Journal of Thrombosis and Haemostasis, 6: 1113–1121
DOI: 10.1111/j.1538-7836.2008.03001.x
ORIGINAL ARTICLE
Validation of a composite score for clinical severity of hemophilia ¨ M , * G . S T A˚ H L B E R G , * J . O D E B E R G à and M . B L O M B A¨ C K § S. SCHULMAN,* A. EELDE,* M. HOLMSTRO *Coagulation Unit, Department of Hematology, Karolinska Institute, Karolinska University Hospital, Stockholm, Sweden; Department of Medicine, McMaster University, Hamilton, ON, Canada; àAtherosclerosis Research Unit, Department of Medicine, Karolinska Institute, Karolinska University Hospital, Stockholm; and §Department of Molecular Medicine and Surgery, Karolinska Institute, Karolinska University Hospital, Stockholm, Sweden
To cite this article: Schulman S, Eelde A, Holmstro¨m M, Sta˚hlberg G, Odeberg J, Blomba¨ck M. Validation of a composite score for clinical severity of hemophilia. J Thromb Haemost 2008; 6: 1113–21.
Introduction Summary. Introduction: Evaluation of modulators of the phenotypic expression of hemophilia may benefit from a scoring system that reflects several aspects of the clinical severity instead of only one dimension. Methods: We describe here how we constructed a composite Hemophilia Severity Score (HSS) and performed validation. The items in the HSS are annual incidence of joint bleeds, World Federation of Hemophilia Orthopedic joint score, and annual factor consumption. The latter two were adjusted for age at start of prophylaxis and body weight. Data for 100 adolescent or adult patients with hemophilia A or B in the mild, moderate or severe form without inhibitors were collected for the 1990–1999 period. We evaluated the reliability (multidimension property, test–retest) and validity (content, convergent, discriminant and known groups) of the score. Results: The HSS ranged from 0 to 0.94 and was higher in severe hemophilia A than severe hemophilia B (median 0.50 and 0.24; P = 0.031). The validation indicated that the HSS is reliable and reflective of the clinical severity of hemophilia. The presence of factor V G1691A or prothrombin G20210A polymorphisms was found in 13 patients. The clinical severity, measured as the HSS or each of the three components, appeared to be modified by prothrombin G20210A but not by FV G1691A. Conclusion: The HSS is a well-defined tool that provides a comprehensive representation of the clinical severity of hemophilia in adults. It would be useful in larger studies on the assessment of modulators of the phenotypic expression of hemophilia. Keywords: arthropathy, bleeding, hemophilia, phenotype, thrombophilia, validation.
Correspondence: Sam Schulman, Thrombosis Service, McMaster Clinic, HHS – General Hospital, Hamilton, ON L8L 2X2, Canada. Tel.: +1 905 5270271 ext 44479; fax: +1 905 5211551. E-mail:
[email protected] Received 21 February 2008, accepted 24 April 2008 2008 International Society on Thrombosis and Haemostasis
The phenotypic expression of hemophilia is poorly correlated with the plasma level of the deficient factor [1]. Furthermore, whereas the plasma trough levels observed in a study on prophylaxis with factor concentrate showed a statistically significant correlation with the number of joint bleeds per year, the correlation coefficient was weak (r2 = 0.085) and there was no correlation with the incidence of non-joint bleeds [2]. It has therefore been of interest to assess possible modulators of the phenotypic expression of hemophilia. The half-life of factor VIII [3], the degree of thrombin generation [4], the level of von Willebrand factor [5], components of the fibrinolytic system such as tissue plasminogen activator and thrombinactivatable fibrinolytic inhibitor [5], and the presence of thrombophilic single nucleotide polymorphisms such as FV G1691A [6–9], prothrombin G20210A [10,11], tissue factor 5¢untranslated region [4] and methylene tetrahydrofolate reductase C677T [8], have all been evaluated. In some reports, a protective modulating effect of different hemostatic components on the severity of hemophilia was found [8–11], but not in others [4,7]. This may have been due to insufficient power to detect an effect. The characterization of the phenotypic expression has, however, varied between studies. Some have compared hemostatic components against the number of joint bleeds per year [4,7,9,10], the age at onset of symptomatic bleeding [11], the orthopedic joint score or number of affected joints [5,8–10], or the annual utilization of factor concentrate [5,8–10]. In one study, all of these, albeit separately, were used for comparison with the half-life of FVIII [3]. As the joint score increases with the number of bleeds suffered [12], and the incidence of hemarthrosis decreases with the amount of factor concentrate used [13], we propose a composite score to describe the phenotypic severity of hemophilia in adults. The objective was to construct a score that provides a more comprehensive measurement of the clinical severity of the disease than when a single variable is used. In this article, we describe how we have constructed this score and performed the validation. We also explored the influence of the
1114 S. Schulman et al
FV G1691A and prothrombin G20210A polymorphisms on the score proposed by us. Materials and methods Patient population
Patients with hemophilia A or hemophilia B with basal factor levels from 0 to 0.30 IU mL)1 were included in the population used for validation. Additional inclusion criteria were that they had been regularly examined at our center during the period January 1990 to December 1999, and that reliable data on factor concentrate utilization for this period were available. The exclusion criteria were presence of an inhibitor at any time before or during this period, and age < 18 years at the end of this period. Informed consent was obtained from the patients to use their data, and the Ethics Committee of the hospital approved the study. Data collection
