Document K0LDJp6aq6O089Eko9qq0qnK

in Youth Predictive of Adult-Onse Leukemias : Brief Communication 1 233 A B S ~ ~ ~ c T - C o l l e gheealth records of 50,000 male alumni of University (Cambridge, Mass.) and the University of pmnsylvania (Philadelphia, Pa.) were examined for characteris* a t were predictive of risk of fatal lymphaden, blood, and ~ rtin cancers. Among these alumni, 45 men with Hodgkin's disease, 89 with non-Hodgkin's malignant lymphoma, 45 with malig- melanoma, 27 with lymphatic leukemia, 41 with myeloid wkemia, and 30 with other and unspecified leukemia died from dise?,'e in 1.71 million person-years of observation. Men v p e d :.;cording to the type of cancers they had, identified *omdeath certificates, were contrasted with four times as many ,,,niving classmates in terms of potential predictive characterbdcs. Relative risks of these six cancers may be grouped by prdictive characteristics in youth as follows: 1) History of com- contagious diseases of childhood was predictive of a lower rblr of Hodgkin's disease and lymphatic leukemia, but varicella Micated increased risk of other malignant lymphomas and wpecified leukemia. 2) Tonsillectomy was associated with wher risk of unspecified leukemia. 3) Obesity was a predictive M o r of excess risk of Hodgkin's disease, whereas leanness was associated with higher risk of other lymphomas. 4) Both coffee drinking and heavy cigarette smoking pointed to higher L k of Hodgkin's disease, myeloid leukemia, and lymphatic leukemia. 5) Outdoor environmental exposure indicated in- creased risk of malignant melanoma. In the tendencies noted, Hodgkin's disease and lymphatic leukemia had parallel predicWe factors that were opposite to those of non-Hodgkin's maligMnt lymphoma, myeloid leukemia, and unspecified leukemias. bfignant melanoma shared none of these predictive character&ICs. These findings are presented as clues deserving further WJ~orationfor any etiologic significance they may hold for mphaden and blood c a n c e r s . 4 Natl Cancer lnst 60: 89-92, ina. jects on entering Harvard University, 1916-50, or while attending the University of Pennsylvania in 1931-40. Characteristics compiled from these records were assessed as potential predictive factors (2, 3 ) . Less than 1% of the 50,000 men has been lost to follow-up observation by alumni offices without notification of deaths.' Causes of death were identified and classified from official death certificates with the use of the coding rules of t h e Seventh Revision of the ICD (1957) ( 4 ) . In this review, 4 control subjects for each site-specific cancer decedent were chosen randomly from classmates born in the same year and known to survive the decedent. Relative risk estimates for each cancer were computed for 17 characteristics previously studied for their relationship to risk of fatal Hodgkin's disease ( I ) and 2 additional characteristics suspected as precursors of skin cancer ( 5 ) . A method of matched set analysis was used (6-8).Statistical probability values are quoted for twotailed tests. Due to the bimodal structure of age-specific death rates for Hodgkin's disease ( I ) , relative risk estimates for predictive characteristics were examined separately for men younger and older than 45 years of age. Inasmuch as no differences were found, `results were presented-for these decedents as a group. Similarities in the predictive characteristics and relative risk estimates of reticulum cell sarcomas, lymptiosarcomas, multiple myelomas, and other non-Hodgkin's lymphomas permitted combination of these specific entities into a single group, "non-Hodgkin's lymphomas," as well. Characteristics recorded at college physical examinations of j0,OOO male former students from Harvard University (Cambridge, Mass.) and the University of Pennsylvania (Philadelphia, Pa.) were reviewed for their relationship to risk of fatal lymphaden, blood, and skin Qncers. Students from a 35-year period were traced for 1.71 million person-years of observation during *hich 277 deaths from these cancers were identified. %pothc.,-s of causation of Hodgkin's disease ( I ) were [wed against non-Hodgkin's malignant lymphoma, mab a n t melanoma, lymphatic leukemia, myeloid leukemia, and other and unspecified leukemia. Relative risks Of these cancers were computed for potential predictive characteristics in youth: history of contagions, familial `rndencies, physical attributes, personal traits, and SO"era1 influences. O u r purpose was to identify leads for tubsequent study of these cancers. Ph\sical examinations were performed on study sub- RESULTS In an earlier report ( I ) , we concluded that the risk of ABBREVIATION USED: ICD = International Classification o f Disease. I Received May IO, 1977; accepted August I . 