Comparison of the long-term mortality in Hodgkin’s disease patients with that of the general population*

Annals of Oncology 10: 1199-1205, 1999. © 1999 Kluwer Academic Publishers. Printed in the Netherlands.

Original article Comparison of the long-term mortality in Hodgkin's disease patients with that of the general population* M. Provencio,1 F. J. Garcia-Lopez,2 F. Bonilla1 & P. Espana1 'Department of Oncology, 2Unit of Clinical Epidemiology, Hospital Universitaho 'Clinica Puerto de Hierro', Madrid, Spain

overall survival rates were 80% at 5 years, 70% at 10 years, and 64% at 15 years after diagnosis. The standardised mortality Background: Although Hodgkin's disease can be treated suc- ratios were 10.8 (95% confidence interval: 9.0-12.8, P < 0.0001) cessfully, its long-term survival rate has yet to be definitely overall, 5.5 in patients in favourable stages (IA, IIA), and 15.2 established. We compared the long-term mortality rate of in those with unfavourable stages (IB, IIB, III, IV). There was patients diagnosed as having Hodgkin's disease with that of a decreasing trend in mortality by calendar period (standardised mortality ratios for 1967-1975, 1976-1974 and 1985-1993: the general population. Patients and methods: We studied a retrospective cohort of 16.8,10.3 and 5.1, respectively). Higher mortality was observed 477 patients who received pathology-confirmed diagnoses of in all periods of follow-up after diagnosis, even after 20 years. Conclusions: Despite the improvements in treatment, morHodgkin's disease between 1967 and 1993 and were treated with combined chemotherapy or radiotherapy with follow-up from tality in Hodgkin's disease remains higher than in the general the day of diagnosis. Standardised mortality ratios were com- population in all disease stages, even 20 years after diagnosis. puted with reference rates taken from the Spanish population. Results: The follow-up was complete for 427 (89.5%) of the patients. The median follow-up time was 8 years, 133 patients Key words: Hodgkin's disease, longitudinal study, mortality, (28%) died and the median survival time was 21 years. The survival analysis Summary

Introduction

Having been considered an incurable disease for many years, Hodgkin's disease first began to be successfully treated about 25 years ago [1]. The definition of cure of Hodgkin's disease accepted at that time [2, 3], based on a disease-free survival of over one decade after treatment and a mortality rate similar to that of the normal population of the same age and sex, still applies. The development of radiotherapy techniques together with modern chemotherapy have made Hodgkin's disease almost curable, as shown by the fact that around 90% of patients in favourable stages I and II and 70% of all patients are considered to be cured [4-7]. Although there is little evidence of specific causes of mortality among patients diagnosed with Hodgkin's disease [8], the criterion used for cure refers to tumour progression-free survival. It is assumed that patients cured of Hodgkin's disease and followed for a long period with no evidence of relapse have survival rates similar to those of the general population, despite the fact that only a few studies have compared causes of death between these patients and the general population of the same sex and age [5, 9]. * Presented in part at the annual meeting of the American Society of Clinical Oncology, Los Angeles, CA, USA, 16-19 May 1998.

In this study, our aim was to determine the long-term survival of patients with Hodgkin's disease after treatment and compare it with the expected survival of the general population. Our purpose was to test the hypothesis of an eventual definitive cure for Hodgkin's disease. Patients and methods Population study Because the university hospital 'Clinica Puerta de Hierro' has been a national reference centre for lymphomas since the 1960s, we proposed to retrospectively identify all patients diagnosed with Hodgkin's disease since 1967, the year when storage of complete patient clinical records was initiated, to 1993. To be considered for inclusion, all patients had to have had a pathologically-confirmed diagnosis of Hodgkin's disease. In the 48 cases in which the diagnosis was uncertain, the pathology findings were reviewed and 19 cases in which the diagnosis remained unclear were excluded. The histopathological classification was based on the Lukes and Butler criteria [10]. Those patients diagnosed according to the Jackson and Parker classification and those in whom the subtype could not be determined were categorized as having non-specified Hodgkin's disease. A total of 477 patients were studied.

