Prognostic Factors in Hodgkin’s Disease: Multivariate Analysis of 327 Patients from a Single Institution

Owing to the progress in its treatment,HD² has become a potentially curable disease. An appropriate therapy scheme, chosen in compliance with prognostic factors, is a very important condition for cure (1). However, the identification of patients at high risk for subsequent treatment failure still remains a very important unsolved question in HD. Despite a number of publications, often controversial, it is still suitable and necessary to present the results of observations on large groups of HD patients. Twenty-five years of experience with the model of combined treatment seems to be a good groundwork for a critical view and evaluation of the real value of factors considered important in the prognosis in HD (2, 3, 4).

In the last several years there has been a marked tendency to search for new, more specific and sensitive prognostic factors in HD. Among a number of laboratory parameters considered valuable in the prediction of outcomes in neoplastic patients, the alterations of some oncogenes involved in the pathogenesis of malignancies have become a particularly interesting problem. In solid tumors, as well as in non-Hodgkin’s lymphomas, the prognostic value of expression of PCNA and genes such p53 and bcl-2 on neoplastic cells has already been suggested (5). There are only few, rather inconsistent reports concerning this problem in HD.

On the basis of a retrospective study of 327 patients with HD, we have analyzed the prognostic significance of several factors with regard to response to treatment and survival time. Apart from a number of well known and traditionally used clinical and laboratory parameters, we have also examined the pretreatment expression of some oncoproteins on Reed-Sternberg cells and their mononuclear variants, Hodgkin’s cells(R-S/H cells), as a potentially important prognostic factor in HD.

A retrospective study was carried out in a group of 327 patients with HD, 132 females and 195 males, ages 17–65 years (mean, 41.3), who were treated and followed up in the period 1970–1995. Characteristics of the patients are presented in Tables 1 and 2. In all patients, a uniform scheme of chemotherapy was applied: between 1970 and 1981, COPP or MOPP regimens were used, and since 1982, alternating MOPP and ABVD regimens were used. After that, appropriate radiotherapy was usually performed (Table 3). The data discussed below were analyzed.

The clinical parameters used were sex, age at diagnosis, time between the first symptoms of disease and the diagnosis, presence of systemic symptoms, clinical status before treatment according to Karnofsky’s index, type and number of nodal areas involved (peripheral,mediastinal, or abdominal nodes), extranodal localization of the disease, huge primary tumor mass (“bulky disease”), response to treatment after four courses of chemotherapy, and adequacy of dosage of the first-line treatment.

Clinical parameters were evaluated on the basis of individual history of the disease in combination with physical examination, X-ray results,ultrasonography, CT methods, and, if necessary, exploratory laparotomy(Table 4). Clinical staging was performed in compliance with the Ann Arbor Conference principles.

One lymph node or group of nodes involved was treated as one nodular site of the disease. We have created three subgroups for statistical analysis: (a) one nodular site; (b) two or three nodular sites; and (c) four or more sites of lymph nodes involved. The bulky disease was determined using the following criteria: more than 1/3 of mediastinal diameter was involved,and/or the horizontal diameter of peripheral or abdominal nodular mass was ≥5 cm (6).

These parameters were histological type; ESR; number of leukocytes,erythrocytes, and platelets; absolute number of lymphocytes, monocytes,and eosinocytes; serum LDH activity; alkaline phosphatase activity;serum iron concentration; level of albumin; serum fractions of globulins (α, β-1, β-2, and γ-globulins); and renal and hepatic sufficiency tests. All morphological and biochemical parameters were estimated using standard laboratory methods.

These parameters included the expression of PCNA, as well as the protein products of the p53 and bcl-2 genes, on R-S/H cells. This retrospective examination could be performed in 194 patients on suitably retained, paraffin-embedded lymph nodes specimens(stored routinely in our center from the 1970s). Immunohistochemical examinations using monoclonal antibodies (DAKO A/S, Glostrup,Denmark) with PCNA and the protein products of the p53 and bcl-2 genes were performed in the lymph nodes obtained from nontreated HD patients. In all cases, the index of positively stained R-S/H cells was determined. The color reaction for standard immunohistochemical examination in the avidin-biotin complex system was obtained by 3,3-diaminobenzidine.

Morphological and biochemical factors analyzed in this study were considered according to commonly established cut-off points,determined by normal laboratory ranges. However, to take into account some biochemical parameters and immunohistochemical factors, as well as to estimate the treatment properly, the cutoff points were established according to statistical principles. Namely, on the basis of histograms (not shown) and frequency distribution tables for those factors examined, the middle value of the class comprising the median number of cases was chosen as a cutoff value.

