Document a1gDnxbErap1wy7GEg5vgoJxX

British Journalof Haerna[o/q!j.1992. 80. 5 5-6 1 Impact of active and passive exclusions on the results of a clinical trial in multiple myeloma M A R T I N H J 0 R T H . I E R I K H O L M B E R G . Z S T I C RODJER' A N D J A N W E S T I N 3 FOR T H E M Y E L O M A GROLJPOF WESTERN S W E D E N * I Depnrtment of Medicine. Lidkoping Hospital. 'Oncology Center and 3Hematology Section. Department of .IIedicinP 11. Sahlyrcnska Hospital. and 4Departnicnt o j Medicine. Ostra Hospital. University OJ Ciiteborg. Swderi Received 10 ]utle 1992: ncceptcdjor publication 8 September 1991 ~~~ Yummary. During the 3 years 1984-86. 314 cases of multiple myeloma were diagnosed in the Health Care Region )f Western Sweden. 180 of these cases were included in a Ainical trial: 71 were notified to the trial but excluded: 49 :ases were not reported to the trial; 14 were diagnosed post 'nortern. . The crude incidence rate of myeloma was 6.3 cases per 100000 inhabitants per year. corresponding to an agerdiusted (world standard population) incidence rate of 2.9 ases per 100000 inhabitants per year. The excluded and the non-reported patients had a signifim t l y shorter survival than those included in the clinical rial (median survival 2 2 . 1 3 and 3 3 months, respectively). h i s was partly due to differences in age and proportion of Ictively treeated patients between the groups, but the same tendency remained also after correction for these factors. Considering the included patients separately. the effect of tentative application of presumptive exclusion criteria corresponding to major prognostic factors was studied. Prolonged survival was seen when the upper age limit was lowered and when patients with renal failure or low performance status were excluded. The results illustrate the fact that for multiple myeloma the survival in a trial population is markedly influenced by active and passive exclusion of patient groups with unfavourable prognostic characteristics. When reporting results of clinical trials. discussion of the representativeness of the trial population for the total patient population is recommended in order to facilitate application of the trial results to medical practice and comparisons between trials. byhen results from clinical trials in multiple myeloma are to he transferred to clinical practice one has to consider some ' x t s t h a t make the results difficult to interpret. Treatment 'esdts differ considerably between different clinical trials but .Omparisons are difficult to make because the patient *lPulations often differ from each other with regard to 'atient characteristics of prognostic importance. e.g. age. 'The followinginvestigators and centresof the Myeloma Groupof . Sweden collaborated in this study: Bjom Fagerberg. Bengt IaRnusson and Jan Westin. Sahlgrenska.Hospital. and Stig Riidier. Hospital. Goteborg: Mats Willstrand. AlingsBs: Carl Magnus iokBorAs: Jan knnartsson. Eckefors: Rolf Svensson. Fakoping: Hellqvist. Hdmstad:knnart Langer. Kungalv: Martin Hjorth. 'dkOPing:Peter Hoffman.Lysekil:Soren Larsson. Mariestad: Sigvard "'ison.Molndal: h e Rindner. Skene: Ulla Evaldsson. Skovde: Kurt "s-n. Striimstad:Bo Grande.Trollhattan:OlofLindqvist.Uddevalla: ' m s bnnart Blom, Varberg: Maiu Sarg. Vanersborg. This study .ceivedfinancialsupport from the Oncology Center, Goteborg. '"Wmdence: Dr Martin Hjorth. Department of Medicine. hdkop'I: tiosPital. S-53185 Lidkoping. Sweden. stage distribution and performance status. This problem has recently been discussed by Paaske Hansen L% Galton (198 5 ) and by Kelly et al(1988).Furthermore, hardly any reports on clinical trials discuss the representativeness of the study population in relation to the total population of myeloma patients. In some studies the material contains the majority of myeloma patients with the diagnosis confirmed during the study period, whereas in other studies the material is highly selected due to different active and passive exclusion mechanisms. Knowledge of these factors may be of great importance when results of clinical trials are to be applied in clinical practice. The purpose of this study is to describe a total population of myeloma patients diagnosed during 3 years within a defined geographical area and to define the relationship between a study material recruited during the same period from the same population for a clinical trial and the total patient population. The total population is analysed and compared with the trial population with respect to prognostic factors and survival. The fate of the non-study patients is described 55 56 Martin Hjorth et a1 and the effect on survival of different plausible exclusion criteria tentatively applied to the trial population is analysed. Table 1. Revised diagnoses of 346 cases notified as myeloma to the Swedish Cancer Register MATERIAL Multiple myeloma. reported to the clinical trial 237 ! During the period 1 January 1984 to 31 December 1986 a Multiple myeloma. not reported to the clinical trial Other plasma cell disorders S9 randomized clinical trial on myeloma was performed in Solitary plasmocytoma 18 Western Sweden (Hjorth et a/. 