Measurement of variables The classification by stage was based on the criteria defined at the Rye meeting in 1972 and subsequently modified at the Ann Arbor [11] and

1200 Cotswolds [12] meetings. For statistical purposes, the stages were designated 'favourable' (stages IA and IIA) or 'unfavourable' (stages IB, IIB, III and IV). The chemotherapy treatment consisted of the combinations MOPP [13] (mechlorethamine, vincristine, procarbazine and prednisone), ABVD [14] (doxorubicin, bleomycin, vinblastine and dacarbazine), or a mixture of the two. The radiotherapy treatment was applied according to the recommendations of Kaplan [15] and adapted to the needs of our hospital by Aragon et al. [16]. The following criteria were used to define the type of response to treatment: a complete response was defined as the disappearance of all evidence of disease, both clinical and radiological, for at least four weeks after the end of treatment. In instances of a residual mass, response was defined as complete if the mass was stable for at least one month after the end of radiotherapy treatment or for the duration of two complete chemotherapy cycles, and the patient had no other clinical or analytical signs of lymphoma activity [17]. An incomplete response was defined as any response that was not complete: either some kind of partial response or none whatsoever. A relapse was defined as the confirmation of active disease, either by histology or radiology, occurring at least four weeks after the achievement of a complete response. When a patient died during the initial treatment, he or she was excluded from the response classification. The cause of death was classified as 'primary tumour' if it was directly related to Hodgkin's disease, 'secondary tumour' if it was due to a secondary neoplasm, and 'other'. All of the clinical information was extracted from the clinical records and from the 40 autopsies. To ascertain the vital status, all patients were contacted and invited to come to the clinic. All patients who could not be contacted were considered lost to follow-up. The calendar period was divided into three equal periods chosen a priori, 1967-1975, 19761984 and 1985-1993.

Main outcome variable The outcome variable was the standardised mortality ratio (SMR), which is the ratio between the number of deaths observed in the cohort and the number of cases expected if the cohort had the same mortality rates as the general population. This measure of association is interpreted as the ratio of two standardised rates, the one from the cohort studied and the one from the general population, when the age and sex distribution of cohort of patients with Hodgkin's disease are used as the standard. If the estimated SMR is equal to 1, the study group is considered to have had a mortality similar to that of the general population, whereas if the estimated SMR is greater than 1, the study group is considered to have had a mortality rate higher than that of the general population as many times as the number of units of the estimated SMR. To arrive at the follow-up time for each subject, the time interval between the date of diagnosis and the date of death, the date the patient was lost to follow-up, or 31 December 1993, whichever came first, was computed in years. The follow-up times for all subjects were stratified according to sex, age,five-yearage groups, and calendar time from 1967 to 1993. In 67 cases (14%) in which the day and month of birth were missing, 1 July was the date used. The total number of person-years was computed for each of these strata using a statistical programme. We used the Spanish population for comparison. The Spanish annual mortality rates from 1975 to 1993, the last year for which data were available, were provided by the Instituto Nacional de Estadistica, stratified by sex and age groups (less than one year of age, one to four years, and thenfive-yearage groups from 5-9 up to 75-79, plus a group of 80 years and over, until 1979 and two groups, from 80-84 and 85 and over, from 1980 on) [18]. From 1967 to 1974 the mortality rates were computed by the authors with data taken from two sources, the Statistical Annuals from the Instituto Nacional de Estadistica, which gave the births occurring annually in Spain and the deaths classified by sex and five-year age groups up to 90 years, and the population censuses of 1960, 1970, and 1981, classified by sex and five-year age groups up to 15 years and 10-year age groups up to 65 years. As there were no accurate data on migratory movements in that period, we estimated the population for the years between censuses in each age

stratum as the proportional part of the difference between censuses corresponding to the number of years passed since the previous census. The deaths not classified by age in the Statistical Annuals (0.51%) were assigned to all strata following weights proportional to the number of deaths in each stratum. For each sex and age stratum, mortality rates were computed dividing the estimated total deaths over the estimated population. For the period 1967-1974, mortality rates are shown in five-year age groups up to 15 years, in 10-year age groups up to 65 years and a group of 65 years and over. In each stratum, the total expected deaths for the general population were determined by multiplying the person-years by the corresponding mortality rate for that stratum. The total expected cases for the study population as a whole were determined by adding together the expected cases for each stratum. For each standardised mortality ratio, a 95% confidence interval (95% CI) was computed using the ratio between the corresponding limits of the Poisson distribution of the number of observed cases and the number of expected cases [19]. When the number of observed cases exceeded 100, the limits of the Poisson distribution were extrapolated according to the Freeman and Tukey method [20]. For each standardised mortality ratio, hypothesis testing was performed using the Byar approximation to the exact Poisson test [21], The excess of deaths, the difference between the number of observed deaths in the study group and the number of expected deaths in the general population, was also computed and reported as the number of additional deaths per 100 person-years. Cumulative survival probability The survival probability in the cohort as a whole, overall and stratified by sex, was computed by the product-limit estimate method. For each median, 95% confidence intervals were calculated [22]. To compare the cohort survival probability to that of the general population, we used the life table method [23]. The partial and total survival probabilities were estimated for each year after the date of diagnosis, taking into account the number of patients alive at the beginning of the year in question, the number of patients who died during the succeeding year and the number of patients lost to follow-up during that period. The method of transformation to a logarithmic scale was used to compute the standard errors of the annual partial and total probabilities [22]. The cumulated survival probability curves, with their corresponding 95% confidence limits, were then compared with the expected cumulative survival probability from the general population standardised to an age and sex distribution similar to that of the studied cohort. The annual probability of death in the general population was first calculated by dividing the number of expected deaths over the total number of person-years observed for each year. Then, the probability of survival was calculated as the complement of the probability of death and the cumulative survival probability as the product of each annual survival probability. Finally, relative survival estimates, the ratios between the observed and the expected cumulative probabilities at different years of follow-up and their corresponding confidence intervals were computed. The time when relative survival becomes constant indicates the moment when deaths due to the disease under study are no longer recorded [24].