In case of PCNA, p53, and BCL-2 expression, on the basis of frequency distribution for each antigen, the following cutoff values were chosen:for PCNA, 40%; for p53, 20%; and for BCL-2, 10%.

All of those parameters were analyzed for the OS and DFS and were correlated with response to the first-line therapy. DFS was defined as the time interval between the end date of the first-line treatment, if CR was obtained, and the date of eventual relapse.

Two categories of response to the first-line therapy were noted: CR and resistance to first-line treatment, i.e., ≤PR, which included partial remission, no change, or progressive disease.

According to standard criteria, CR was defined as a disappearance of all evidence of the disease without development of new lesions. Partial remission was defined as a greater than 50% decrease of all of the measurable pathological lesions, also without a development of new lesions. No change was defined as a decrease in pathological lesions less than 50% or an increase less than 25%. Progressive disease was defined as increase in the size of measurable lesion by at least 25%or appearance of any new lesion.

The whole panel of follow-up procedures and tests for HD patients used routinely in our center comprised physical examination,complete blood count and sedimentation ratio, blood chemistry, chest X-ray (or CT scan if necessary), and abdomen ultrasonography (or CT scan if necessary). Frequency of those examinations was as follows:every 3 months for the first 2 years after completing the whole treatment required and achieving CR; next, every 6 months for 3–5 years after the end of successful treatment; and finally, once a year after the 5th year after the treatment. Despite of the panel of follow-up tests, we have also controlled the CR patients with only physical and blood examination (if no clinical symptoms of relapse are present) during every month’s visits (1st and 2nd years after treatment) and then every 2–3 months up to 5 years, and every 6 months after the end of treatment.

Because the results of therapy may be significantly affected by the real doses of cytostatics given, the DI was assessed for all of the patients in this study. DI was calculated on the basis of the model of Hryniuk and Bush (7) and was expressed as the ratio of actually received average dose to the planned average dose over the same time frame as prescribed in the original protocols. The received actual DI was calculated on a per-patient/per-cycle basis and averaged across the patients who received the same regimen of chemotherapy. The total number of mg/m² was determined for each drug given throughout cycle of the treatment, for each patient. Cumulative dose of drug was calculated as its summary dose in all cycles applied to patient. Then, the total number of days between the first day of chemotherapy and one cycle time after the day of last treatment was divided by 28 (recommended time frame of one cycle) to give the divisor to determine the actually received average dose for each drug/for cycle.

The programmed cumulative dose of cytostatics was calculated in the same way. This differed in particular patients, depending on stage of the disease. The following optimal intensity of treatment(i.e., number of cycles and/or summary dose of radiotherapy planned) was stated for statistical calculations: for early stages(IA–IIA), radiotherapy, with a total dose 40 Gy (30 Gy for extended fields and 10 Gy for involved regions); for intermediate stages(IIB–IIIA), four cycles of chemotherapy and complementary radiotherapy(40 Gy); and for advanced stages (IIIB–IV), eight cycles of chemotherapy, with optional local radiotherapy for bulky sites (8). However, in the analyzed group were six patients in stage I B, with either bulky mediastinal tumor mass or infradiaphragmatic disease, who received 2–3 cycles chemotherapy as a complementation to mantle radio-therapy.

Statistical analysis was carried out on the basis of logistic forward stepwise multivariate regression test, as well as proportional hazard Cox regression model and Kaplan-Meier survival analysis, tested by the log-rank test. For the analysis of differences between means/medians,the following nonparametric tests were used when necessary:Mann-Whitney U test, Kruskal-Wallis ANOVA, and median test. Results were considered significant if P was less than 0.05.

The first-line chemotherapy with COPP, MOPP, or alternating MOPP/ABVD regimens was applied in 284 (86.8%) patients. Complementary mantle radiotherapy was given in 240 cases (73.4%), and local radiation was given in 40 cases (12.2%; Table 3).

CR was achieved in 257 patients (78.6%) after the first-line therapy. Eighty-seven of these patients (37.0%) subsequently relapsed: 40 of the relapses occurred during the first 12 months from the end of the treatment, whereas 47 relapsed after 1 year from CR. Of those 87 relapsed patients, 49 achieved a second long-lasting CR after salvage therapy (median DFS, 6.0 years), 16 entered CR but subsequently again relapsed (median DFS, 3.0 years), and remaining 22 either did not respond or only partially responded to the treatment (median DFS, 1.5 years).

