The Changing Prognosis of Classic Hemophilia (Factor VIII "Deficiency") Paul K. Jones, PhD, and Oscar D. Ratnoff, MD

Objective: To estimate relative risk of mortality and median life expectancy for patients with classic hemophilia by a life-table analysis, taking into account deaths that may have occurred in infancy or childhood before the onset of symptoms. Design: Retrospective chart review of clinical series. Setting: Referral-based university medical center. Patients: Seven hundred one patients with classic hemophilia (hemophilia A; factor VIII "deficiency") were studied for the years from 1900 to 1990; patients were identified in 289 families. Measurements and Main Results: Relative risk for mortality and median life expectancy among hemophiliacs were compared with those among normal U.S. males. Overall, mortality (relative to that of contemporaneous U.S. males) was increased about sixfold among severely affected patients, more than twofold among moderately affected patients, and was equivalent to that of U.S. males among mildly affected patients. Median life expectancy at 1 year of age had reached almost 68 years in the decade 1971 to 1980, but declined to only 49 years in the decade 1981 to 1990. Conclusions: After improvement in survival from 1971-1980 (corresponding to widespread treatment with lyophilized concentrates of antihemophilic factor [factor VIII]), relative mortality is now increasing, especially among severely affected patients, in large measure because of the acquired immunodeficiency syndrome (AIDS).

Annals of Internal Medicine. 1991;114:641-648. From Case Western Reserve School of Medicine and University Hospitals of Cleveland, Cleveland, Ohio. For current author addresses, see end of text.

1 he longevity of persons with classic hemophilia (hemophilia A; factor VIII ''deficiency") has increased in industrialized countries during this century. Two indices of longevity are the average age of patients receiving treatment and their average age at death. These averages are now said to resemble, especially in mildly affected hemophiliacs, the average age of living males and average age at death in the normal male population (1-9). Longevity has increased among all males in national populations, raising the question of how much of the increased longevity of hemophiliacs is due to secular trends in the general, nonhemophilic population. Further, the age at diagnosis and the length of time that patients are followed affect both the average observed age of living patients and age at death (10-13). Survival analysis and life-table methods have been applied to

study changes in hemophilic mortality in a more rigorous fashion (9, 14-17). Larsson (16) applied standard survival and life-table analysis methods to all known patients with either classic hemophilia or Christmas disease (hemophilia B; factor IX "deficiency") in Sweden, which has had a national registry for hemophiliacs since the 1750s. He estimated that, during the period from 1961 to 1980, patients with mild or moderate disease had a median life expectancy of about 72 years, compared with 76 years among all Swedish males. In contrast, patients with severe disease had a median life expectancy of 57 years. Median life expectancy has increased for Swedish hemophiliacs since the period 1831-1920, when it was estimated to be 28 years for moderately affected patients and 11 years for severely affected patients (16). Notably, Larsson treated classic hemophilia and Christmas disease as if they were a single entity. Larsson concluded: " . . . the infant mortality in mild and moderate haemophilia was grossly understated in relation to the total male population, but of course the haemophiliacs as a sub-population in Sweden faced at least the same infant mortality as the rest of the population. Several infant deaths were simply not recognized as having been caused by haemophilia" (p. 600, emphasis added). Similar anomalies have been described by others. Aronson (8) noted the low number of reported deaths among U.S. patients aged less than 1 year with classic hemophilia, which is surprising because cerebral hemorrhage at birth is common (8); he attributed the low number of deaths to delayed diagnosis. Rizza and Spooner (9) calculated median life expectancy for patients in the United Kingdom with classic hemophilia or Christmas disease; among severely affected patients, life expectancy was nearly equal to that for normal males, which they attributed to incomplete reporting of deaths. Twenty years ago, Ramgren (18) described the time of onset and type of clinical symptoms among Swedish hemophiliacs. He reported that the median age at which patients with classic hemophilia experienced the onset of clinical symptoms was 5.0 years, 3.0 years, and 1.5 years for mildly, moderately, and severely affected patients, respectively. Ramgren's data suggest that hemophilia was not recognized in most infants and young children until well after the peak of infant mortality in the general population had passed. Application of survival and life-table analyses to hemophilia has been relatively recent, and only one type of data censoring has been studied: right-censoring, in which survival times of patients alive at the most recent time of observation are used only up to the last date of contact. The possibility of left-censoring (12, 13) poses ©1991 American College of Physicians

Downloaded From: https://annals.org/ by a Glasgow Univ Library User on 09/14/2018

641

Table 1. Patient Characteristics

by Severity of Disease

Patient Characteristic

Severe (n = 413)

Moderate (n = 105)

Mild (n = 183)

%—

< Residence Ohio Adjoining state Elsewhere in United States Outside United States Unknown Year of birth 1844-1899 1900-1920 1921-1940 1941-1960 1961-1970 1971-1980 1981-1990 Race White Black Test status Patient tested Related patient tested Neither tested Circulating anticoagulants* Tested positive Tested negative Cause of deatht Bleeding or trauma mentioned AIDS mentioned Liver disease Other cause listed Cause unknown Total

Total (/i = 701)

