor-prognosis factors that could be specific to 1 disease and not the others.

Definitions

Score parameters were defined as follows:

Renal insufficiency: creatinine ≥150 μmol/L, measured at its stabilized peak level;

Cardiac insufficiency: based only on the presence of its clinical symptoms (for example, pulmonary edema), excluding laboratory values for biologic parameters, like brain natriuretic peptide, that were not determined for the majority of patients, and, in the absence of clinical symptoms, echocardiography or other complementary investigations were not considered for diagnosis of cardiac insufficiency;

Only severe gastrointestinal involvement: that is, bowel perforation, bleeding, and pancreatitis, with neither appendicitis nor cholecystitis being integrated into the score when it was the sole vasculitis manifestation or found fortuitously in a histologic specimen;

Ear, nose, and throat (ENT) involvement: the presence of clinical symptoms confirmed by physical examination by an ENT specialist and by computed tomography scan. When ENT symptoms were not present, specialized investigations and scans were not systematically performed.

Clinical symptoms comprising the score were directly attributable to active vasculitis. For example, when cardiac insufficiency was associated with preexisting arterial hypertension or when gastrointestinal bleeding was drug-induced, the symptoms were not taken into account to establish the score.

Statistical Analyses

For univariable analyses, standard parametric methods were applied, that is, the chi-square or Fisher exact test for small numbers, when appropriate, was used to compare qualitative variables and the Student t test was used to compare quantitative variables expressed as means (± standard deviation). Mortality rates and hazard ratios were calculated for every item that could be a predictor of death. The variables tested were age, sex, weight loss, myalgia, arthralgia, polyarthritis, mononeuritis multiplex, skin manifestations, glomerulonephritis, proteinuria <1 and ≥1 g/d, microscopic hematuria, renal insufficiency, gastrointestinal tract involvement, severe gastrointestinal tract symptoms, digestive tract surgery, cardiomyopathy, asthma, pleural effusion, orchitis, pericarditis, arterial hypertension, ophthalmologic manifestations, inflammatory signs characterized by erythrocyte sedimentation rates <40 and ≥40 mm 1st h or C-reactive protein, and hepatitis B virus (HBV) infection. We did not test comorbidity factors, for example, diabetes or coronary artery disease, which were not within the scope of this analysis specifically devoted to symptoms of the 4 SNV.

Patient survival was assessed by life-table analysis using the Kaplan-Meier method.¹⁹ Survival was calculated from the date of vasculitis diagnosis to the censoring date or the time the database was closed for this study. We evaluated 5-year mortality as a function of the demographic, clinical, and laboratory parameters recorded at vasculitis diagnosis, and Cox proportional hazards models⁴ were fitted to examine the individual and combined effects of these variables.

The effect of continuous variables was analyzed as continuous covariates but we determined thresholds for age and creatininemia with receiver operating characteristics curves to transform these variables into categorical variables providing the best discrimination. To establish a score simple to use in clinical practice, we did not weight each item according to severity or different outcomes. The multivariable model included all the parameters with p ≤ 0.10 in univariable analysis and used a backward-selection procedure. We calculated the 1996 FFS with this method then adjusted with the new significant variables.

For all statistical analyses, p < 0.05 was considered to be significant. All confidence intervals were calculated at the 95% level. Statistical analyses were computed using the SAS statistical package, version 8.2 (SAS Institute Inc., Cary, NC).

RESULTS

The patient population comprised 349 patients with PAN (including 122 HBV-infected), 311 with WG, 230 with CSS, and 218 patients with MPA. The mean patient age at diagnosis was 53.5 years. Six hundred twenty-five patients were male and 483, female (M/F sex ratio, 1.3). Their main clinical and biologic manifestations observed at the time of diagnosis are detailed in Table 1.

TABLE 1: Main Clinical Manifestations of Systemic Necrotizing Vasculitides Included in the 2009 FFS Calculation

Mortality was evaluated for the entire patient population and for each of the SNV. The mortality rate was the highest for MPA (60/218, 27.5%), followed by PAN (86/349, 24.6%), CSS (32/230, 13.9%), and WG (41/311, 13.2%) (p < 0.001) (Figure 1). For the PAN group, HBV infection influenced mortality: 42/122 (34.4%) for HBV-PAN versus 44/227 (19.4%) for PAN without HBV (p < 0.002) (Figure 2).