Multivariate statistical analysis To study the age- and sex-adjusted influence of disease staging and calendar period on mortality, we performed an SMR multivariate analysis following a Poisson multiplicative model, OBS,vt = EXP,* • SMR • exponential (x/tP), or, in other terms, Log OBS,* = Log EXP,* + u + xJ;tp, where OBS/t is the expectancy of the number of deaths for each stratum of sex, age (/) and the exposed or studied variable (k), EXP is the number of expected deaths for each stratum of sex and age (/) in the general population after application of the population mortality rates, u denotes the SMR logarithm in the reference group of the exposed variable, Xy^ is the configuration of variables in stratay and k, and p indicates the parameters for the variables under study,

1201 Table I. Patient characteristics of patients diagnosed with Hodgkin's disease (n = 477).

Population study

Characteristics Sex

Male Female Age at diagnosis (years) 0-19 20-29 30-39 40-59 60+

Calendar year at diagnosis 1967-1975 1976-1984 1985-1993 Clinical stage at diagnosis IA IB IIA IIB

IIIA IIIB IVA IVB

Response to treatment Complete Non-complete Relapses Follow-up time Less or equal 5 years > 5 years and less or equal 10 years > 10 years and less or equal 15 years > 15 years and less or equal 20 years > 20 years Deaths Due to primary tumour Due to secondary tumour Due to other causes Infectious disease Digestive diseases Cardiac disease Cerebrovascular disease Pulmonary fibrosis Accident Unknown

Results

301 (63) 176(37) 94 (20) 143 (30) 117(25) 98(21) 25(5) 109(23) 175(37) 193 (40) 85(18) 8(2) 139(29) 61(13) 56(12) 62(13) 16(3) 50(11) 437 (92) 36(8) 138(29) 146(31) 170(36) 86(18) 49(10) 26(5) 133(28) 49 (37) 26(19) 58 (44) 33 10 7 3 1 1 3

Because of rounding up the total of some percentages is not 100. In the 'Response to treatment' section, four cases were missing. The percentages in the death causes are calculated over the 133 deaths.

age, sex, disease staging and calendar period, whose exponentiation multiplies the SMR in the reference exposure group. Since comparisons between two or more SMRs may be misleading if the specific age rates for the cohorts to be compared are not proportional to the reference population rates [19, 25], we indirectly checked that condition with the goodness of fit of the model. The final model included all of the variables studied, sex, age, stage and period, as independent variables. The goodness of fit was checked through the deviance (G2), the Pearson residuals (x2), the addition of interactions, and the graphical analysis of residuals [19, 26]. The chi-squared likelihood ratio test was used to assess whether a variable should be included in the model. Survival rates by sex were compared using the log-rank test. The level of statistical significance was set at 0.05 for all comparisons. All statistical analyses were carried out with the statistical software GLIM 4.0 [27].