Severity of Disease

>

54 8 22 5 11

46 28 14 0 12

48 9 28 4 11

49 11 25 4 11

10 15 23 23 13 11 5

5 16 21 31 10 13 4

4 9 15 39 16 13 4

6 12 18 34 14 12 4

97 3

98 2

92 8

94 6

50 39 11

34 47 19

33 17 50

38 27 35

0 100

3 97

21 79

13 87

28 7 0 26 39 100

48 3 0 23 26 100

45 13 3 12 27 100

42 11 2 16 29 100

* Number tested were 257: 65 with mild disease, 37 with moderate disease, and 155 with severe disease. t Number of deaths was 239: 43 with mild disease, 27 with moderate disease, and 169 with severe disease.

a more difficult problem as some hemophiliacs may have died before they were identified. (The designations of "left" and "right" derive from the convention of a time line running from left [beginning at birth] to right [terminating at death].) We have done a life-table analysis of patients with classic hemophilia; the data indicate an interruption in the trend toward improved survival among patients with classic hemophilia because of the acquired immunodeficiency syndrome (AIDS) epidemic (19-22). Reported human immunodeficiency virus (HIV) seropositive rates among U.S. patients with classic hemophilia average about 70%; most of these serologic surveys have been done at hemophilia treatment centers, where patients with severe disease are more likely to be encountered (19). The cumulative incidence of AIDS among severely affected patients with classic hemophilia during the period January 1981 to September 1987 was estimated by the Centers for Disease Control as 4.2 cases per 100 persons (21). More recently, the incidence rate after seroconversion was 2.67 per 100 person-years and was directly related to age (22). Methods To assess the changing prognosis of classic hemophilia, the records of 289 affected families were examined retrospectively 642

15 April 1991 • Annals of Internal Medicine

to determine the year of birth and the last known contact with each patient, and the year of death in patients known to have died. In 181 families (63%), at least 1 patient with classic hemophilia had been studied at University Hospitals of Cleveland. In the other families, 1 or more female relatives of a patient had been tested for the carrier state at University Hospitals of Cleveland, and a positive diagnosis in her affected relative had been attested by the referring hospital, physician, or hemophilia center. Whenever possible, the family trees in each family were examined to determine the dates of birth and death of affected relatives in earlier generations, assuming that in each family all males with a life-long history of bleeding had classic hemophilia. A total of 701 patients with classic hemophilia were identified in the 289 families. Severity of disease was estimated in each family from the coagulant titer of antihemophilic factor (AHF, factor VIII) determined in at least one affected individual. The patients were classified into one of three groups: mildly affected (factor VIII coagulant activity level, 0.05 U/mL or higher; n = 183), moderately affected (0.01 to 0.04 U/mL, n = 105), or severely affected (< 0.01 U/mL, n = 413), assuming that the mean procoagulant titer in a normal male from 21 to 39 years of age is 1.00 U/mL. When the level of antihemophilic factor procoagulant activity was unknown for a particular patient, it was imputed from values of other affected family members. Information concerning severity of illness was consistent among hemophiliacs diagnosed within the same family; in only 4 families was there an individual patient whose severity level was different from other patients in the family who were tested. Information regarding year of birth, year of death if the patient was dead, or year last known to be alive was elicited from patients, other family members, attending physi-

• Volume 114 • Number 8

Downloaded From: https://annals.org/ by a Glasgow Univ Library User on 09/14/2018

cians, or agencies providing services for hemophiliacs. This information was collected and updated from 1951 to September 1990. Life-table analysis was done using age-specific death rates for all U.S. males for the years from 1900 to 1986. Before 1933, death rates were available only for those states referred to as the "death registration" states (23, 24). Age-specific death rates for U.S. males for 1986 (the latest available at this writing) were used for 1987 to 1990 (25). The age intervals used were less than 1 year, 1 to 4 years, 5 to 14 years, and thereafter in 10-year intervals (23, 25). The basic unit of observation for our life-table analysis was the person-year (each year in which the patient was known to have been alive). For each person-year, the corresponding U.S. male age-specific death rate was ascertained as an "expected" death. An "observed death" was said to occur in a given person-year if the patient had died in that year. The numbers of observed and expected deaths were then summed over all the person-years within subgroups of interest. A given patient could contribute person-years to more than one subgroup. We anticipated that the relative mortality risk ratio (10, 17) of observed-to-expected deaths would exceed unity for all hemophiliacs, except perhaps in the youngest age groups. It was further anticipated that this ratio would be higher in moderately affected patients than among mildly affected patients and highest among severely affected patients. Two estimates of the relative mortality ratio, crude and adjusted, were computed {see Appendix). The crude estimate was obtained by dividing the summed observed deaths (O) in each subgroup by the summed expected deaths (E), where E is based on the death rates for all U.S. males. The adjusted estimate of relative mortality was obtained by calculating the number of deaths before the onset of symptoms in each ageseverity-specific subgroup, assuming that in any cohort of live births an unknown number of individuals born with hemophilia died before diagnosis. During the early years of life, the disease is recognized in only a fraction of hemophiliacs; this fraction depends on the severity of disease and the age of the hemophiliac, and was estimated from published data of Ramgren (18). The formula for the adjusted relative risk therefore combines two mortality components: observed deaths occurring in diagnosed patients and presumed deaths among undiagnosed hemophiliacs. In order to calculate the presumed deaths, we assumed that the relative risk of mortality was a linear function of age (26, 27). Survival curves, as well as median life expectancy at selected ages, were computed for normal males and for hemophiliacs. Mortality among patients with classic hemophilia was tabulated for each historical period (1900 to 1920, 1921 to 1940, 1941 to 1960, 1961 to 1970, 1971 to 1980, and 1981 to 1990),