1977. ? This report is No. XV in a series o n chronic diseases in former college students. JSupported by Public Health Service grants HD04753 from the National Institute o f Child Health and Human Development and 'AGO0309 from the National Institute on Aging. Department of Biomedical and Environmental Health Sciences, University o f California School o f Public Health. 303 Earl Warren Hall, Berkeley, Calif. 94720. Department o f Epidemiology. Harvard University School of Public Health, 677 Huntington Ave., Boston, Mass. 02115. e Address rcpn'nt requests lo Dr. Paffenbarger, California- State De- 'partment of Health. 2151 Berkeley Way, Berkeley, CaIif. 94704. For example, as o f the summer of 1975. mailing addresses were unavailable for only 206 (0.62%)o f 33,512 surviving Harvard University men who matriculated in 1916-50. Since Harvard University often receives notification upon the death o f these men, the followup status o f many o f them ultimately becomes known. 'OL. 60, NO. 1 , J A N U A R Y 1978 89 J NATL CANCER INST ,/ 90 P A F F E N B A R C E R , W I X G , A N D H Y D E developing Hodgkin's disease was decreased among subjects in this-population who had experienced common contagious diseases of childhood before entering college a n d was increased among those in the population who had a history o f early parental death, especially from cancer, who were obese, or who were heavy cigarette smokers or coffee drinkers. We had fcund no relationship between case cluster- ing, a history of tonsillectomy, or sibship status and risk of Hodgkin's disease, though others had reported such a relationship (9,10). We have tested similar hypotheses of causation for ocher lyrnphaden, blood, and skin cancers, together with hypotheses that might implicate exposure to sunlight (outdoor work, physical activity, and geographic origin), and present these findings in tables 1 and 2. T h e cancers shown a r e Hodgkin's disease (ICD #201), other lymphomas (ICD #200, 202, 203, and 205), malignant melanoma of the skin (#190), lymphatic leukemias (#204.0), myeloid leukemias (#204.1), and other and unspecified leukemias (#204.2, 204.3 and 204.4). In this series the mean ages at death from these cancers, respectively, were 45.3, 50.0, 48.2, 54.5, 49.2, and 47.2 years. Neither age at onset of infectious diseases, urban versus rural origin, public versus private secondary schooling, nor socioeconomic status indicators distinguished the cancer patients from the controls. Unlike Hodgkin's disease, none of the other cancers showed evidence of bimodal, age-specific incidence. No clustering of any cancers was found in relation to college class, geographic origin, athletic status, sibship, or orher circumstances potentiating transfer of infectious agents. None of the cancers, except Hodgkin's disease and glsome tendency in lymphatic leukemia, was found ciated wirh early death of a parent or familial histodd cancer. Other characteristlcs predictive of Hodgkin's (45 men) included failure of men to common contagious diseases before (especially pertussis, scarlet fever, sity, heavy cigarette smoking, and death rate for Hodgkin's disease person-years of observation. There were 89 fatal cases o f non-Hodgkin's of early varicella; this finding was quite the the experience found with Hodgkin's diseas ficiency o f other contagious diseases of found predictive of Hodgkin's disease, show tionship to the non-Hodgkin's lymphomas. relative risk for Hodgkin's disease, the risks of died from these other lymphomas tended to be e less often in outside remunerative work duri youth and were more apt to.have been coffee than were their classmates. ': T-E 1.4haracteristics recorded at eo Hodgkin's diseuse, other lymph0 Characteristic Frequency of characteristic; % Frequency of characteristic, 5% -Ponderalindex, Cl2.9 Cigarette smoker 48.9 43.2 0.9 0.78 55.0 52.3 64.4 63.6 0.9 0.84 61.7 70.9 24.3 37.8 1.9 0.06 25.9 14.9 29.7 35.0 1.4 0.9 0.54 58.0 62.2 1.5 0.12 65.5 , 68.2 0.5 0.04 19.8 25.0 1.3 1.1 1.3 J NATL CANCER INST PREDICTORS OF MALIGNANT LYMPHOMAS, MELANOMAS, A N D LEUKEMIAS 91 !