1990). Two university Primary amyloidosis 2 hospitals and 16 local hospitals, covering the entire Health Care Region of Western Sweden (1.65million inhabitants), co-operated in the study. The participating centres were Immunoblastic lymphoma Monoclonal gammopathy of unknown significance No plasma cell disorder 1 24 5 requested to report all newly diagnosed cases of myeloma, including patients not eligible for the trial. During the study period, a total of 2 5 5 patients were reported. 180 patients were included in the trial. 75 were excluded. The reasons for exclusion were high age (19). heart disease precluding anthracycline chemotherapy ( 12). other severe concurrent illness ( 12). unwillingness to participate ( 151, renal failure necessitating initial haemodialysis (4). non-secreting myeloma (2), plasma cell leukaemia (1).multiple myeloma complicated by a large plasmocytoma necessitating irradiation ( 1 ). terminal state ( 1), administrative reasons (4) and diagnosis post mortem (4). In order to find patients not reported to the study. the Regional Swedish Cancer Register was surveyed for all and the non-reported patients. cytostatic therapy, when given, usually consisted of traditional MP therapy. Follow-up. All patients were followed up until death or until November 1989. with a median follow-uptimeof 53 months. Statistical methods. Comparisons between the group of included patients and the other groups were performed using i 1Pitman's test (Bradley. 1968). When studying the relation- ship between presumptive prognostic variables. a special survival test was applied in the univariate case (Merck et d, 198 3 ) and Cox's regression model in the multivariate case. The survival curves were constructed by the lifetable method and compared by the log-rank test (Armitage Br Berry. 1987). notified cases of myeloma from the same catchment area. diagnosed during the study period. For cases notified to the cancer register but not reported to the study. the medical records were reviewed considering criteria for diagnosis, treatment and survival. which is a conventional method when studying the relationship between a zero-one variable and survival. The surviv& test (Merck rt al. 198 3 ) extends the applicability to general variables (e.g. serum calcium) avoiding the idealizing assumptions made by the Cox's model. Two-sided tests were used. $1 ETH0DS The eflects o j plausible exclusion criteria were calculated as follows: For age. upper age limits with 3-year intervals from Diagnostic criteria. In the clinical trial the following criteria for the diagnosis of multiple myeloma were used: ( A ) Serum %I-proteinconcentration > 30 g/l (IgC) or >10 g/l (IgA) and,or Bence Jonesproteinuria > 1 g/24 h. t B) bone marrow plasma cells > lo?/,. and IC) osteolytic bone lesions. A +diagnosis of multiple myeloma was accepted if criteria A B or A +C were fulfilled. In the evaluation of patients identified by the survey of the Swedish Cancer Register. in order to include all symptomatic multiple myeloma patients. the somewhat broader Southwesi Oncology Group (SWOG) diagnostic criteria tDurie & Salmon. 19i7)were used. Irrcidence rate cdcubtion. To calculate age-standardized 85 to 65 years were applied to the whole stage I1 and In patient group and to the included stage 11 and I11 patients separately. For each age limit. the median age of the remaining patients was determined and the median survival estimated from the survival curve. Regarding renal failure. patient groups with serum creatinine values > 500. > 300. >LOO. > 1 5 0 and > 120 pnol/l were excluded consecutively and the survival of the remaining patients determined. In a similar manner, patients with a /ow prrfurrriurx~estatus. defined as Karnofsky index <200/, and 230%. were excluded and the survival of the rernaming patients determined. incidence rates the 'direct' method was used with the world population as standard (Waterhouse rt a/. 