FIGURE 1: Kaplan-Meier survival estimations for each systemic necrotizing vasculitis (p < 0.001 between curves).

FIGURE 2: Kaplan-Meier survival curves for PAN with and without HBV (p < 0.002).

We assessed mortality during 3 periods: 1953-1974, 1975-1989, and 1990-2005. Between 1957 and 1974, 33 patients were registered in the database and 16 (48.5%) have since died. From 1975 to 1989 the cohort included 254 new patients, among whom 92 (36.2%) have since died. During the most recent period, 1990-2005, 821 new patients joined the cohort and 111 (13.5%) of them have since died.

The multivariable analysis (Table 2) retained the following items as being significantly associated with a higher risk of mortality: age >65 years (see ROC curve, Figure 3), cardiac insufficiency, gastrointestinal involvement, and renal insufficiency (see ROC curve, Figure 4). The hazard ratio for each item and each of the SNV is reported in Table 3. Among kidney manifestations, proteinuria and/or microscopic hematuria were not associated with higher mortality: 330/890 (37%) survivors versus 98/218 (45%) nonsurvivors (not significant [NS]) had proteinuria, and 307/890 (34.5%) survivors versus 87/218 (39.9%) nonsurvivors (NS) had microscopic hematuria.

FIGURE 3: Receiver operating characteristics (ROC) curve for age (cutoff age = 65 yr).

FIGURE 4: Receiver operating characteristics (ROC) curve for creatininemia level (cutoff creatinine level = 150 μmol/L).

TABLE 2: Results of the Multivariable Analysis

TABLE 3: Global Prognostic 2009 FFS for 5-Year Mortality and Differences According to Each of the SNV

In addition to the symptoms associated with a poor prognosis, the presence of ENT involvement was evaluated as an independent factor significantly associated with a better outcome. Because ENT manifestations were observed mainly in WG and CSS, this prognostic factor was only relevant for these 2 SNV (Figures 5A, B).

FIGURE 5: Kaplan-Meier survival curves for (A) Wegener granulomatosis or (B) Churg-Strauss syndrome patients with ear, nose, and throat (ENT) involvement vs. those without, with respective between-curve differences of p < 0.01 and p < 0.03.

The impact of several symptoms on prognosis was also analyzed for each of the SNV considered in the study (Table 3). The following independent factors predictive of poor prognosis were retained: age and renal insufficiency for MPA; age and gastrointestinal involvement for PAN; age and cardiac insufficiency for CSS; and age, and renal and cardiac insufficiencies for WG. For the subgroup of patients with PAN, different impacts of organ involvement on mortality were observed according the presence or absence of HBV. Renal insufficiency (stabilized peak creatinine ≥150 μmol/L) was observed in 18.9% and 11.8% of the patients with and without HBV (p < 0.03), respectively. Severe gastrointestinal manifestations were also more frequent in PAN patients with HBV (50.4%) than without (31.1%) (p < 0.001).

ENT manifestations were an independent factor associated with improved survival of WG or CSS patients. According to Kaplan-Meier curves, the estimated 5-year mortality rates for patients with or without ENT involvement, respectively, were 23% vs. 43% for those with WG (p < 0.01) and 92% vs. 84% for those with CSS.

The 2009 FFS is composed of the following manifestations, which were accorded +1 point when present: age >65 years; cardiac insufficiency, renal insufficiency (stabilized peak creatinine ≥150 μmol/L) and gastrointestinal involvement. Because the presence of ENT manifestations was associated with a better prognosis, for practical reasons, their absence was scored +1 point. When the score was applied to our entire SNV population, the 5-year mortality rates for a 2009 FFS = 0, 1, and ≥2 were, respectively: 9%, 21% (p < 0.005), and 40% (p < 0.0001) (Figure 6A). The 5-year mortality rates calculated for the same SNV population using the 1996 FFS = 0, 1, and ≥2 were 12%, 18%, and 29%, respectively (Figure 6B). When applied individually to each of the SNV, the 2009 FFS reproduced the results established for the entire population (Figures 7A-D).