Table 2. Crude and Adjusted Disease from 1900 to 1990

Estimates

Age, y Mild (/i = 183) < 1 1-4 5-14 15-24 25-34 35-44 45-54 55-64 65-74 75-84 85+ Total Correlation of age and mortality

0.00 0.00 0.55 0.62 1.95 1.67 0.49 1.32 0.58 1.04 0.58* 0.76 NA

and survival was calculated at or near the midpoint of each period (1910, 1930, 1950, 1965, 1975, and 1986, respectively); 1986 was used as the comparison year for the last historical period because it was the most recent year for which U.S. mortality data were available (25). In each historical period we summarized the mortality experience for patients alive at any time during the period; it was assumed that, in each age interval, the relative mortality was at least unity, the expectation for U.S. males. The chi-square test and the chi-square test for trend were used to compare numbers of observed versus expected deaths between adjacent historical periods and over several periods, respectively (10, 13, 28). We were especially interested in whether survival improved after widespread treatment with lyophilized concentrates of antihemophilic factor (factor VIII), introduced about 1967 (29), and whether survival deteriorated after the recent increase in reported transfusionrelated cases of AIDS, beginning in 1984 (19-22).

Results Sixty percent of the patients resided in Ohio or the adjoining states of Pennsylvania, West Virginia, Kentucky, Indiana, and Michigan (Table 1). Thirty-six percent of the patients were born before 1941. Six percent of the patients were black. In the total sample, 38% of patients had been tested at University Hospitals of Cleveland. In an additional 27% of patients, a hemophilic relative of the patient (another of the 701 patients) had been tested at University Hospitals of Cleveland. Circulating anticoagulants were found in 21% of those severely affected patients who were tested. Of 701 patients, 239 had died by September 1990; 42% of deaths were attributed to trauma or bleeding (including postoperative deaths), a percentage not varying significantly among severity groups (P > 0.05). One patient in our study, who was severely affected, died before 1900; he was excluded from subsequent calculations. Table 2 presents crude and adjusted relative mortality by age of patient and by severity of disease, excluding the single patient who died before 1900. No deaths were observed among mildly affected patients before the age of 5 years, and none was observed among moderately affected patients before the age of 1 year; hence the

of Relative Mortality Risk by Age of Patient in Years and Severity of Crude Relative Risk Moderate Severe (n = 105) (n = 412) 0.00 0.64 2.99 2.27 2.33 2.18 0.39 2.15 1.69 0.00* NA* 1.23 NA

0.63 6.19 6.78 6.67 8.29 4.83 3.49 2.42 1.96 0.00* NA* 3.42 NA

Total (n = 700)

Mild (n = 183)

0.35 3.11 4.08 3.98 4.85 2.98 1.41 1.76 1.00 0.93 0.58* 1.86 NA

0.93 0.68 0.89 0.78 1.95 1.67 0.49 1.32 0.58 1.04 0.58* 0.97 0.21

Adjusted Relative Risk Total Moderate Severe (n = 105) (n = 412) (n = 700) 2.96 2.42 3.50 2.34 2.33 2.18 0.39 2.15 1.69 0.00* NA* 2.17 -0.71t

7.09 7.81 6.81 6.67 8.29 4.83 3.49 2.42 1.96 0.00* NA* 6.05 -0.64t

4.66 4.45 4.30 4.04 4.85 2.98 1.41 1.76 1.00 0.93 0.58* 3.13 -0.78t

* Estimate was either unavailable for patient subgroup (indicated by NA, if not available) or was based on fewer than 10 person-years. Relative mortality computed for age-severity-specific subgroups using deaths occurring any time during the years from 1900 to 1990. t P < 0.001 for decreasing linear trend of relative risk for mortality with age. 15 April 1991 • Annals of Internal Medicine

Downloaded From: https://annals.org/ by a Glasgow Univ Library User on 09/14/2018

• Volume 114 • Number 8

643

Table 3. Adjusted Relative Mortality Risk and Corresponding Median Life Expectancy Years of Age by Disease Severity and Historical Period Disease Severity

Historical

Period Mild (n = 183)

Moderate (n = 105)

Severe (n = 412)

Total (n = 700)

U.S. males$

1900-1920 1921-1940 1941-1960 1961-1970 1971-1980 1981-1990 All years 1900-1920 1921-1940 1941-1960 1961-1970 1971-1980 1981-1990 All years 1900-1920 1921-1940 1941-1960 1961-1970 1971-1980 1981-1990 All years 1900-1920 1921-1940 1941-1960 1961-1970 1971-1980 1981-1990 All years 1910 1930 1950 1965 1975 1986

PersonYears* 606 1288 1817 1163 1191 856 6921 280 576 1015 654 706 538 3769 587 1228 2594 2078 2394 1746 10 627 1473 3092 5426 3895 4291 3140 21 317