-rrsLE ?.--Characteristics recorded at colLege physical examination for men who later died from myeloid, lymphatic, or other and unspecified I !- - leukemia, and for their respective controls Lymphatic leukemia Myeloid leukemia Other and unspecified leukemia !.::aracteristic / Rist0l-y of: PertUSSls Scarlet fever Diphtheria Sleasles (rubeola) \raricella Mumps Rube112 Ip~flur!:x ,a parent dead Only-child status ?t siblings T~nsillectomY pondera1 index, 42.9 Cigarette smoker io+ per day cotfee drinker fiercise, <5 h r h k Outside remunerative work s e w England origin Frequency of characteristic, % Of108 Of27 controls ddeencets- Relativerisk 68.9 19.6 2.8 90.7 68.2 57.1 50.0 17.7 i6.0 12.1 55.1 59.4 24.1 30.5 9.7 46.3 34.7 9.5 40.9 34.6 11.5 7.7 80.8 ' 38.5 53.8 ' 100.0 17.4 26.9 4.0 64.0 65.4 26.9 34.8 17.4 66.7 45.0 7.7 44.4 0.2 0.6 2.7 0.4 0.3 0-.9 1.1 i.8 0.3 1.3 1.2 1.2 1.3 2.7 2.7 1.6 1.0 1.3 . Frequency of characteristic, % Of164 Of41 controls ddeencet-s Relativerisk <0.01 0.36 0.26 0.16 0.01 0.90 - 0:90 6.24 0.26 0.56 0.65 0.76 0.57 0.17 0.06 0.48 1.00 0.69 63.0 13.6 4.9 84.0 59.9 53.2 33.3 23.4 15.4 9.4 70.6 55.7 29.0 30.9 7.0 49.0 31.9 11.4 46.3 53.7 17.1 7.3 87.8 51.2 51.3 33.3 8.8 20.0 15.0 62.5 62.5 15.0 50.0 19.4 66.7 50.0 20.0 51.9 0.7 1.3 1.5 1.4 0.8 0.9 0.8 0.4 1.4 1.8 0.7 1.2 0.4 2.4 3.6 3.2 2.4 1.8 1.2 Frequency of characteristic, % of 120 RelaOf30 tiverisk controls dents 0.26 0.54 0.55 0.54 0.46 0.84 0.81 0.13 0.47 0.30 0.27 0.60 0.08 0.03 0.03 0.02 0.16 0.27 0.66 60.8 12.5 5.9 89.1 62.2 42.4 20.0 22.1 E.! 12.5 , "55.0 64.1 28.0 24.3 6.4 57.0 47.1 13.0 40.0 66.7 13.3 6.7 90.0 90.0 34.5 0.0 ' 23.1 - 16.7 13.3 46.7 89.3 31.0 29.6 3.8 52.6 56.3 9.1 40.0 1.3 1.1 1.1 1.1 5.5 0.7 - 1.0 -1.1 1.1 0.7 4.7 1.1 1.3 0.6 0.8 1.9 1.0 0.8 0.51 0.90 0.90 0.87 cO.01 0.-38 1.00 8.91 0.90 0.40 0.01 0.77 . 0.63 0.65 0.73 0.35 0.97 0.82 43 men died from malignant melanoma, an incidence rate of 2.6 per 100,000 person-years of observation. Of the potential predictive characteristics examined, only outside (presumably outdoor daytime) employment before college was significantly associated with malignant melanoma: Outdoor workers had a relative risk 3.9 Limes greater than that of men who did not report outdoor work. In contrast to students who came from other areas, students from New England had only half the risk of malignant melanoma, but the difference was not statistically significant. The 27 lymphatic leukemias (incidence rate of I.6/ 100,000 person-years o f observation) were notably similar to Hodgkin's disease in predictive characteristics o f the subjects. Risks were significantly lower for men who had experienced common contagious diseases o f childhood before entering college, particularly pertussis and varicella. T h e relative risk for heavy cigarette smoking and for coffee drinking was 2 . 7 , another characteris:;:: parallel to that seen with Hodgkin's disease pa tienr ,, The 41 men with myeloid leukemia (an incidence rate of 2.4/100,000 person-years of observation) did not hare the predictive factor of early history of contagion. A s was found for students who later developed nonHodgkin's lymphomas, lean rather than obese college students tended to develop these leukemias. T h e rela[Ire risk o f myeloid leukemia was 2.4 for cigarette Imokers, 3.6 for heavy cigarette smokers, and 3.2 for drinkers of any amount of coffee; all tendencies were greater than unity. found 30 other and unspecified leukemias (acute, Is; monocytic, j; and unspecified. 7) that gave an s 60, N O . [, J A N U A R Y 1978 incidence rate of 1.7 per 100.000 person-years of observation. An early history of varicella was associated strongly with these leukemias (relative risk, 5 . 