FIGURE 6: Kaplan-Meier survival curves for SNV patients according to their (A) 2009 FFS or (B) 1996 FFS, with respective between-curve differences of p < 0.001 and p < 0.01.

FIGURE 7: Kaplan-Meier survival curves for (A) polyarteritis nodosa, (B) microscopic polyangiitis, (C) Churg-Strauss syndrome, and (D) Wegener granulomatosis patients according to their 2009 FFS.

DISCUSSION

SNV are a heterogeneous group of diseases. Despite some shared characteristics, their differences have been clearly identified since classification systems and testing for antineutrophil cytoplasmic antibodies (ANCA) have become available. Notably, the outcome and prognosis of each of the SNV can be different. In the past, treatments were applied in a uniform manner, regardless of the vasculitis diagnosed. The results of therapeutic trials^12,15,33 showed that outcomes and prognoses vary among the 4 SNV groups.

In addition to those variable outcomes, disease severity also has proven therapeutic implications. Several scores have been designed and used in prospective trials,^27,36 mainly to follow the impact of treatment on the disease course. Establishing a uniform score for vasculitis assessment is not easy, and that was the BVAS objective. That score has been mainly studied in ANCA-associated vasculitides but needs to be validated for other groups of vasculitides. A universal assessment system was the OMERACT goal³⁰ but it too focuses mainly on ANCA-associated vasculitides and has to be extended to other vasculitides.

BVAS was shown to be correlated with the 1996 FFS³⁰ but has not yet been analyzed for the 2009 version. The 1996 FFS was intended to evaluate vasculitis severity at the time of diagnosis to predict outcome and, if possible, to help choose among therapeutic options. However, the 1996 FFS concerned only PAN, MPA, and CSS.¹¹

The 2009 FFS was calculated for those 3 SNV and, for the first time, WG was also included. Moreover, this new prognostic score has been validated on a larger patient population. Because vasculitis treatment regimens have become more effective, thereby improving outcomes, mortality decreased from 1953 to 2005. However, because only 33 patients were included during the cohort's first period (1953-1974), their impact on overall mortality and identification of prognostic factors was diluted in the >1000+ patients added since then. But it must be kept in mind that all 1108 patients included in this analysis received the best treatment available at the time of diagnosis, and that corticosteroids and immunosuppressants were already being prescribed in the 1970s. Moreover, the 50% mortality rate in the earliest period was not too bad for that era. Today, we have improved regimens and better adapted supportive care that have limited adverse events.

The FFS takes into account only clinical symptoms present at the time of diagnosis. Symptom changes over time are also parameters that clinicians consider in making first-line treatment decisions or modifying ongoing regimens. But those aspects are beyond the scope of this study. The creation of an evolving prognostic parameter, by integrating the progressive kinetics of symptoms initially present, would surely be useful and could achieve another objective that has not been obtained to date, to our knowledge.

Results of FVSG prospective trials previously demonstrated the therapeutic impact of the 1996 FFS.^3,9,35 In a retrospective study,⁸ we showed that the most severe forms of necrotizing PAN without HBV and CSS should be treated with combined steroids and cyclophosphamide. Conversely, our findings obtained in a prospective study³⁵ indicated that patients without poor-prognosis factors could be treated initially with steroids alone, and that only a quarter of them required a cytotoxic agent to control their SNV. This use of the prognostic 1996 score to guide treatment changes is not valid for WG. Published results of therapeutic trials based on the impact of treatment stratification using the 1996 FFS would favor treatment adapted for patients with CSS, PAN, or MPA without poor-prognosis factors. Survival of patients initially treated with corticosteroids alone was excellent. However, subsequent relapses were frequent, and then about half of the patients required the adjunction of an immunosuppressant to control their disease.³⁴ Therapeutic strategies based on prognostic factors will probably have to be adapted to some SNV and validated in other studies, also comparing the benefits/risks of the regimens prescribed.