Relative Mortality Riskt At 30 Years At 1 Year 1.75 0.47 0.95 0.34 0.13 0.97 0.95 2.61 2.57 5.34 0.63 9.28 2.72 3.23 2.53 9.56 12.70 10.48 1.71 13.86 7.81 2.55 4.64 8.55 6.94 2.85 9.38 4.91 1.00 1.00 1.00 1.00 1.00 1.00

1.20 0.80 1.68 0.54 0.38 1.64 1.11 1.97 1.87 3.31 1.54 3.66 2.58 2.26 4.18 4.30 8.61 5.73 3.11 11.26 6.12 2.04 2.49 5.07 3.60 2.15 7.33 3.54 1.00 1.00 1.00 1.00 1.00 1.00

at 1 Year of Age and at 30 Median Life Expectancy^ At 1 Year At 30 Years 57.9 57.2 61.7 69.2 69.5 60.0

34.5 30.0 34.2 40.8 41.3 32.6

37.3 58.8 58.7 57.3 67.1 64.1

11.2 38.3 32.7 29.1 40.6 37.4

29.9 27.4 39.7 53.8 60.5 39.8

14.6 32.0 21.0 27.6 36.1 24.4

40.9 52.6 54.9 61.3 67.5 49.1

30.6 31.7 29.4 35.8 40.3 30.9

62.0 65.5 68.9 69.3 71.1 74.0

36.9 38.3 40.7 41.0 42.9 45.4

* A given patient could contribute person-years to more than one historical period. t Relative mortality risk estimated by linear regression within each historical period. $ Estimates of median life expectancy may differ from published data because 10-year age intervals were used to estimate mortality.

crude relative mortality was zero in these patient subgroups. Among the 700 hemophiliacs alive any time during the years from 1900 to 1990, 238 deaths occurred, compared with 127.7 expected deaths from the U.S. male mortality rates for the corresponding years. Thus the crude relative mortality was 238/127.7, or 1.86. This crude relative mortality estimate understates the risk associated with the disease; it is likely that some deaths in infancy or childhood were attributed to other causes or that information obtained from family members was incomplete. Therefore, adjusted estimates of the relative mortality were obtained. From Ramgren's (18) data and the U.S. male death rates, an adjustment of 161.7 deaths was calculated to correct for deaths that had occurred before the disease could be recognized among hemophilic infants and children. The adjusted estimate of relative mortality was (238 + 161.7)/127.7, or 3.13. The 238 observed deaths represent about 60% of the total deaths. Although the effect of adjustment was substantial in infancy, it decreased with age; for patients aged 25 years or more, crude and adjusted relative mortality were identical. The average adjusted relative mortality 644

for all mildly affected patients was 0.97, not significantly different from unity (the defined level of risk in normal males). Moreover, there was no apparent association between age and the adjusted relative mortality estimates for mildly affected patients. For moderately affected patients, the average relative mortality across all ages was 2.17 and, for severely affected patients, the average relative mortality was 6.05. The need to perform some kind of an adjustment may be seen in several ways (Table 2). The relative mortality should be at least unity in each table cell, yet the pattern of underestimation corresponds to those ageseverity strata in which some of the hemophiliacs had not yet been recognized to be bleeders because symptoms had not yet appeared. In addition, comparison of crude with adjusted relative mortality across all ages indicates that failure to adjust reduces the estimate by 22% for mildly affected patients, by 43% for moderately affected patients, and by 43% for severely affected patients (40% reduction overall). Finally, for all patients, a declining linear trend of adjusted relative mortality was observed with age (weighted correlation coefficient r = - 0.78, P < 0.001).

15 April 1991 • Annals of Internal Medicine • Volume 114 • Number 8

Downloaded From: https://annals.org/ by a Glasgow Univ Library User on 09/14/2018

Historical Trends Table 3 shows the adjusted relative mortality and median life expectancy at ages 1 and 30 years by disease severity and historical period. The 700 patients contributed 21 317 person-years of observation. Calculations were done using the U.S. male death rates in each year from 1900 to 1990 and the estimates of relative risk obtained in severity-age-specific subgroups within each historical period; it was assumed that the true relative risk could not be less than unity for any subgroup of hemophiliacs. Median life expectancy was calculated at age 1 year rather than at birth to avoid a disproportionate influence of infant mortality (24), which was high at the beginning of the century and was subject to incomplete reporting. From 1910 to 1986 the median life expectancy at age 1 year among U.S. males increased from 62 to 74 years. The median life expectancy at age 1 year for all hemophiliacs increased from 41 to almost 68 years, but has declined recently to only 49 years. Although mildly affected hemophiliacs had life expectancies only slightly less than contemporaneous U.S. males, moderately and severely affected patients had substantially shorter life expectancies. The historical changes in life expectancy can be further understood through inspection of regression estimates of relative mortality risk in each historical period (Table 3). As U.S. males have had increased life expectancies, so too have hemophiliacs. Among those mildly affected, adjusted relative mortality was essentially unity at both age 1 and age 30; hence changes in the life expectancy of mildly affected patients have corresponded to changes in normal U.S. males. Relative mortality risk estimates for moderately and severely affected hemophiliacs are generally higher at age 1 year than at 30 years of age. Figure 1 displays projected survival curves from age 1 year for normal males (left) and all hemophiliacs (right); these curves show how mortality patterns have changed over the study period. Survival of normal males increased steadily in the years 1910, 1950, 1975, and 1986 (curves for 1930 and 1965 were omitted to ensure readability). Survival among all hemophiliacs increased in the years 1910, 1950, and 1975, based on mortality in the corresponding historical periods; by 1975 the per-