We studied 477 patients, with a median follow-up time of 8 years (range 0-26 years). The follow-up was complete for 427 (89.5%), although for 58 cases known to be alive the last information was obtained between 1992 and 1993. The median age at diagnosis was 30 years (range 4-85). The sex ratio (male: female) at diagnosis was 1.7:1. Of the total group of patients, 133 (28%) died (Table 1) with a median time to death after diagnosis of 3 years (mean 5 years, range 0-25 years). The median time to death of those who died of a progression of the primary tumour (37%) was 2.2 years (range 0-14.3 years); of those who died of a second cancer (19%), 7.2 years (range 1.3-21.1 years) and of those who died of other causes (44%), 3.3 years (range 0-25.1 years). The overall survival rate was 80% at 5 years, 70% at 10 years, and 66% at 15 years after diagnosis. The median survival time was 21.1 years (95% CI: 18.8-23.3 years). Survival time was longer in women (median 24.5 years, 95% CI: 22.8-26.1) than in men (median 17.8, 95% CI: 16.3-19.4 years, P = 0.0004). Standardised mortality ratios In the overall analysis, the SMR for patients with Hodgkin's disease was 10.8 (95% CI: 9.0-12.8, P < 0.0001). The SMRs were consistently high for both sexes (Table 2). The excess of deaths per 100 person-years in the overall sample of Hodgkin's disease patients was 3.2 (Table 2). The analyses by five age strata (under 20, 20-29, 30-39, 40-59 and 60 years or over) showed SMRs significantly greater than 1 in all age groups; they tended to diminish with age whereas the excess of deaths increased slightly with age (Table 3). The SMR was higher in the 'unfavourable' stages IB, IIB, III and IV than in the 'favourable' stages IA and IIA (Table 4). The SMRs and excess of deaths declined with calendar periods under study, 1967-1975,1976-1984, and 1985-1993 (Table 5). In the analysis of long-term mortality, in five-year periods after diagnosis, the SMRs and excess of deaths were consistently higher than 1 in all periods, even 20 years after diagnosis (Table 6). We compared the mortality unrelated to Hodgkin's disease with the general population assuming that, given the low incidence of Hodgkin's disease in the general population, its contribution to the general mortality rates is negligible. A total of 84 patients died of causes unrelated to Hodgkin's disease and 12.4 were expected, giving an SMR of 6.8 (95% CI: 5.4-8.4, P < 0.0001, excess of deaths: 1.9/100 person-years). In men, the SMR was 5.8 (95% CI: 4.4-7.5, P < 0.0001) and in women, it was 10.9 (95% CI: 7.0-16.1, P < 0.0001).

1202 Table 2. Standardised mortality ratios (SMR) in patients with Hodgkin's disease. Overall analysis: all the patients. 1967-1993. Group

No.

Personyears

O

E

SMR

Total Men Women

477 301 176

3805 2215 1591

133 98 35

12.4 10.1 2.3

10.8 9.7 15.3

95% CI

9.0 7.9 10.7

12.8 11.9 21.3

Excess of deaths/ 100 person-years

/"-value

3.2 4.0 2.1

< 0.0001 < 0.0001 < 0.0001

Abbreviations: No - total number of subjects; O - observed deaths; E - expected deaths; SMR - standardised mortality ratio; CI - confidence interval; P - probability.

Table 3. Standardised mortality ratios in patients with Hodgkin's disease: analysis by age groups at diagnosis. Age group

No.

Personyears

O

E

SMR

95% CI

Excess of deaths

/'-value

0-19

Total Men Women

94 61 33

918.9 578.3 340.6

13 11 2

0.7 0.6 0.1

17.3 18.6 12.5

9.3 9.3 1.5

29.6 33.4 45.1

1.3 1.8 0.5

< 0.0001 < 0.0001 0.0147

20-29

Total Men Women

143 73 70

1178.7 517.9 660.8

34 21 13

1.3 0.8 0.4

26.5 25.0 29.6

18.4 15.5 15.7

37.1 38.2 50.5

2.8 3.9 1.9

< 0.0001 < 0.0001 < 0.0001

30-39

Total Men Women

117 78 39

952.2 615.4 336.8

32 26 6

2.2 1.8 0.4

14.7 14.6 15.1

10.0 9.5 5.5

20.7 21.3 32.7

3.1 3.9 1.7

< 0.0001 < 0.0001 < 0.0001

40-59

Total Men Women

98 69 29

633.6 414.6 219.0

44 31 13

4.5 3.8 0.7

9.7 8.1 18.6

7.1 5.5 9.9

13.0 11.5 31.8

6.2 6.5 5.6

< 0.0001 < 0.0001 < 0.0001

60+

Total Men Women

25 20 5

122.0 88.5 33.5

10 9 1

3.6 3.0 0.6

2.8 3.0 1.7

1.3 1.4 0.0

5.1 5.7 9.4

5.2 6.8 1.2

0.0043 0.0042 0.4415

Abbreviations: No - total number of subjects; O - observed deaths; E - expected deaths; SMR - standardised mortality ratio; CI - confidence interval; P - probability. The excess of deaths is expressed in 100 person-years.

Table 4. Standardised mortality ratio in patients with Hodgkin's disease. Analysis by disease stage at diagnosis. Stage

No.