The current revisitation analysis established the reliability of the 2009 FFS, compared to the 1996 version, despite the introduction and deletion of some items. When the 1996 FFS was applied to the present SNV population, no difference in the mortality rates was observed. Pertinently, the FFS is based on mortality and not on morbidity. Morbidity also has a major impact on outcome and quality of life, but it was not taken into account in the formulation of this score.

The cohort used to establish the score was amassed over several decades. During that time, treatments obviously changed and the general care of patients was certainly improved, which could have modified patients' general outcomes. Our objective was not to analyze the influence of treatments but to evaluate the impact of some clinical symptoms on mortality. Hogan et al¹⁴ evaluated the effect of cyclophosphamide on their patient cohort and found that odds ratios decreased when a cytotoxic agent was compared to steroids. A combination of steroids and another immunosuppressant is the optimal treatment for severe SNV. This combination needs to be assessed in patients with less severe SNV, as identified by prognostic scores, and ANCA-independent vasculitis in an attempt to achieve steroid-sparing. Because our analysis was made on a cohort of French patients, the 2009 FFS should be validated on other international cohorts to confer to this score a broader impact.

The 2009 FFS still comprises 5 factors and 5 points, but only 4 of them are associated with a poor prognosis (age, renal insufficiency, cardiac involvement, and gastrointestinal manifestations), each accorded +1 point, while the fifth factor, ENT manifestations, is associated with a better outcome, and their absence is scored +1 point. ENT symptoms should be a part of the FFS used only for patients with WG or CSS. ENT signs were also present in a minority of MPA patients, but their ENT involvement was usually minor and we did not consider it a significant marker of prognosis. The favorable impact of ENT manifestations on survival had been shown previously¹ and could embody the different pathogenic mechanisms at work in WG, probably pleading in support of different WG phenotypes.²²

Cardiac involvement, as in the 1996 FFS version, is still associated with a poor prognosis. Not surprisingly, it is more frequently present in CSS and it is the first cause of death for this SNV. One limitation is that the link between cardiac manifestations and vasculitis is indirect in some patients, based on the coexistence of both diseases and the absence of a cause other than vasculitis to explain the occurrence of cardiac manifestations.

Renal insufficiency is another factor associated with poor prognosis.³² In a 2007 study¹⁶ on patients with creatininemia >500 μmol/L, mortality was higher. According to the 2009 FFS, the death rate rose once the peak creatinine threshold (≥150 μmol/L) was reached. We were unable to obtain glomerular filtration rates for all patients, but the apparently moderate creatininemia increase should be considered as severely impaired renal function, when seen before treatment in newly diagnosed SNV patients who had generally lost weight and were sometimes elderly.

The mortality rate was higher for patients with HBV-PAN than for PAN patients without HBV. However, the poor-prognosis factor, as defined by multivariable analysis, was not HBV itself, but rather some symptoms associated with poor prognosis, like those comprising the FFS.^34,35 For HBV-PAN patients, they were renal insufficiency and gastrointestinal involvement, which are already FFS items.

Compared to the 1996 version, age is now included among the factors predictive of prognosis, reflecting the severity of vasculitis in the elderly;³¹ this probably argues for different therapeutic approaches, at least to limit the number and severity of infectious side effects. Steroids and cytotoxic drugs favor the occurrence of infections, as demonstrated by the results of several prospective trials,^2,6,10,15 especially in the elderly,³¹ that explored different therapeutic strategies, for which infectious side effects are less frequent when drugs are prescribed at a lower cumulative dose or are switched rapidly from intensive induction to less aggressive maintenance therapy.

Among the items included in the 1996 and 2009 versions of FFS, we did not modify the definitions for gastrointestinal and cardiac involvements. Kidney manifestations counted for 2 points in the 1996 FFS, with proteinuria >1 g/d being considered but not microscopic hematuria. Those parameters were evaluated separately in this analysis and were not retained as being associated with a poor prognosis.