cent survival was close to that among normal males. By 1986, however, survival had dropped sharply, especially among hemophiliacs older than 15 years. To clarify how mortality has changed over time, the relative risks of mortality were compared across historical periods for selected age ranges (Table 4). Among all hemophiliacs (particularly among severely affected patients) there was a declining linear trend in mortality from 1941 to 1980 (P < 0.01). Mortality among severely affected children aged 1 to 14 years decreased significantly in the period 1971 to 1980 from that of the previous decade (P < 0.05). Mortality among all patients rose significantly in the decade 1981 to 1990 (P < 0.001), due to deaths from AIDS, an increase especially pronounced among severely affected patients aged 15 to 34 years (P < 0.001). Of 52 deaths among study patients in the period 1981 to 1990, 26 patients had AIDS (22 severely affected, 1 moderately affected, and 3 mildly affected hemophiliacs). The median age at death for patients with AIDS was 33 years (range, 8 to 71 years). Mortality was also examined in relation to race and to the presence of circulating anticoagulants. Weighted averages of the race-specific relative risks were computed to control for disease severity and the under-representation of black patients among the mildly and moderately affected patients. Black hemophiliacs had a 22% higher death rate than whites (P > 0.05). In the general U.S. population, normal black males had age-adjusted mortality rates averaging 39% higher than those for normal white males during the same years. Other Risk Factors Circulating anticoagulants were generally not identified in mildly and moderately affected patients. There were 155 severely affected patients who had been tested for circulating anticoagulants; 40 deaths occurred over the period from 1955 to 1990. Among 33 severely affected patients who tested positive for circulating anticoagulants, 14 died (10 from bleeding or trauma, 1 from AIDS, 2 from other causes, and 1 from unknown cause); among the 122 who tested negative, 26 died (9 from bleeding or trauma, 13 from AIDS, 3 from liver disease, and 1 from other cause). Although overall mor-

Figure 1. Projected survival from one year of age among U.S. normal males (left) and among all hemophiliacs (right). Survival curves are presented only for selected years (1910, 1950, 1975, and 1986) to ensure readability. Curves for 1930 and 1965 were intermediate. 15 April 1991 • Annals of Internal Medicine • Volume 114 • Number 8

Downloaded From: https://annals.org/ by a Glasgow Univ Library User on 09/14/2018

645

Table 4. Adjusted Relative Mortality Risk and Number of Observed Deaths for Selected Age Ranges by Disease Severity and Historical Period* Disease Severity

Mild (n = 183)

Moderate (n = 105)

Severe (« = 412)

Total (n = 700)

Historical Period

< 1

1 to 14

Age Range, y 15 to 34

>35

Total

1900-1920 1921-1940 1941-1960 1961-1970 1971-1980 1981-1990 All years 1900-1920 1921-1940 1941-1960 1961-1970 1971-1980 1981-1990 All years 1900-1920 1921-1940 1941-1960 1961-1970 1971-1980 1981-1990 All years 1900-1920 1921-1940 1941-1960 1961-1970 1971-1980 1981-1990 All years

0.93 0.93 0.93 0.93 0.93 0.93 0.93 2.96 2.96 2.96 2.96 2.96 2.96 2.96 6.46 7.08/3 7.28/5 7.65/2 8.40/2 6.46 7.09/12 3.87 4.29/3 5.56/5 5.49/2 6.04/2 4.42 4.66/12

1.09/1 0.53 0.51 0.47 0.47 0.48 0.76/1 2.37/1 1.24 5.75/2 0.97 12.93/1 0.93 2.84/4 2.99/4 9.00/17 12.89/17 9.75/6 0.69t 7.45/1 7.41/45 2.14/6 4.50/17 8.82/19 6.54/6 2.18/1 4.33/1 4.39/50

1.80/2 1.36/2 2.46/4 0.08 0.09 0.06 1.35/8 0.03 4.12/3 2.72/2 2.10/1 0.03 6.71/2 2.33/8 4.93/6 5.92/8 8.95/14 6.36/8 4.15/9 15.08/21§ 7.37/66 2.70/8 3.59/13 5.06/20 3.88/9 2.57/9 10.52/23§ 4.41/82

0.33/1 0.73/4 1.66/11 0.60/3 0.78/6 1.10/9 0.95/34 10.17/1 0.00 2.64/6 2.00/3 0.74/2 1.12/3 1.49/15 6.16/1 0.72/1 4.24/14 3.50/7 1.98/6 3.71/16 3.17/45 0.92/3 0.65/5 2.54/31 1.53/13 1.04/14 1.85/28 1.56/94

0.88/4 0.84/6 1.62/15 0.58/3 0.72/6 1.04/9 0.97/43 2.53/2 2.22/3 3.02/10 2.07/4 1.06/3 1.69/5 2.17/27 5.35/11 6.42/29 7.33/50 6.12/23 3.59/17 6.51/38§ 6.05/168 2.89/17 3.15/38 4.43/75 2.98/30 1.78/26 3.00/52§ 3.13/238