Personyears

O

E

SMR

95% CI

Excess of deaths

P-value

IA, IIA

Total Men Women

224 129 95

1779.5 952.2 827.3

31 18 13

5.6 4.3 1.3

5.5 4.2 9.9

3.7 2.5 5.3

7.8 6.6 16.9

1.4 1.4 1.4

< 0.0001 < 0.0001 < 0.0001

IB, IIB, III ylV

Total Men Women

253 172 81

2025.9 1262.5 763.4

102 80 22

6.7 5.7 1.0

15.2 13.9 22.7

12.4 11.1 14.2

18.5 17.3 34.3

4.7 5.8 2.8

< 0.0001 < 0.0001 < 0.0001

See footnote to Table 3.

Cumulative survival probability

Results ofstatistical modelling

The estimates of the overall survival probability with their 95% CI were always lower than the expected survival probability of the general population (Figure 1). The overall relative survival estimates at 5, 10, 15 and 20 years were 0.82 (95% CI: 0.78-0.85), 0.73 (95% CI: 0.68-0.78), 0.71 (95% CI: 0.65-0.76) and 0.60 (95% CI: 0.51-0.69), respectively. The estimates of survival probability in men and women were also lower than the corresponding expected probabilities of the general population (Figure 2).

Age over 40, diagnosis before 1985 and unfavourable stages were risk factors for increased mortality in Hodgkin's disease when compared to the general population in both the univariate and multivariate analyses (Table 7). Female sex, a risk factor for mortality in the univariate analysis, was no longer a risk factor in the multivariate analysis when age, calendar period and disease staging were included in the model.

1203 Table 5. Standardised mortality ratio in patients with Hodgkin's disease: analysis by calendar period. Period

No.

Person-years

O

E

SMR

95% CI

Excess of deaths

f-value

1967-1975

Total Men Women

109 68 41

1258.0 687.6 570.4

67 47 20

4.0 3.2 0.8

16.8 14.9 24.4

13.1 10.9 14.9

21.4 19.8 37.7

5.0 6.4 3.4

< 0.0001 < 0.0001 < 0.0001

1976-1984

Total Men Women

175 112 63

1541.4 904.3 637.1

46 35 11

4.4 3.5 0.9

10.3 9.9 11.7

7.5 6.9 5.9

13.7 13.8 21.0

2.7 3.5 1.6

< 0.0001 < 0.0001 < 0.0001

1985-1993

Total Men Women

193 121 72

1006.0 622.8 383.2

20 16 4

3.9 3.4 0.5

5.1 4.7 7.6

3.1 2.7 2.1

7.9 7.7 19.4

1.6 2.0 0.9

< 0.0001 < 0.0001 0.0027

See footnote to Table 3.

Table 6. Standardised mortality ratios in patients with Hodgkin's disease. Analysis by years of follow-up. Follow-up period

No.

O

E

SMR

95% CI

Excess of deaths

/•-value

0-5 years

Total Men Women

477 301 176

87 69 18

6.3 5.2 1.1

13.8 13.3 16.4

11.1 10.3 9.7

17.1 16.8 25.9

16.9 21.2 9.6

< 0.0001 < 0.0001 < 0.0001

6-10 years

Total Men Women

333 196 137

29 20 9

3.5 2.8 0.7

8.5 7.2 13.7

5.7 4.4 6.3

12.1 11.1 26.0

7.7 8.8 6.1

< 0.0001 < 0.0001 < 0.0001

11-15 years

Total Men Women

161 89 72

6 1 5

1.4 1.1 0.3

4.3 0.9 16.9

1.6 0.0 5.5

9.3 5.0 39.6

2.9 -0.1 6.5

0.0036 0.2993 < 0.0001

16-20 years

Total Men Women

75 39 36

9 8 1

0.9 0.7 0.2

9.8 10.7 5.8

4.5 4.6 0.1

18.6 21.0 32.5

10.8 18.6 2.3

< 0.0001 < 0.0001 0.1588

> 20 years

Total Men Women

26 11 15

4 2 2

0.2 0.1 0.1

23.1 18.4 30.8

6.3 2.2 3.7

59.0 66.5 111

14.7 17.2 12.9

< 0.0001 0.0082 0.0040

See footnote to Table 3.

Expected survival for females Expected survival for males

Observed survival

Observed survival for females

95% confidence interval

Observed survival for males 6 0

2

4

6

8

10

12

14

16

18

.. Time after diagnosis (years) No at ^ 477 413 362 312 227 161 114 84 66 44

20

22

24

26

16

6

28

Discussion Despite the advances introduced in the treatment of Hodgkin's disease over the past three decades, in our

10

12

14

16

18

20

301

254

213

186

137

89

63

47

31

19

11

6

Women 176

159

149

126

90

72

51

37

35

25

15

10

Men

Figure 1. Cumulative probability of annual survival for the cohort of patients with Hodgkin's disease when compared with the estimated cumulative probability for the general population.