Seventy (21.4%) patients did not achieve CR after the first-line treatment (≤PR, i.e., patients with only partial remission,no response, or even progression of the disease during therapy); in 36 of them (52.1%) a CR after second-line therapy was obtained.

The mean time of survival in the whole examined group of HD patients was as follows: for OS, 8.5 years; for DFS, 7.0 years (Fig. 1). The longest time of observation was 22.5 years. At the time of submitting the manuscript (in July 1997), 244 patients were alive and free of disease, 58 died or were lost from observation, and 25 had the symptoms of an active disease and were receiving treatment. In patients who achieved continuous CR after the first-line treatment (CR group), the time of survival, both OS and DFS, was significantly longer than in ≤PR(P < 0.001).

The real prognostic significance of traditionally used clinical,morphological and biochemical parameters, which were often of value in univariate analysis (Tables 1 and 2), was verified by multivariate regression tests. Only 6 of 32 factors analyzed showed their independent statistical significance. Table 5 shows the data from regression analysis. Pretreatment factors adversely correlating with OS were as follows: the presence of systemic symptoms (P = 0.002),poor clinical status according to Karnofsky’s scale (score < 70; P = 0.003), age at diagnosis of >45 years(P = 0.005), advanced (III/IV) clinical stage(P = 0.007), multisite nodal localization of disease(more than three sites; P = 0.01), and huge primary tumor mass (bulky disease; P = 0.03).

Parameters influencing DFS in multivariate regression analysis were as follows: age at diagnosis of >45 years (P = 0.001),Karnofsky’s scale (score < 70; P = 0.002), the presence of B-symptoms (P = 0.003), advanced clinical stage (P = 0.003), multisite nodal localization of disease (P = 0.01), and bulky disease(P = 0.02).

The following parameters showed their important, statistically significant value for the prediction of poor response to the first-line treatment (≤PR patients): bulky disease (P = 0.02),advanced clinical stage (III/IV; P = 0.03),Karnofsky’s scale score <70 (P = 0.03), and age at diagnosis of >45 years (P = 0.04; Table 5).

Pretreatment serum LDH activity in the ≤PR ranged from 143 to 592 IU(mean, 323.4 ± 110.1 IU; normal range, 120–230 IU). These results were significantly higher than in CR patients, in whom the level of LDH was 110–458 IU (mean, 204.6 ± 73.8 IU; P = 0.02). However, serum LDH levels did not show statistical significance as an independent prognostic factor. It did not influence either the response to the treatment in the logistic multiple regression test or the time of survival in the Cox regression analysis.

Other morphological and biochemical parameters that were analyzed in this study (Table 2), such as pretreatment ESR; hemoglobin level;absolute number of leukocytes, lymphocytes, monocytes, eosinocytes,erythrocytes, and platelets; alkaline phosphatase activity; serum iron levels; the concentration of serum proteins in electrophoresis; and renal and hepatic sufficiency tests, did not influence the survival time or the response to treatment in multivariate analysis.

Clinical analysis enabled us to detect an additional predictive factor, i.e., the intensity of treatment. The DI of COPP or MOPP,and ABVD regimens is shown in Table 6. The majority of therapeutic agents in these regimens were applied in satisfactory dosages, comparable to those of the original protocols. However, sometimes it was necessary to modify the drug dosages(especially adriamycin) due to age, performance status, side effects of chemotherapy, or lack of cooperation with the physician. For these reasons, in the whole analyzed group, 49 (15.0%) patients did not receive adequate therapy (DI ratio under 70% of planned total doses). The differences in the time of survival between those patients and the rest of the group were statistically significant (for OS, P = 0.009; for DFS, P = 0.02).

Doses of mantle radiotherapy were also modified in individual cases. In 21 of 327 (6.4%) patients analyzed, the total dose of radiotherapy applied was less than 70% of the planned dose. Among them there were 12 patients in stage I/II and 9 in stage III. Furthermore, in the other 12 (3.7%) patients (5 in stage I/II and 7 in stage III), due radiotherapy was not performed because of their resignation from this kind of treatment. All of these 12 patients achieved CR after an additional 2–4 cycles of standard chemotherapy, but 10 of them(83.3%) subsequently relapsed, with DFS from 1.5 to 8 years. The differences in the time of survival between those patients and the patients treated with the full planned doses of radiotherapy were statistically significant (for OS, P = 0.03; for DFS, P = 0.04).