* Significant change in adjusted relative mortality from previous historical period using chi-square test to compare observed with expected deaths. Observed deaths follow slashes (/); zeros are omitted for conciseness. t P < 0.05. XP < 0.01. §/> < 0.001.

tality was not significantly higher among patients who had circulating anticoagulants (23% higher, P > 0.2), there was a significant inverse association between presence of circulating anticoagulants and death from either AIDS or liver disease (odds ratio = 0.05, P < 0.001). A possible interpretation is that, although the presence of circulating anticoagulants may have raised risk for death from bleeding or trauma, such patients were less likely to be transfused (because they were assumed to be refractory to treatment) and therefore were less likely to contract AIDS or liver disease. Discussion We did this study to assess the change in prognosis of classic hemophilia because of the current epidemic of AIDS. The data presented demonstrate a striking decrease in life expectancy, particularly among severely affected patients. Early studies of the prognosis of classic hemophilia were hampered by lack of knowledge concerning the multiplicity of hereditary hemorrhagic disorders, by failure to identify mildly affected individuals, and by the lack of laboratory procedures to confirm the diagnosis and assess the degree of severity. Even now, evaluation of prognosis has been confused by lumping classic hemophilia and Christmas disease under the same rubric, hemophilia, a practice that makes the a priori assump646

tion that the disorders run identical courses (for example, Larsson [16], Rosendaal and colleagues [17]). We limited our analysis to classic hemophilia, and corrected for the fact that many patients are alive at the time of this writing by using standard life-table methods (10-13). We also corrected for the fact that some hemophiliacs died before they could be identified; failure to make this correction would have underestimated relative mortality risk and overestimated median life expectancy; in some subgroups median life expectancy could not have been estimated. We do not have adequate information on individual patients regarding the age of onset of symptoms of hemophilia. Several lines of evidence support the rationale of our approach: previous underestimation of the prevalence of hemophilia in national populations, specification of the reasons for delayed or missed diagnosis, and documentation of hemophilic morbidity and mortality in the first year of life. Excellent research concerning trends in mortality among hemophiliacs has been done in Scandinavia (2, 6, 7, 15, 16, 18, 29). Factors conducive to complete enumeration include establishment of national patient registries and performance of most testing of patients and possible carriers in a small number of hospitals, thereby facilitating accurate assessment of hemophilia prevalence. Yet previous identification of hemophiliacs has

15 April 1991 • Annals of Internal Medicine • Volume 114 • Number 8

Downloaded From: https://annals.org/ by a Glasgow Univ Library User on 09/14/2018

been incomplete. Larsson (16), for example, noted that disease prevalence in Sweden had been underestimated as recently as 1960 by 50%. Historically, a family history of hemophilia has often been concealed from other family members or from society, even after the onset of symptoms in an affected male. Bulloch and Fildes (30), in 1912, identified two potent motivations for concealment: "The disease runs in families, and in the course of time an attitude of fatalism is developed towards the resources of medicine, so that in all probability many cases do not come to the attention of the medical man at all. . . . Another point which conduces to concealment is the fact that women being conductors [carriers] of the disease may have their chances of matrimony considerably reduced if their anomaly were widely known" (30). A half-century later, Kerr (31) recognized the same motivations, noting early opposition of the medical profession and society to marriage of hemophiliacs or of their female relatives. Kerr emphasized that it may take a long time to modify attitudes toward the illness in hemophiliacs and family members; thus patient behavior may reflect older societal expectations rather than current medical management and understanding of the underlying disease mechanisms. Concealment of the diagnosis may also reflect fears concerning acceptance for employment as well as the stigma of having "bad blood." Baehner and Strauss (32), studying the occurrence of bleeding episodes in diagnosed hemophiliacs in the first year of life, stated: ". . . apart from circumcision, hemophilic newborn infants rarely have bleeding. This is all the more surprising when the significant trauma in a normal delivery is considered." However, they also noted: "In 76 percent of the infants who had moderate or severe bleeding difficulties in the newborn period, the diagnosis of hemophilia was not established at that time. Most of these had severe disease and a negative family history for bleeding, so that hemophilia was not suspected." Baehner and Strauss (32) also reported an increase in frequency of bleeding about the ninth month of life, presumably due to the increased opportunity for trauma among toddlers. In our series, relative mortality was elevated in infancy as well as in early childhood. Of the 12 recorded infant deaths (all in severely affected infants), 8 involved a history of bleeding or trauma, and three were of unknown cause; only 1 was known to have a lethal illness not associated with bleeding or trauma. We found the same underestimation of mortality in younger age groups reported previously (8-9, 16). Clinicians who care for children and adolescents have noted the special problems of regulation of behavior. As Agle (33, 34) emphasized, a minority of hemophilic youths have unusual risk-taking behavior that exposes them to increased risk for death from hemorrhage or trauma. Surviving hemophiliacs who mature into their late twenties and beyond may adapt their behavior so as to lower the risk for death from trauma. One speculation is that hemophilia may protect from ischemic heart disease, which, if true, would lower death rates in middle age and beyond (8, 17). Only a few instances of myocardial infarction have been recognized in patients with classic hemophilia (35). Results