8

Time after diagnosis (years)

Figure 2. Cumulative probability of annual survival for the cohort of patients with Hodgkin's disease when compared with the estimated cumulative probability for the general population by sexes.

patients with Hodgkin's disease we found the rate of mortality to be more than 10-fold higher than that of the general population and there were over 3 deaths more per 100 person-years. This higher mortality rate was observed in both men and women, all age groups, all

1204 Table 7. Mortality in patients with Hodgkin's disease compared to the general population. Relative risk estimates. Variable

Sex Female Age (years) 20-29 30-39 40-59 >60 Period 1976-1984 1985-1993 Stage Unfavourable

Univariate analysis

Multivariate analysis

RR

95% CI

/"-value

RR

95% CI

P-value

1.57

1.1 2.3

0.03

1.37

0.9 2.1

0.13

1.53 0.84 0.56 0.16

0.8 2.9 0.4 1.6 0.3 1.0 0.1 0.4

< 0.0001

1.47 0.78 0.52 0.22

0.8 2.8 0.4 1.5 0.3 1.0 0.1 0.5

< 0.0001

0.61 0.31

0.4 0.9 0.2 0.5

<0.0001

0.81 0.53

0.5 1.2 0.3 0.9

0.04

2.77

1.9 4.1

<0.0001

2.25

1.5 3.4

0.0001

Abbreviations: RR - relative risk obtained from a Poisson multiplicative regression model of the observed death cases on the expected cases of the general population and the following variables: sex, age, calendar period and disease staging; CI - confidence interval. P - probability. Unfavourable stage indicates stages IB, MB, III or IV. The reference groups are the following: in sex, men; in age, subjects under 20; in calendar period, 19671975; in disease staging, the favourable stages (IA, IIA).

stages of disease, and every calendar period; in particular, younger ages, unfavourable stages and earlier periods had the highest rates. The higher mortality was sustained throughout the entire follow-up period, even after 20 years, and it never stabilised. As only 37% of deaths were due to the primary tumour, the excess of mortality must also have been caused by secondary tumours and other factors, which explains the almost seven-fold increase in the risk for death. The reasons for this excess are still uncertain [28], although there is some evidence of a higher incidence of second malignant neoplasms [29, 30], thought to be related either to the toxicity of treatment or to intrinsic host factors linked to the initial Hodgkin's disease, and sparse evidence for a higher incidence of cardiovascular complications [31—33]. The direction of our findings is consistent with other studies comparing the long-term mortality of patients with Hodgkin's disease with that of the general population [5, 34] but we found an overall standardised mortality ratio 1.5 higher than the figure of 7.09 reported previously by the International Data Base on Hodgkin's Disease (IDHD) [5]. Our findings also agree with the observation of a decrease in mortality from Hodgkin's disease in Europe in recent decades [35, 36]. In contrast to the IDHD study, which affirms that 67% of patients die of disease progression, in our study we found it to be the cause of death in only 37%. The reasons for this discrepancy are unclear, although the proportion of ascertainment by necropsy and the coding system of causes may have been different. Our study has some strengths. The proportion of patients lost to follow-up was quite low (3.8%). Moreover, the follow-up time was very long, spanning the entire period since Hodgkin's disease was diagnosed and treated effectively. Furthermore, our study helps to correctly interpret the influence of some factors on mortality when a population approach