1982). The age- RESULTS specific incidence rate was calculated for each 5-year inter- Survey o j the Swedish Cuncrr Rtgistrr val. During the study period 346 cases of myeloma from the Treatment regrrriens. The design of the clinical trial is Health Care Region of Western Sweden were notified to the described in detail elsewhere (Hjorth et al. 1990). Briefly. cancer register (Table I). out of which 2 3 ; cases were also stage I patients were randomized between traditional inter- reported to the clinical trial: 109 cases were not reported to mittent melphalan and prednisone therapy (MPt and initial the trial. Out of these, the diagnosis of multiple myeloma observation without therapy: when progression occurred, according to SWOG criteria was confirmed in 59 cases, 46 of MP was started. Stage I1 patients were randomized between which also fufilled the diagnostic criteria of the clinical trial. M p and vincristine. melphalan. cyclophosphamide and pred- Ten of these 59 cases were diagnosed post mortem. The nisone (VMCP)every 4 weeks. and stage 111patients between diagnoses of the 50 cases that were not considered to be MP and vincristine, carmustine. doxorubicin and prednisone confirmed myeloma in the present re-evaluation are shown (VBAP)a n d VMCP alternately every 4 weeks. In the excluded in Table I. Active and Passive Exclusions in Myeloma Trials 5 7 - Males -0- Females A 3 -Ie If 35- 40. 45- 50- 55- 6& 65- 70- 75- 80- 85+ 39 44 49 54 59 64 69 74 79 84 Age iniervai Fig 1. Age and sex-specitic inc,idence rates per 100000 inhabitants per year for multiple myeloma in \Vestern Sueden from 1984 to 1986. 1rr:idence qf nigelottrci Thus, during the study period of 3 years a total of 314 confirmed cases of multiple myeloma were diagnosed in Western Sweden,with a population comprising 1.65 million Inhabitants.This yields a crude incidence rate of 6.3cases per 100000 inhabitants per year, which corresponds to a world qe-standardized incidence rate of 2.9 cases per 1 0 0 0 0 0 inhabitants per year. For men. the crude annual incidence rate was 6.8 and the age-standardized incidence rate 3.6 cases per 100000 inhabitants per year. For women. the c'omponding figures were 5.9 for crude incidence rate and 2 . 3 for age-standardized incidence rate. The age-specific Incidence rates increased with age as shown in Fig 1. I'utient groups The 14 cases of myeloma that were diagnosed post mortem dre included in the incidence calculations but omitted from dl other analyses in this study. The material used for all cinalyses except the incidence calculations consists of three Rroups of patients: ( 1 ) 180 patients reported to and included In the clinical trial. These patients are henceforth referred to 'mcluded'. ( 2 ) 71 patients reported to the study but not c'llgible for the clinical trial. These patients are referred to as rrcluded'. ( 3 ) 49 patients not reported to the study. These Patients are referred to as 'non-reported'. IJPtJentcharacteristics ('hical and laboratory characteristics for the different Patient groups are shown in Table 11. As expected. age was '%her in the excluded group and also in the non-reported ErouP compared to the included group. At age intervals the median age of the total material. an increasing proportion of patients were excluded from the trial or not "ported (Fig 2). Haemoglobin was significantly lower in the " " n - W ~ r t e d group than in the included group ( P < O - 0 1). I 'Ir other patient characteristics of major prognostic impor- tance- with reservation for the possible omission of some clinically important difference. due to the rather small number of patients-no statistically significant differences were seen between the patient groups. Fa@of the included patients The results of the randomized trial comparing MP with multidrug chemotherapy for stage I1 and 111 patients are reported elsewhere (Hjorth et af. 1990). In summary. no differences were found in response rate or survival between the treatment arms. Considering the two arms taken together. the response rate as defined by a decrease of the 11protein concentration by > 50(;;, was 63% for stage I1 patients and 58%for stage 111 patients. The median survival was 36 months for stage If and 2 5 months for stage Ill patients. For stage I patients. randomized between MP and initial observation. iriterim results have been reported (Hjorth et a]. 