from a study among Dutch patients with classic hemophilia or Christmas disease reported an 80% reduction in ischemic heart disease relative to normal males (17). Our data on relative mortality do not show a uniform pattern of improvement in recent decades. Relative mortality was slightly (but not significantly) lower during the years 1900 to 1940, perhaps reflecting a tendency to recall longer-lived hemophiliacs in the early years. From 1941 to 1980 the relative mortality fell, presumably attributable to improved methods of treatment. The increase in life expectancy among hemophiliacs parallels the reduction in mortality rates at all ages and the consequent increase in life expectancy in the U.S. male population. The recent and sudden deterioration in life span for patients with classic hemophilia reflects the growing number of deaths from AIDS. Given the high level of HIV seropositivity among patients who received transfusions between 1978 and 1985, we fear that mortality will increase (19-22). Precisely when the AIDS-related mortality will peak is not known given the uncertainty regarding the incubation period and the future survival times of persons with hemophilia-associated AIDS (19, 21-22). Appendix. Derivation of Relative Risk

Estimators

The usual (crude) estimator of relative mortality risk is of the form O/E where O is the number of observed deaths occurring among individuals in a cohort of size N at a specific age, and where E = dN is the expected death rate obtained by applying the age-specific death rate d in the population of normal males to the N individuals. The adjusted estimator of relative mortality risk is also a ratio; it is assumed that there are N = N, + N2 hemophiliacs of a specific age in a cohort, in which N, of the individuals have been diagnosed and N2 have not been diagnosed. We define the fraction diagnosed at a particular age to be: f = N,/(N, + N2) and assume that f can be estimated by the fraction of patients who had experienced onset of symptoms at that specific age in a set of patients who were eventually diagnosed. The adjusted estimate of relative mortality risk is the ratio of the sum of observed deaths O among the N, diagnosed hemophiliacs and assumed deaths occurring among the N2 undiagnosed hemophiliacs to the expected deaths among N normals of the same age. In order to the calculate the adjustment, we first estimate a provisional regression equation for the relative risk Rreg = a + b (age) over a range of ages for which onset of symptoms has already occurred. Among the N2 undiagnosed hemophiliacs, the calculated number of assumed deaths is Rreg d N2 (it was assumed that deaths were uniformly distributed over the person-year [10, 26]). Hence the adjusted relative mortality risk among the N individuals of a given age is Radj = [O + Rreg d N2]/[d (N, + (N2)] = [O + Rreg d N (1 - f)]/[dN] = [O + Rreg E (1 - f)]/E where E = dN as before. Because the regression intercept (a) and slope (b) could not be expressed in terms of a simple formula, a trial-and-error systematic search was made to estimate these parameters within each severity group; the search typically converged to the final estimates within a few steps. Starting values for the search were obtained from the portion of the life span for which relatively complete data were available. At each step new values for a and b were used to compute values for Rreg,

15 April 1991 • Annals of Internal Medicine • Volume 114 • Number 8

Downloaded From: https://annals.org/ by a Glasgow Univ Library User on 09/14/2018

647

which in turn led to refined values of Radj. If the adjusted relative risk is required for a group of hemophiliacs of varying ages, the estimate is constructed by summing the numerators and summing the denominators across specific ages and taking the ratio of these sums. Sensitivity analysis was conducted to examine whether the findings depended upon the values of the fraction f calculated from Ramgren's (18) data. A variation of plus or minus 10% in values of f was inversely associated with a variation of 2% in the relative risk and directly associated with a variation of 0.2% in the median life expectancy at 1 year among all hemophiliacs. Because variations in the value of f did not affect the tests of significance, the conclusions are not altered. A possible limitation of our approach is that the relative risk may not be a linear function of age, as we have assumed; a nonlinear function may be more appropriate for estimating true relative risk. If AIDS-related deaths are excluded, the relative risk appears to be somewhat more linear with age. Acknowledgments: The authors thank the patients, family members, physicians, and other health care personnel who contributed information. Grant Support: By grant HL01661 from the National Heart, Lung, and Blood Institute, the National Institutes of Health, U.S. Public Health Service; and by grant M01 RR00080-25 from the General Clinical Research Centers Branch, Division of Research Resources, National Institutes of Health, U.S. Public Health Service. Requests for Reprints: Oscar D. Ratnoff, MD, Department of Medicine, University Hospitals of Cleveland, Cleveland OH 44106. Current Author Addresses: Dr. Jones: Department of Epidemiology and Biostatistics, Case Western Reserve University School of Medicine, Cleveland OH 44106. Dr. Ratnoff: Department of Medicine, University Hospitals of Cleveland, Cleveland OH 44106. References 1. Nour-Eldin F. Longevity of hemophiliacs [Letter]. Br Med J. 1961; 1:824. 2. Ramgren O. Haemophilia in Sweden. V. Medico-social aspects. Acta Med Scand. 1962;Suppl 379:37-60. 3. Nour-Eldin F. Blood Coagulation Simplified. New York: AppletonCentury-Crofts; 1967:87-8. 4. Biggs R. Haemophilia treatment in the United Kingdom from 1969 to 1974. Br J Haematol. 1977;35:487-504. 5. Eyster ME, Lewis JH, Shapiro SS, et al. The Pennsylvania hemophilia program 1973-78. Am J Hematol. 1980;9:277-86. 6. Larsson SA, Nilsson IM, Blomback M. Current status of Swedish hemophiliacs: I. A demographic survey. Acta Med Scand. 1982;212: 195-200. 7. Larsson SA, Wiechel B. Deaths in Swedish hemophiliacs, 1957-1980. Acta Med Scand. 1983;214:199-206. 8. Aronson DL. Cause of death in hemophilia A patients in the United States from 1968 to 1979. Am J Hematol. 1988;27:7-12. 9. Rizza CR, Spooner RJ. Treatment of haemophilia and related disorders in Britain and Northern Ireland during 1976-80: report on behalf of the directors of haemophilia centres in the United Kingdom. Br Med J [Clin Res]. 1983;286:929-33.