is taken. While men seem to have a worse prognosis than the women in the cohort when their mortality rates are compared to those of the general population, both men and women with Hodgkin's disease have the same relative increase in mortality as their counterparts in the general population, for men usually have higher mortality rates than women. Likewise, although as compared with patients with Hodgkin's disease under 20 years of age, older patients have an increased risk of dying, especially those aged from 20 to 60, when compared with the general population, the relative risk of mortality is the same up to 40 years of age and begins to decline for those over 40, probably because of competing causes of death. Finally, our study shows that, when analysed simultaneously, age, sex, disease stage and calendar period all play a role in increasing the risk of death in the general population. This study has several limitations. Firstly, the sample studied may not be representative of the population of patients with Hodgkin's disease since it is not taken from a population-based cancer registry. Although we believe that Hodgkin's disease is usually diagnosed and treated in hospitals and is not frequently missed, each hospital may have special characteristics with respect to social class, age, sex or health care received that might to some extent affect prognosis. Moreover, as our hospital has been a reference centre for Hodgkin's disease for many years, other centres may have referred those patients with worse prognoses to ours. However, the age, sex and disease stage distribution in our sample did not differ from the large cohorts reported so far [5]. In our cohort, the complete response rate after the first treatment was 92% for the whole sample, slightly higher than in the IDHD (87%) [5]. We found a relapse rate of 29%, similar to that in other reports [37-40]. Secondly, the data collection was retrospective, which may have caused some misclassification. In 4% of patients, we could not identify a precise histopathological type despite the review of preparations, although other authors have found percentages of uncertainty as high as 19% [41]. To reduce these problems, the diagnosis and disease stage were determined in accord with well established criteria. Thirdly, the mortality rates during the first 8 years studied were provided for 10-year age intervals instead of the usual intervals of 5 years. This may have affected the assumption underlying the life-table method of equal risk of death within each interval. In addition, the lack of accurate data on population migration during the years between censuses may have biased the estimates of population mortality rates. However, the consistency of our findings across all age groups and of the decreasing time trend suggest that these inaccuracies should not have had a strong influence on the results. Lastly, only a few patients were followed for 20 years or longer. Nevertheless, the standardised mortality ratio in these patients was still higher than in the general population. In conclusion, our findings do not support the hypothesis that there will ultimately be a definite cure

1205 for Hodgkin's disease, and although the rate of mortality tends to diminish over time, it is still quite high. This excess of mortality could be due in part to the treatment administered, and further studies on the causes of the long-term mortality may shed light on identifying potential preventable factors for death.

Acknowledgement We are greatly indebted to M. Messman for her help in editing the manuscript. References 1. Farah R, Wechselbaum RR. Management of stage I and II Hodgkin's disease. In Williams SF, Farah R, Golomb HM (eds): Hematology/Oncology Clinics of North America. Philadelphia, PA: WB Sanders Company 1989; 253-64. 2. Frei E, Gehan EA. Definition of cure for Hodgkin's disease. Cancer Res 1971; 31: 1828-33. 3. Easson EC, Russel MH. The cure of Hodgkin's disease. BMJ 1963; 1: 1704-7. 4. Rosenberg SA, Kaplan HS. The evolution and summary results of Stanford randomized clinical trials of the management of Hodgkin's disease: 1962-1984. Int J Radiat Oncol Biol Phys 1985; 11: 5-22. 5. Henry-Amar M, Somers R. Survival outcome after Hodgkin's disease: A report from the International Data Base on Hodgkin's disease. Semin Oncol 1990; 17: 758-68. 6. Farah R, Ultmann J, Griem M et al. Extended mantle radiation therapy for pathologic stage I and II Hodgkin's disease. J Clin Oncol 1988; 6: 1047-52. 7. Mauch P. Controversies in the management of early stage Hodgkin's disease. Blood 1994; 83: 318-29. 8. Henry-Amar M, Joly F. Late complications after Hodgkin's disease. Ann Oncol 1996; 7 (Suppl 4): SI 15-26. 9. Hoppe R. Hodgkin's disease: Complications of therapy and excess mortality. Ann Oncol 1997; 8 (Suppl 1): SI 15-8. 10. Lukes RJ, Butler JJ. The pathology and nomenclature of Hodgkin's disease. Cancer Res 1966; 26: 1063-81. 11. Carbone PP, Kaplan HS, Musshoff K et al. Report of the Committee on Hodgkin's disease staging classification. Cancer Res 1971; 31: 1860-1. 12. Lister TA, Crowther D, Sutcliffe SB et al. Report of a committee convened to discuss the evaluation and staging of patients with Hodgkin's disease: Cotswolds meeting. J Clin Oncol 1989; 7: 1630-6. 13. DeVita VT, Serpick AA, Carbone PP. Combination chemotherapy in the treatment of advanced Hodgkin's disease. Ann Intern Med 1970; 73: 891-5. 14. Bonadonna G, Zucali R, Monfardini S et al. Combination chemotherapy of Hodgkin's disease with adriamycin, bleomycin, vinblastine, and imidazole carboxamide versus MOPP. Cancer 1975; 36: 252-9. 15. Kaplan HS, Rosenberg SA. Extended field radical radiotherapy in advanced Hodgkin's disease. Cancer Res 1966; 26: 1268-76. 16. Aragon G, Cardenes H, Otero J et al. Individual risk of abdominal disease in patients with stages I and II supradiaphragmatic Hodgkin's disease. A rule index based on 341 laparotomized patients. Cancer 1989; 63: 1799-803. 17. Urba WJ, Longo D. Hodgkin's disease. N Engl J Med 1992; 10: 678-87. 18. Data provided by the Instituto Nacional de Estadistica, Subdireccion General de Estadisticas Demograficas, 1996. 19. Breslow NE, Day NE. Statistical Methods in Cancer Research. Vol II. The Design and Analysis of Cohort Studies. Lyon, France: International Agency for Research on Cancer .1987.