According to our multivariable analysis, only peak creatininemia was a prognostic factor, and it overwhelmed the other renal parameters. Proteinuria and microscopic hematuria, signs of glomerulonephritis, had little impact on prognosis, probably because its recognition implies treating the patient with steroids and immunosuppressants, and the therapeutic decisions are made early, before renal insufficiency can occur.

In the study by Hogan and coworkers,¹⁴ other factors were associated with poor prognosis. Among them, alveolar hemorrhage increased the risk of death by an odds ratio of 8.64. Our multivariable analysis did not retain alveolar hemorrhage as a factor of poor prognosis, despite its having been selected in the univariable analysis, probably because it was already taken into account by another well-recognized poor-prognosis factor, renal insufficiency (pulmonary-renal syndrome), frequently associated with alveolar hemorrhage (97% of the patients followed by Lauque et al²³), and the profusion of potential mortality-impacting variables. Pertinently, in our cohort, severe alveolar hemorrhage was found only in a minority of patients who also suffered from severe renal insufficiency, and, for many patients, it was merely recorded as blood-streaked sputum that has no effect on outcome.

Few studies regrouping a large number of patients with alveolar hemorrhage have been published,^18,23 and 1 was published several decades ago.¹³ When the patient was hospitalized in an intensive care unit because of alveolar hemorrhage, those investigations focused on the subgroup comprising the most severely ill patients. Even for this subgroup of patients, at higher risk of mortality, the outcome was better than would be expected for some authors²⁰ and better management techniques improved prognosis.²¹

One of the major changes in the 2009 FFS is the absence of CNS involvement as a parameter of poor prognosis. We had arbitrarily introduced it into the 1996 FFS, considering its impact on prognosis, but its rarity prevented it from reaching statistical significance because of the small size effect. For the analysis of this larger cohort, we had decided not to include items that were not significantly associated with a poorer outcome in univariable analysis. That does not mean that CNS involvement is a benign manifestation. We confirmed that peripheral neuropathy has no effect on vital prognosis, even though its impact on function and quality of life is important.

Usually, assessment scores²⁶ grade disease severity by adding symptoms: the more symptoms, the more severe the disease is. In previous studies,²⁸ ENT manifestations were associated with a better WG outcome, even though these manifestations were present together with kidney involvement or other signs of particular severity. The results of the current new analysis clearly demonstrate that, more than the sum of symptoms, the intrinsic characteristics of each one plays a role in the patient's prognosis.

We did not examine the impact of ANCA on prognosis, because some patients were diagnosed before the era of ANCA, and, hence, their antibodies were not systematically sought at the time of SNV diagnosis. However, Hogan et al¹⁴ found that a cytoplasmic versus perinuclear ANCA-labeling pattern increased the relative risk of mortality. In our study, mortality was higher for MPA patients than for those with WG, which is contradictory to the findings of Hogan et al,¹⁴ if we consider cytoplasmic and perinuclear labeling to correspond, respectively, to WG and MPA.

The 2009 FFS, one of the simplest scores for clinicians to use to predict the prognosis of their SNV patients, now needs to be validated as a tool to help clinicians make therapeutic choices. When the 1996 FFS was applied to therapeutic options, we showed that SNV patients with FFS = 0 could receive steroids alone as first-line treatment,³⁵ and that those with PAN or CSS given first-line steroids and cyclophosphamide were overtreated.⁸ For WG, the combined immunosuppressant and steroid regimen is universally accepted, thus probably conferring less usefulness to the 2009 FFS for therapeutic choices. Pertinently, the new therapeutic strategies for the management of the 3 ANCA-associated SNV now being investigated could modify these dogmas in the near future. For SNV without poor-prognosis factors, the respective advantages and disadvantages of treating patients with steroids alone need to be reinvestigated in terms of morbidity and relapse prevention, even though our present results showed lower mortality when poor-prognosis factors were absent.

Although we have not yet tested the 2009 FFS on other vasculitides, it could probably be adapted to cryoglobulinemic vasculitis or Henoch-Schönlein purpura, but adaptation is unlikely for Takayasu arteritis, which has characteristics that differ markedly from those of ANCA-associated vasculitides.