From each patient, we obtained age, weight, height, original nationality (Swedish vs. non-Swedish), place of domicile (Stockholm or non-Stockholm), number of joint bleeds every year 1990–1999, the orthopedic joint score at the end of the period, type of hemophilia, basal level of the deficient factor, and the consumption of factor concentrate for each of the years between 1990 and 1999 as reported by the patients and, in cases of missing reports from the patients, as delivered from the hospital pharmacy. DNA was analyzed for the FV G1691A and prothrombin G20210A polymorphism using a pyrosequencing technique, as previously described [14]. The orthopedic joint score was the combined pain score and physical examination score, evaluated as recommended by the Orthopedic Advisory Committee of the World Federation of Hemophilia [15,16], in which a score of 0–15 is awarded to each of the knee and ankle joints and a score of 0–13 to each elbow, according to the severity of the arthropathy, with a maximum score of 86 [17]. Flexion contracture was assessed in hips, knees and ankles, as recommended [15,16]. If a patient had had arthroplastic surgery in one of these joints during the time period 1990–1999, the score before surgery was used to avoid a false mild reflection of the arthropathy. Construction of the composite score
The composite score, designed to reflect the clinical severity of hemophilia, was named the Hemophilia Severity Score (HSS). It consists of the sum of three components – annual incidence of joint bleeds, orthopedic joint score, and annual factor utilization; these should each have a maximum possible value of 1 to a worst composite value of 3. This would theoretically only be reached by a patient with a high titer inhibitor, suffering bimonthly hemarthrosis, and with completely destroyed joints in spite of massive use of factor concentrate.
The bleeding score component The number of joint bleeds per year was obtained from our medical records, copies from other hospitals, and the home-treatment reports. The annual incidence of joint bleeds was averaged for the 10 years of observation and then divided by 20, yielding the bleeding score. Twenty is typically considered to be the maximum number of joint bleeds that occurs in a patient with a truly severe form of hemophilia without any treatment, and, for example, Fischer et al. demonstrated in two publications interquartile ranges of annual bleeding frequency of 4.0–16.4 [18] and 3.8–24.0 [19]. The joint score component The orthopedic joint score was assessed independently by one of the experienced physicians at our treatment center and by the physiotherapist who regularly examines these patients. The physiotherapist did not see some of the patients with mild hemophilia and completely normal joints. The last score obtained during the 10-year period was divided by the maximum possible score of 86, resulting in the joint score. The factor score component The annual use of FVIII or FIX in kilo-units (kIU) was obtained from home treatment reports, treatment notes from our center, and from other hospitals. A comparison was also made in some unclear cases with data from the three pharmacies that are central distributors of factor concentrates to all local pharmacies and to the hospitals in Sweden. The annual amount was averaged for the 10 years of observation and normalized by dividing by the mean body weight of the patient during this period. We divided this value by 6 (kIU kg)1), which is approximately the maximum consumption of regular prophylaxis. This was derived from a recent publication from Malmo¨, Sweden [2], which showed that patients with hemophilia A who took FVIII 3 times week)1, every other day or 4 times week)1 had an annual use of 4.3, 5.3 and 6.1 kIU kg)1, respectively. Likewise, patients with hemophilia B who took FIX twice a week, every third day, 3 times week)1 or every other day had an annual use of 2.8, 3.2, 5.8 or 6.3 kIU kg)1. A few patients thus used > 6 kIU kg)1, but this was mainly driven by the presence of a weak inhibitor [2]. Adjustment of the components Briefly, there was a need for adjustment of the components to avoid effects on the HSS of parameters that have no obvious association with the clinical expression of hemophilia. Body weight was simply adjusted for as already described in the previous paragraph. Other characteristics that may influence the clinical severity were evaluated, but we did not find any influence of hepatitis C or human immunodeficiency virus (HIV) infective state, proximity of the domicile to a hemophilia treatment center or geographical place of origin, whereas the age when prophylactic factor replacement was started had a clear effect. The age at the start of prophylaxis is important for the development of arthropathy [20]. Many of our patients, mainly older patients or immigrants, had started prophylaxis beyond 2008 International Society on Thrombosis and Haemostasis
Composite score for hemophilic severity 1115
early childhood, and we found it necessary to adjust the joint score for this effect. We introduced this correction by dividing the joint score by the logarithm of (10+ age in years at start of prophylaxis). For patients who had never started on prophylaxis, this age was set at 50 years, because the arthropathy in patients without prophylaxis mainly develops before that age. The joint score was thus obtained from: worst orthopedic joint score during the observation period · 86)1 · log(age at start of prophylaxis + 10))1. We added the constant 10 to avoid division by a number < 1. The factor score also required adjustment for late start of prophylaxis. Arthropathy can cause more bleeding and higher utilization of factor concentrate. The averaged and normalized factor consumption was divided by the natural logarithm (ln) for the age at start of prophylaxis. We added the constant 1.72 to avoid division by a number < 1. The factor score was thus obtained from: average annual use · body weight)1 · 6 )1 · ln(age at start of prophylaxis + 1.72))1. Validation of the score