20. Documenta Geigy. Tablas cientificas, 6th edition. Basilea, Switzerland: JR Geigy SA 1965. 21. Rothman KJ, Boice JD Jr. Epidemiologic analysis with a programmable calculator. Chestnut Hill, MA: Epidemiology Resources Inc 1982. 22. Parmar MKB, Machin D. Survival Analysis. A Practical Approach. Chichester, UK: John Wiley & Sons 1995. 23. Lawless JF. Statistical Models and Methods for Lifetime Data. New York: John Wiley & Sons 1982. 24. Esteve J, Benhamou E, Raymond L. Statistical Methods in Cancer Research. Vol. IV. Descriptive Epidemiology. Lyon, France: International Agency for Research on Cancer 1994. 25. Clayton D, Hills M. Statistical Models in Epidemiology. Oxford, UK: Oxford University Press 1993. 26. Lindsey JK. Modelling Frequency and Count Data. Oxford, UK: Oxford University Press 1995. 27. Francis B, Green M, Payne C (eds). GLIM 4. The Statistical System for Generalized Linear Interactive Modelling. Release 4 manual. Oxford, UK: Oxford University Press 1994. 28. Henry-Amar M, Hayat M, Meerwaldt JH et al. Causes of death after therapy for early stage Hodgkin's disease entered on EORTC protocols. EORTC Lymphoma Cooperative Group. Int J Radiat Oncol Biol Phys 1990; 19: 1155-7. 29. Rubin P, Zagars G, Chuang C, Thomas EM. Hodgkin's disease: Is there a price for successful treatment? A 25-year experience. Int J Radiat Oncol Biol Phys 1986; 12: 153-66. 30. Henry-Amar M. Second cancer after the treatment for Hodgkin's disease: A report from the International Database on Hodgkin's disease. Ann Oncol 1992; 3 (Suppl 4): SI 17-28. 31. Boivin JF, Hutchison GB. Coronary heart disease mortality after irradiation for Hodgkin's disease. Cancer 1982; 49: 2470-5. 32. Mauch P, Kalish L, Marcus K et al. Long-term survival in Hodgkin's disease: Relative impact of mortality, secondary tumor, infection and cardiovascular disease. Cancer J Sci Am 1995; 1: 33-42. 33. Vanghan Hudson B, Vanghan Hudson G, Linch D et al. Late mortality in young British National Lymphoma Investigation patients cured of Hodgkin's disease. Ann Oncol 1994; 5: 565—6. 34. van Rijswijk REN, Verbeek J, Haanen C et al. Major complications and causes of death in patients treated for Hodgkin's disease. J Clin Oncol 1987; 5: 1624-33. 35. La Vecchia CL, Levi F, Lucchini F et al. Hodgkin's disease mortality in Europe. Br J Cancer 1991; 64: 723-34. 36. Levi F, La Vecchia C, Lucchini F et al. Mortality from Hodgkin's disease and other lymphomas in Europe, 1960-1990. Oncology 1995; 52: 93-6. 37. Bodis S, Henry-Amar M, Bosq J et al. Late relapse in early-stage Hodgkin's disease patients enrolled in EORTC protocols. J Clin Oncol 1992; 11: 225-32. 38. Lohri A, Barnett M, Fairey RN et al. Outcome on first relapse of Hodgkin's disease after primary chemotherapy: Identification of risk factors from the British Columbia experience 1970 to 1980. Blood 1991; 77: 2292-8. 39. Viviani S, Santoro A, Negretti A et al. Salvage chemotherapy in Hodgkin's disease. Ann Oncol 1990; 1: 123-7. 40. Fisher RI, DeVita VT, Hubbard RN et al. Prolonged disease free survival in Hodgkin's disease with MOPP reinduction after first relapse. Ann Intern Med 1979; 90: 761-3. 41. Bernhars J, Fischer R, Hubner K et al. Histopathological classification of Hodgkin's lymphomas. Ann Oncol 1992; 3 (Suppl 4): 31-3. Received 14 May 1999; accepted 17 August 1999. Correspondence to: M. Provencio, MD Servicio de Oncologia Hospital Universitario 'Clinica Puerta de Hierro' calle San Martin de Porres, 4 28035 Madrid Spain J£-mail: [email protected]