1989).The median survival is not yet reached: for the two arms taken together. the survival at 48 months is 60(%. Fntc qf tlir exduded patients In this group ( n= 7 1 ) 60 patients I 8 5%) were treated with chemotherapy. usually MP. 28 patients (47'%,1responded with a decrease of 34-protein by at least 50%. The median survival was 22 months for the whole group (Fig 3) and 26 months for the actively treated patients. Fate of the non-reported patients This group was heterogeneous and comprised three categories of patients: ( A ) 1 1 patients with definite multiple myeloma according to SWOG criteria but not fulfilling the formal diagnostic criteria used in the trial. (B) 24 patients that. because of high age ( n = 8 )and/or other coincident illness ( n= 12) and/or terminal state of myeloma disease ( n=8\ or other reasons ( n= 2 j. presumably were never considered for the clinical trial. and IC, 14 patients that were eligible for the trial but n e v c notified. Out of the whole group. 34 patients (69%)were treated with chemotherapy. Considering the three patient categories separately. the number of treated patients was 7/11, 13/24 and 14/14. respectively. A response. as defined above, was noted in 15/34 treated patients (44%).The median survival was 1 3 months for the whole group (Fig 3) and 17 months for the actively treated patients. Prognostic factors Prognostic factors were analysed for the included stage I1 and III patients only (n=149) (Table 111). Univariate analysis showed prognostic importance for age, serum creatinine, haemoglobin, platelet count, serum albumin, corrected serum calcium, performance status and light chain disease. Serum /?Z-microglobulin was not analysed. The multivariate analysis, however, showed independent prognostic importance only for age. serum creatinine. performance status and light chain disease. Survival The median survival for the total material was 26 months. Considering the three patient groups separately (Fig 3). the 58 Martin Hjorth et a1 Table11. Patient characteristics. Clinicaland laboratory Endingsin included,excludedand non-reported patients Included Excluded Not reported All patients No. of patients 180 71 49 Age (medianI 69 76 79 (range) 41-91 44-90 51-89 Sex ratio (M/F) 1.1 0.9 1.6 Stage according to h i e & Salmon I 31 (17)t 15 (21) lO(20) I1 52 (29) 16 (23) 10 (20) rIr 97 (54) 40 (56) 29 (59) Serum creatinine 200 pmol/l 33 (18) 13 (18) 7(14) Haemoglobin <85 g/l 23 (13) 12(17) 9(19) Platelet count < 100x 109/l 8 (4) 3 (4) 2 (5) Serum albumin < 30 g/l 37 (21) 20 (29) 13 (33) Corrected S-calcium' > 2.60 mmol/l 68 (38) 25 (36) 18 (44) M-protein class 110 (61) 33 (47) 31 (63) IgA IgD IgE.IgM or not analysed 37(21) 22(31) 40 11 6(12) 0 1 Bence Jonesonly 27 ( 1 5 ) 12 (17) 10 (201 Non-secreting myeloma 12 2 300 72 41-91 1.1 56 (19) 78 (26) 166 (55) 53 (18) 44 (15) 1 3 (4) 70 (24) 111 (38) 174 (58) 66 (22) 4 3 48 (16) 5 * Corrected S-calciurn=Serum calcium -0.02 (Serumalbumin -40). $ Numbers in parentheses are percentages. 1.o 4 u so i so ssw70 7s . w . 8s . 90 os Fig 2. Agedistribution of 300cases of multip!emyeloma diagnosed in Western Sweden from 1984 to 1986. Included=lncluded in the clinical trial ~n= 180). Excluded= Sotitied but excluded from the clinicaltrial ( n= i l l . Non-reported=Sot reported to the clinical trial In = 49 I. median survival was 3 3 months for the included. 2 2 months for the excluded and 13 months for the non-reported group. The differences between the three survival curves were statistically significant ( P < O . 0 5 ) . After correction for the age differences between the patient groups. there was still a significant difference between the included and the nonreported patients. When a similar comparison was restricted to patients treated with chemotherapy. there was still found a significant difference between the survival curves for included and non-reported patients. also after correction for age (P<O.O5). 1 0.8 w 5 0.8 c -0 6 i 0*4 9 0.2 0.0 0 6 12 18 24 30 36 42 48 Months Fig 3. Probability of survival for the complete patient groups. The differencebetween the survival curves was statistically significant for included versus excluded I P <0.05L included versus non-reported ( P<0.001 and excluded versus non-reported patients IP <0.05). Survival and its dependence on age To study the dependence of survival on age distribution, the patient material was reduced by tentative application of a series of upper age limits with 3-year intervals from 8 5to6 5 years (Fig 4). Considering the included stage II and lll patients first. the number of remaining patients was now consecutively reduced from 149 to 61 and