Because both a low DI of cytostatics and a lack of or too-low dose of radiotherapy influenced the time of survival and response to the first-line treatment in similar statistical strength, we have compared all those cases in the same NAIT group. The total number of NAIT patients was 68, because in 14 cases, both radiotherapy and chemotherapy doses were reduced. The cut-points (for both 70% of planned doses) were established on the basis of DI values distribution in histogram (not shown). The percentage of CR achieved after the first-line therapy in this NAIT patients was 57.4, which was significantly lower than in patients with an adequately intensive treatment (84.2%; P = 0.01). NAIT correlated negatively also with the time of survival (for OS, P =0.02; for DFS, P = 0.03; Fig. 2).

Additionally, short period of time till CR was achieved also appeared to be a factor significantly influencing the time of survival. In HD patients in whom CR was obtained after no more than four cycles of standard chemotherapy, the risk of subsequent relapse was 21.0%, with a mean OS of 12.0 years and DFS of 9.5 years. In contrast, in patients in whom CR was obtained after more cycles of chemotherapy or after complementary radiotherapy, the frequency of relapses was higher,amounting to 39.7%, with a mean OS of 8.0 years and DFS of 6.0 years. The differences in both the risk of relapse and the time of survival in those groups of patients were statistically significant (for the risk of relapse, P = 0.02; for OS, P = 0.01;for DFS, P = 0.009; Fig. 3).

Table 7 summarizes the results of PCNA,p53, and BCL-2 expression on R-S/H cells obtained in the whole group of patients. Positive PCNA reaction was observed on R-S/H cells in 138(71.1%) cases, with the proliferate index in whole group ranging from 0 to 95% (mean, 51.5%). Tumor cells were p53-specifically labeled in 112 HD patients (57.7% %), with the index ranging from 0 to 90% (mean, 26.2%) in the whole group. BCL-2 protein expression was demonstrated in the cytoplasm of R-S/H cells in 91 (46.9%) cases; the index in the whole group ranged from 0 to 60% (mean, 11.0%). The coexpression of two antigens was observed in 37 patients (19.1%), and coexpression of all three antigens was observed in 59 patients(30.4%).

The expression of PCNA, p53, and BCL-2 on R-S/H cells correlated statistically significantly with the time of survival (Figs. 4,5,6). Significant negative correlation was found to occur between the indexes of PCNA and both OS (P = 0.02) and DFS(P = 0.01). Much stronger correlation with survival time was recorded for p53 (OS, P = 0.001; DFS, P = 0.0002) and BCL-2 expression (OS, P = 0.009; DFS, P = 0.002).

High expression of PCNA, p53, and bcl-2 statistically influenced poor response to the treatment. In CR patients, the expression of PCNA was 0–80 (mean, 46.3%) of the R-S/H cells. In this group the index for p53 was 0–60 of the R-S/H cells (mean, 18.4%) and the index for BCL-2 was 0–40 (mean, 4.5%; Table 6). In ≤PR indexes of PCNA, p53 and BCL-2 were much higher than in CR patients and were 10–95%(mean, 78.2%), 20–90% (mean, 49.6%), and 0–60% (mean, 20.5%),respectively (Table 7).

The best prognostic pretreatment pattern was lack of the estimated protein expression, whereas the worst prognosis involved patients with the coexpression of PCNA, p53, and BCL-2 on R-S/H cells (Fig. 7).

The final stage of multivariate regression analysis was applied to summarize all of the known and considered predictive factors, including our new proposition (oncoprotein expression index), for evaluation of their independent prognostic significance. The final results after all steps of the forward stepwise multivariate analysis are as shown in Table 8.

The only independent prognostic factors that negatively influenced OS were advanced (III/IV) clinical stage (P = 0.0001),high expression of p53 (0.003), poor performance status (Karnofsky’s scale < 70; P = 0.01), and age at diagnosis of>45 years (P = 0.02). Prognostic parameters influencing DFS were advanced clinical stage at diagnosis(P = 0.00007), high expression of p53 (0.001),Karnofsky’s scale < 70 (P = 0.004), age at diagnosis of >45 years (P = 0.01), and high BCL-2 expression on R-S/H cells (0.02).

Among all of the pretreatment factors analyzed in regard to poor response to the first-line treatment (≤PR) only four factors had statistically independent, predictive strength: high p53 index(P = 0.0007), poor performance status (Karnofsky’s scale < 70; P = 0.009), multisite nodal localization of disease (more than three sites; P =0.03), and high PCNA index (P = 0.04; Table 8).