648

10. Colton T. Statistics in Medicine. Boston: Little, Brown and Co.; 1974. 11. Gross AJ, Clark VA. Survival Distributions: Reliability Applications in the Biomedical Sciences. New York: Wiley; 1975. 12. Cox DR, Oakes D. Analysis of Survival Data. London: Chapman and Hall; 1984. 13. Kotz S, Johnson NL, Read CB, eds. Encyclopedia of Statistical Sciences. New York: Wiley; 1982. 14. Stafford RS, Hegewald M, Haag C, Wolff L, Lovrien E. Life expectancy in hemophilia [Abstract]. Clin Res. 1980;28:103A. 15. Ikkala E, Helske T, Myllyla G, Nevanlinna HR, Pitkanen P, Rasi V. Changes in the life expectancy of patients with severe haemophilia A in Finland in 1930-79. Br J Haematol. 1982;52:7-12. 16. Larsson SA. Life expectancy of Swedish haemophiliacs, 1831-1980. Br J Haematol. 1985;59:593-602. 17. Rosendaal FR, Varekamp I, Smit C, et al. Mortality and causes of death in Dutch haemophiliacs, 1973-86. Br J Haematol. 1989;71: 71-6. 18. Ramgren O. Haemophilia in Sweden: III. Symptomatology, with special reference to differences between haemophilia A and B. Acta Med Scand. 1962;171:237-42. 19. McGrady GA, Jason JM, Evatt BL. The course of the epidemic of acquired immunodeficiency syndrome in the United States hemophilia population. Am J Epidemiol. 1987;126:25-30. 20. Human immunodeficiency virus infection in the United States: a review of current knowledge. MMWR. 1987;36(Suppl. 6): 1-20. 21. Stehr-Green JK, Holman RC, Jason JM, Evatt BL. Hemophiliaassociated AIDS in the United States, 1981 to September 1987. Am J Public Health. 1988;78:439-42. 22. Goedert JJ, Kessler CM, Aledort LM, et al. A prospective study of human immunodeficiency virus type 1 infection and the development of AIDS in subjects with hemophilia. N Engl J Med. 1989; 321:1141-8. 23. United States Bureau of the Census. Historical Statistics of the United States: Colonial Times to 1970. Bicentennial ed. part 1. Washington, DC; 1975. 24. National Office of Vital Statistics. Vital Statistics of the United States 1950. v I. Washington, DC; 1954. 25. United States Bureau of the Census. Statistical Abstract of the United States, 1989. 109th ed. Washington, DC; 1989. 26. Cochran WG. Sampling Techniques. 3d ed. New York: Wiley; 1978. 27. Little RJ, Rubin DB. Statistical Analysis with Missing Data. New York: Wiley; 1987. 28. Maxwell AE. Analysing Qualitative Data. London: Methuen; 1961. 29. Nilsson IM, Hedner U, Ahlberg A. Haemophilia prophylaxis in Sweden. Acta Paediatr Scand. 1976;65:129-35. 30. Bulloch W, Fildes P. Haemophilia. In: Pearson K, ed. Theory of Human Inheritance, v I. London: Cambridge University Press; 1912. 31. Kerr CB. The fortunes of haemophiliacs in the nineteenth century. Med History. l*963-64;7:359-70. 32. Baehner RL, Strauss HS. Hemophilia in the first year of life. N Engl J Med. 1966;275:524-8. 33. Agle DP. Psychiatric studies of patients with hemophilia and related states. Arch Intern Med. 1964;114:76-82. 34. Agle DP, Matsson A. Psychiatric and social care of patients with hereditary hemorrhagic disease. In: Ratnoff OD, ed. Treatment of Hemorrhagic Disorders. New York: Harper and Row; 1968. 35. Goodnough LT, Saito H, Ratnoff OD. Thrombosis or myocardial infarction in congenital clotting factor abnormalities and chronic thrombocyopenias: a report of 21 patients and a review of 50 previously reported cases. Medicine (Baltimore). 1983;62:248-55.

15 April 1991 • Annals of Internal Medicine • Volume 114 • Number 8

Downloaded From: https://annals.org/ by a Glasgow Univ Library User on 09/14/2018

The changing prognosis of classic hemophilia (factor VIII "deficiency").

To estimate relative risk of mortality and median life expectancy for patients with classic hemophilia by a life-table analysis, taking into account d...
1MB Sizes 0 Downloads 0 Views