there was a concomitant progressive decrease of the median age of the remaining patients from 68 to 60years. The effect on survival was a corresponding prolongation of median survival from ' 30 to 35 months. Active and Piissive Exclusiuiis in Myelorria Trials Table Ill.The prognostic effect of some variables estimated by multivariate Cox analysis 59 Variable Age lyears i Serum creatinine Ipmolll) Haemoglobin Ig/l) Platelet count I x 1 0 " 1) Serum alhuniin ( g 11 Performance status' ",I hght chain diseaxf Mean=kSD Death risk ratio for increase of the variable by 1 SD Confidence interval for death risk ratio Karnofsky index. ;For light chain disease death risk ratio was calculated for the 17 patients with only Bence Jonesproteinuria versus the 1 5 3 patients with other Ig classes. Upper age limit -- All stage 11-111 patients (nd44) Includedstage ll-Ill patients (n=149) 24 26 28 30 32 34 36 Modhn survival (months) IR 4. Ukct on survival of Merent upper age limits in a population of multiple myeloma patients. The median age of the remaining patients in `ach group is given in parentheses. Table IV.meet on survival of the exclusion of patients with renal failure upper limit for Serum creatinine (rrnol/l) NO limit 500 300 `on) 150 120 No.of patients 149 139 I31 Ilh 1Oh 87 Median survival (months) 30 31 32 33 34 36 Secondly, for the sake ofcomparison. a similar calculation its made for all stage 11 and 111 patients (n=244). For the ' holeof this group without any upper age limit. the median Table V. Effect on survival of the exclusion of patients with low performance status Exclusion of patients with Karnofsky index Q 2% Q 3% No. of patients 149 140 122 Median survival (months) 30 31 33 age was 72 years and the median survival 24 months. When the group was reduced by application of upper age limits decreasing from 85 to 65 years, the number of patients was successively reduced to 75 with a concomitant lowering of the median age of remaining patients to 61 years and a corresponding prolongation of median survival to 33 months (Fig 4). 6 0 Martin Hjorth et al Sitrvivul unci its dependence on exclusion of patients with renal /&lure [n a similar manner. the effect on survival for the included stage I1 and I11 patients was studied by applying different upper limits for serum creatinine. As a result. the median survival in the material was substantially prolonged when patients with high serum creatinine values were excluded (Table [V). For instance. if patients with serum creatinine values >200 pnol/l In = 331 were excluded the median survival was prolonged from 30 to 3 3 months. Survival and its dependence on perjorniance status The exclusion of patients with low performance status was shown to bring about a similar effect on survival (Table V). For instance. if patients with a Karnofsky index < 30% were excluded from the analysis. the median survival was prolonged from 30 to 3 3 months. DISCUSSION Multiple myeloma is a disease which primarily affects older patients. In Sweden. the crude incidence of multiple myeloma is high compared to many other countries because of a high and steadily increasing proportion of old patients in the population. But the age-adjusted incidence rate is also high compared to other recent reports (Lines t't rtl. 1981:Turesson vt al. 1984: Shapira 8;Carter. 1986:Nandakurnarrl id. 1988: Karjalainen Clr Palva. 1989: Hansen rt ul. 1989: Cadi et ul. 19891. The high incidence tigures in this study do not necessarily reflect a higher real incidence than in other countries but rather a high diagnostic intensity combined with a high notitication rate. resulting from the requirement of notitication of all newlv diagnosed cases both to the ongoing clinical trial and to the Swedish Cancer Register. The Swedish Cancer Register is based on the notification duty for both clinicians and pathology cytology laboratories concerning all newly diagnosed malignant and certain benign tumours. The reliability 0 1 the Kegister has been evaluated by blattsson ( 1984I . who. by comparison of the notitications in the Cmcer Register during 1978 with cancer diagnoses notitied to the Cause-of-Death Register and the InPatient-Care Register of Stockholm County. found a registration deticit for multiple myeloma of S.8':{, for Stockholm County and 16.3'%, for the whole of Sweden. In the present study we found a registration deticit in the Cancer Register of 4.5','(,.On the other hand. the review o f the medical records of patitmts notitied t o thc L'anccr Kegister but not to the