There are many clinical, morphological, and biochemical parameters considered to be important for the prognosis in HD. However, the application of statistical multivariate analysis has dramatically reduced the number of factors, separately evaluated as important, to only those retaining a true independent prognostic role (9, 10, 11). In this study, we have proven that six independent clinical pretreatment factors are negatively correlated with survival time or response to the first-line treatment. The predictive value of factors commonly considered important for the prognosis in HD, such as sex, ESR, number of leukocytes, or hemoglobin level determined at diagnosis, has not been confirmed. Our observations are consistent with the results of some other authors, but the outcomes from different studies are still controversial (12, 13, 14, 15). In any case, the majority of authors emphasize the strong negative prognostic value of large tumor mass at the diagnosis. After using a univariate and multivariate statistical analysis, most of the hitherto known prognostic factors appeared to correlate with the total tumor burden, and they lack independent prognostic significance (16).

In the last several years, scientists have been searching for new, more specific and sensitive prognostic factors. For example,aneuploidy and S-phase fraction are well-known prognostic factors in solid tumors and non-Hodgkin’s lymphomas. Erdkamp et al.(17), analyzing these parameters in relation to clinical characteristics of HD patients, concluded that the S-phase fraction could be a good indicator in cases in which the outcomes may be worse. However, DNA aneuploidy did not correlate either with time of survival or with other known prognostic factors.

A number of other laboratory parameters, such as serumβ-2-microglobulin level (18), interleukin-2 (19), CD-8 antigen (20), or urinary excretion of pseudouridine (21), have been also considered as valuable for the prognosis in HD. According to some authors, one of the most important prognostic factors in HD is histological type of the disease (22). Additionally, Alavaikko et al.(23) suggested that the presence of follicular dendritic cells in neoplastic areas of lymph nodes may predict a favorable outcome in the HD patients.

In non-Hodgkin’s lymphoma patients, as well as in pediatric HD, a valuable prognostic factor seems to be the pretreatment serum LDH activity (24, 25). In our study, high LDH activity significantly correlated with worse response to treatment, and it was the only biochemical parameter investigated that appeared to have an independent prognostic importance.

Abnormal expression of p53 and BCL-2 oncoproteins on neoplastic cells has been recently shown in carcinomas, sarcomas, and lymphoid neoplasms. The expression of PCNA, a marker of cell proliferation, can also correlate with the progression of some of human malignancies. However, data from the studies concerning HD are thus far scarce and controversial.

Lack of p53 expression on R-S/H cells in the lymphocyte-predominant type of HD and positive staining in the remaining histopathological types of the disease were reported by Lauritzen et al.(26). Overexpression of the BCL-2 was demonstrated in a wide range of HD cases, and it can be speculated that this abnormality may play a role in the pathogenesis of HD (5). PCNA expression without any statistical significant differences between the different types of HD was found by Hell et al.(27) and Kordek et al. (28).

Therefore, thus far, only Xerri et al.(29) have investigated the correlation between the oncoprotein expression in R-S/H cells and some clinical parameters in HD, such as clinical staging, B-symptoms, probability of relapse, and DFS. The authors were not able to find any statistical correlation between the p53 expression on R-S/H cells and these parameters. Additionally, Trümper et al.(30) found that the activity of soluble p53 in serum of 21 of 33 HD patients bears no statistical correlation with other clinical prognostic factors.

Our immunohistochemical examinations of PCNA and the protein products of the p53 and bcl-2 genes, performed in the lymph nodes obtained from patients with HD, showed that the high expression of all these antigens correlated with poor response to the treatment and short time of survival. The lowest pretreatment indexes of PCNA, p53, and BCL-2 were found in the patients who achieved CR and were alive and free from disease during the whole period of our observation. Statistical analysis of survival time led us to the conclusion that PCNA index, as well as the expression of the p53 and BCL-2 proteins on R-S/H cells, can be taken into consideration as a new prognostic factor in HD.

Treatment modality is an important factor influencing the rate of relapses (1). Simultaneously, we found that the early CR,achieved by the fourth cycle of chemotherapy, can be used as an independent positive prognostic factor. Similar results were reported by Somers et al. (31). An inadequate treatment(usually due to lack of radiation therapy) can be also a cause of failure in the treatment in a number of patients with HD.

The data on the real predictive value of many factors may not be fully consistent, mainly because the number of examined patients is still too small and because of differences in statistical tests used for the analysis. Therefore, the standardization of methods and multicenter collaborative studies on large groups of patients (2, 3, 31, 32) is necessary to establish the set of factors with a real prognostic value, with the view of selecting the most adequate therapeutic modality in HD.