We used the following criteria to evaluate the HSS: internal consistency, content validity, convergent validity, discriminant validity, and known-groups difference. Internal consistency We have three components in the composite score, and consistency is not an objective. Indeed, it is important in this situation to reflect different rather than similar or homogeneous characteristics of the disease. Rsquared statistics to measure the amount of variation explained by one score on the other as well as CronbachÕs alpha were performed. The latter is usually employed to demonstrate that different items measure the same thing, which is defined as CronbachÕs alpha of > 0.70 for a summary score [21]. In our case, an alpha of < 0.70 was desirable. Test–retest reliability was assessed by independent evaluation of the orthopedic joint score by the physician and the physiotherapist. Content validity This criterion is used to describe how representative the contents of a scale or a score are of the conceptual domain that it is designed to illustrate, in this case the Ôclinical severity of hemophiliaÕ. We tested this by asking 11 patients with hemophilia and 11 physician experts on hemophilia what items best described the clinical severity of hemophilia. The five most frequent responses from each group were then provided to another set of 10 patients with hemophilia, and 10 other experts, who were asked to rank the most pertinent items by importance from 1 to 5, with 5 being the most important. Convergent validity This examines whether the scale or the score is correlated with other measures of the same or a similar construct. For this purpose, we examined the correlation between the HSS and the factor level for the entire population. 2008 International Society on Thrombosis and Haemostasis
Discriminant validity This criterion is used to show that the scale or score is independent of other measures that would confound the picture. We assessed the HSS against nationality, against place of domicile, and against active hepatitis C infection. It is possible that immigrants have received different or no treatment and that patients living far from the hemophilia treatment center receive less optimal treatment than those living nearby, but this must not influence the score. We also assessed whether the HSS was dependent on age in any subset. Conversely, the score should discriminate between groups with an established difference in severity, obviously corresponding to the classic types of hemophilia, which is termed known-groups difference. Statistical analysis
Categorical data from three groups was compared with PearsonÕs chi-squared test with YateÕs correction, and in cases of statistical significance or for two-group comparisons, we used a two-tailed FisherÕs exact test. For non-parametric data, we initially used the Kruskal–Wallis test for difference between three groups, and in cases of statistical significance, we proceeded with a two-tailed Mann–Whitney U-test or, in cases of normal distribution, StudentÕs t-test to detect from where the significance arose. A P-value of < 0.05 was considered to be statistically significant. Interobserver agreement for assessment of scores was also evaluated with the Bland & Altman plot. Results One hundred patients with hemophilia, all males, were recruited for validation of the HSS, and the distribution between type and severity of hemophilia is shown in Table 1. Whereas all eligible patients with moderate or severe forms of hemophilia fulfilling the recruitment criteria were included, we could have increased the number of patients with mild hemophilia. We chose not to do that, as the main problem with mismatch between the bleeding phenotype and factor levels is among the more severe forms. The characteristics of the patients are shown in Table 1. Patients with the severe form were significantly younger than those with the moderate form (Kruskal–Wallis P = 0.0499; Mann–Whitney for severe vs. moderate P = 0.017). There was no statistically significant difference in age between patients with hemophilia A or hemophilia B (data not shown). The proportion of immigrant patients with the three severities of hemophilia was not statistically different (Pearson chi-square P = 0.15). A higher proportion of the patients with severe hemophilia had been infected with hepatitis C virus (HCV) than of those with the mild or moderate form (Pearson chi-square P = 0.003; FisherÕs exact test P < 0.001 and P = 0.025, respectively). The patients, who were HCV-antibody positive but polymerase chain reaction negative for HCV RNA, were either spontaneously negative or had received successful therapy against the infection. There were also numerically more patients with the severe form than with the mild or moderate form of hemophilia
1116 S. Schulman et al Table 1 Characteristics of the patients according to severity of disease
n Hemophilia A/B Age at mid-term, mean (± SD) Immigrants (%) Domicile near HTC (%) HCV-negative/Ab+, PCR)/PCR+ HIV-positive (%) Median age at first bleed, years On prophylaxis (%) Median starting age, years Joint bleeds per year Orthopedic joint score (range) Factor consumption per year, kIU
Mild
Moderate
Severe
28 19/9 36.8 (16.7) 2 (7) 11 (39) 10/10/8 0 (0) 6.5 (3.8–18.2) 0 NA 0 (0–1) 0 (0–0.2) 0.9 (0–2.1)
29 21/8 42.8 (20.8) 6 (21) 13 (45) 7/8/14 1 (3) 4 (1.6–7) 5 (17) 16 (16–17) 1 (0–3) 3 (0–9) 6.7 (1.5–26)
43 37/6 31.0 (14.0) 11 (26) 21 (49) 2/20/21 13 (30) 1 (0.5–2) 34 (79) 8 (4.2–29.8) 3 (2–5) 10 (3–20) 152 (78–232)
P-value*
< < < <