trial revealed an over-notitil.ation in the Cancer Register by 1h'lh. partly due to co-registration of other malignant plasma cell disorders. mostly solitary plasmocytomas. partly to incorrect notitication of nonmalignant plasma cell disorders. mostly MGCS. Thus, according to this study, the figures of the Cancer Register give in an overestimation of the real incidence of multiple myeloma by about 10%. Prognostic factors in multiple myeloma have been analysed and reported from several studies. Recent reviews have been published by Paaske Hansen & Calton ( 1983 ) and Kelly et ul I 1988,. In the present study. old age. impaired renal function. performance status and !igh: chain diseEse -:ere found to have independent negative effects on survival, Among prognostic factors. high age. renal failure and low performance status constitute patient characteristics that are liable to cause passive exclusions from a clinical trial. with a potential corresponding impact on the results of the trial (see below I. When the results from a clinical trial are to be applied in clinical practice. it is necessary to know whether the s t u d 4 patient population is representative ofthe majority ofpatienb with the disease or not. In a disease like multiple m y e l o w with a steeply increasing age-related incidence. this point ha special significance because clinical trials are mostly c a m out on patients of considerably lower ages than the majoh9 of patients in clinical practice. This issue is addressed bg Cohen & Bartolucci (1985). who. based upon experience from a clinical trial performed in 1972-76 by the Southe& tern Cancer Study Group. concluded that older patients the same responses and survival rates as younger patients. that study. 2 1%of the patients were 2 70 years. cornparedto 47% projected for the total multiple myeloma patient popula- tion. The corresponding figures for the present study are jo% and 61%. In the present study. the matter of representativeness is illustrated in two ways. Firstly. the population ofpatients that were not included in the clinical trial was identified a d studied regarding prognostic factors. treatment and survival. Secondly. the impact on survival of excluding patient group with poor prognostic characteristics from the trial mated4 was analysed. In this study. 6004of all cases diagnosed ante mortem were included in the clinical trial. Out of the remaining patients. 24% were actively excluded from the trial. 1h',Y,were nevu reported to the trial and thus passively excluded. majority of these patients. if notitied. would also have beep actively excluded from the trial because of high age and otha coincident diseases and at most 5% of the total patient population could possibly have been Lidded to the trial. Thus. the representatives of this trial population was limited to patients that were under the age o f 8 5 years and free from other severe diseases. The effect of other exclusion criteria on the results of the trial is illustrated by the calculations of survival after the application of different plausible exclusion criteria to the trial material. The results show chat the exclusion of higher age groups. patients ivith renal failure and patients with low performance status has a considerable etlect on the survival figures. in ii clinical trial. exclusion of patients with these characteristics may be .ictive. according to the exclusion criteria in the study protocol. or passive. because of r e f e d patterns and participating clinicians' hesitance to consida certain categories of patients for randomized trials. Publica tions on clinical trials seldom contain information regardins these exclusion possibilities. A management trial. trying as far as possible to imita& day-to-day clinical practice. should ideally include a & population that mirrors the total patient population. Thisbe achieved by including all cases in a given population o r b including a randomly selected group of patients. In haemat@ logy, the latter is generally not feasible because of i n s a c i d 1. Active and Passive Exclusions in Myeloma Trials h 1 number of available cases. Therefore. when management trials are planned. every effort should be made to include as many actively treatable patients as possible out of the given population. But even with a masimal effort. a considerable proportion of patients will not be included in the trial. Knowledge of the differences between included and nonincluded patients is important-not for the interpretation of the results of a comparison between treatment arms in a randomized study. but when the results of the trial are to be applied to the whole patient population and when comparisons with other trials are made. In conclusion. this study illustrates how the survival of a study patient populatiLm is influenced by the inclusion and!or rsclusion of patient groups with unfavourable prognostic characteristics. When the results of a clinical trial is to be transferred to clinical practice it is necessary to know how the patient population of the trial has been selected from the totai patient population. N'hen a clinical trial is reported. a discussion regarding the representativeness of the study Population is recommended. \CKNOWLEDGUENTS \'.e wish to express our gratitude to h d e r s Oden for statistical advice. Susanne .\msler-Nordin for secretarial mistance. and the members of the Myeloma Group of estern Sweden for rewarding collaboration. REFERENCES lrmitage. p. & Bern-. c;. i 19x7) Stnriarirul .Mrtlimk in Medirul Hrsearch.2nd edn. pp. 419-43 3. Blackwell ScientificPublications. Oxford. hdley. J.W. ( 19681 Distrihutiori,frre Statistical Trsts. pp. 68- 86. I'rentice-Hall. London. '.*irkP.M.. Bailly. F.. Tavernier. C.. Milan. C.. Hendes. 1).& Lavault. 1.F. ' 19891Multiple myeloma: epidemiological features in a well- kldined population in Burgundy. France. Irltrrriutror!nl /ournuf O/ bidrmiology. 18, 3 30-3 3 3. ('ohen. H.1. & Bartolucci. A. 11985)Age and the treatment of multiple myeloma. American journal ojMedicine. 79. 3 16-324. Ihlhe.B.C.M. &Salmon.S.E. ( 197i)Multiplemyeloma. macroglobulinaemia and monoclonal gammopathies. Recent Advances in Hwmatology (ed. by A. V. Hoffbrand. M.C. Brown and 1. Hirxh). PP. 243-261. Churchill Livingstone. Edinburgh. 'Ian-. %E.. Karle. H. & Olsen. 1.H. I 19891 Trends in the incidence or multiple myeloma in Denmark 1943-1982: a study of 5 5 0 0 patients. European /ournal o/ Haernnto1og.y. 42. 72-76. Hjorth. Mi..Hellquist. L.. Holmberg. E.. Magnusson. B..Riidjer. S. & Westin. J. I 19x91 Results of cytostatic therapy differentiated according to stage in an unselected series of multiple myeloma patients. (AbstractI. 5th European Conference on Clinical Oncology. London. Hjorth. %I..Hellquist. L.. Holmberg. E.. Magnusson. B.. RWjer. S. & Westin. 1. t 19901 Initial treatment in multiple myeloma: no advantage of multidrug chemotherapy otrer melphalan-prednisone. British /ourrid o,/ Hcrcwtutologq. 74. 1 X 5- I9 1. Karjalainen. S.6; Palva. I . I 1 9 X Y i 1)o treatment protocols improve end results: :I study of survival o f patients with multiple ni\:elomi! in Finland. Hritisli , t f r d i d/ournol. 299. 106% 1072. Kelly. K..9.. Duric. R. 6; SlacLcnnan. I.C. I I 9 S X I Prognostic factors and staging systems for multiple myeloma: comparisons between the Sledical Research C'ouncil Studies in the I'nited Kingdom and the Southwest Oncology Group Studies in the I.ni!ed S!ates. Hrtnakriqqic-uI Oncology. 6. 1 3 1- 140. Linos. .A,. Kyle. R.A.. O'Fallon. W.M. 6; Kurland. L.T. 198 11 Incidence and secular trend of multiple myeloma in Olmsted County. Uinnesota: 1965-77. /oitrrtnl of tfw .Sariorid C ' a r i c w Irrstrflrtc. Ihh. i 7 - 2 0 , .Mattsson.8.I 19841('anccr registration in Sweden. Thesis. Depart- ment o f Oncoiog> a n d C'ancrr tipidemiology. Kadiumhcmmet. Karolinska Hospital. Stockholm. Sweden. Merck. C.. Angervall. L.. Kindblom. L.G. 8: Oden. A. 119831 Myxofibrosarcoma. .A malignant soft tissue tunior of fibroblastichistiocytic origin. A clinicopathologicand prognostic study of 1 I O cases using multivariate analysis. Actu Patfiologica.i\.licrohiologic.n CI Imrnunologico Scandinnvica. Section A. 91. Suppl. 282. Nandakumar. A.. English. D.R.. Dougan. L.E. 8- Armstrong. B.E. ( 1 988 I Incidence and outcome of multiple myeloma in Western Australia. 1960 to 19x4. Australian and Srw Zvuland /ournrtl ffl Mrdicinr. 18. 7i4-779. Paaske Hansen. 0. & Galton. D.A.G. (19851 Classification and prognostic variables in myelomatosis. Scandinavian /ournu/ qf Harrnatologj. 1 3 5 . 1 0 - 19. Shapird. R. 6 Carter. A. 1 Y X h I llultiple myeloma in northern Israel. 19iO-1979. Ctrric-cr. 58. 206-209 Turesson. I.. Zettervall. 0.. Cuzick. J.. Waldenstrom. J.C.& Velez. R. (1984) Comparisons of trends in the incidence of multiple myeloma in Malmo. Sweden. and other countries. 1950-1979. New England Journal of Medicine. 310. 421-424. Waterhouse. I.. Shanmugaratnam. K.. Muir. C. & Powell. J. (eds) (1982) Cancer Zncidence in Five Continents. pp. 671-675. IARC ScientificPublications 42: IV.