5 ) , as was a history of tonsillectomy (relative risk, 4.7). In such predictive relationships these leukemias are parterned after the non-Hodgkin's lymphomas. None of the other characteristics studied presage any of this medley of leukemias. DISCUSSION Whether or not the data reviewed here identify causal relationships, the differences in predictive characteristics between Hodgkin's disease and other malignant lymphomas d o suggest differing etiologies. By the same token, similarities in predictive factors in youths who later develop Hodgkin's disease or lymphatic leukemias could signify common etiologies as could similar predictive factors in youths who later develop non-Hodgkin's malignant lymphoma, myeloid leukemia, or unclassified leukemia. Malignant melanoma shares no host predictive characteristics with the other cancers and seems to result from another cause. Suggested hypotheses of causation for Hodgkin's disease, based primarily on history of reduced frequencies of childhood contagions by the time the youth enters college, are as follows: 1) Birth cohorts ivith inadequate immune mechanisms either succumbed to or escaped childhood infections, leaving the survivors predisposed to Hodgkin's disease whether or not Hodgkin's disease is of infectious origin. 2) Originally normal or abnormal immune mechanisms were inadequately challenged by early childhood infections and so were not precondi- J NATL CANCER INST . `92 PAFFENBARGER, W I N G , A N D HYDE tjoned to withstand the assault of Hodgkin's disease later in life. 3) Hyperimmune mechanisms aborted clinical attacks of childhood contagions (other than the ubiquitous measles) and fostered an autoimmune phenomenon leading to Hodgkin's disease. Any of these hypotheses fit the observed inverse relationship of childhood contagious diseases and subsequent risk of Hodgkin's disease and are equally applicable to lymphatic leukemia patients who have a similar array of predictive characteriscics. The observation that a predictive factor of Hodgkin's disease is reversed for nomHodgkin's malignant lymphomas, i.e., occurrence of childhood infections (notably varicella), invites for other lymphomas some adaptations of the Hodgkin's hypotheses of causation: 1) An inadequate immune system might fail to contend with the effects of varicella infection, leading to development of non-Hodgkin's maligant lymphoma. 2) T h e varicella itself may be a slow-acting virus that directly causes lymphoma. 3) T h e varicella infection may have altered o r sensitized the lymphatic system by a single o r repeated process that renders, the subject receptive to development of the lymphoma either as a subsequent infection o r as an autoimmune phenomenon. T h e observation that childhood varicella is predictive of non-Hodgkin's malignant lymphoma and unclassified leukemia has some parallel to the observation that a history of tonsillectomy also tends to be predictive ofthese cancers. It is likely that the increase in tonsillectomies accompanies the increase in infectious diseases of childhood. At this point, the aforementioned hypotheses are merely leads in the search for causation, though in several animal models evidence that autoimmunity may lead to certain lymphoreticular diseases has been found (11). Others have reported associations between influenza or varicella and the subsequent development of leukemias a n d lymphomas (12-14). Contrary to o u r findings, one group of investigators found an association between varicella and lymphatic leukemia (12), another between varicella and unspecified and lymphatic leukemias ( 1 3 ) , and a third between influenza and lymphatic-hematopoietic neoplasms (14). These associations have been based, however, on in utero infection with varicella o r influenza rather than on childhood infection, and the end points of interest have been childhood leukemias and lymphomas rather than adult lymphaden and blood cancers. Patients with Hodgkin's disease and non-Hodgkin's malignant lymphomas have opposing-predictive in obesity and leanness that occur during youth such as childhood contagions. Nevertheless, tremes of weight for height differ significa the norms of classmate controls and deserve in account. T h e patterns of associations involving heavy ci munity or sensitizing processes, the roles of dictive characteristics are more likely those implicated secondary characteristics. REFERENCES ( I ) PAFFENBARGERRS JR, WING AL, HYDE RT: Cham youth indicative of,adult-onset Hodgkin's disease. cer Inst 58:1489-1491, 1977 (2) PAPFENBARCERRS JR. Wow PA, NOTKIN disease in former college students. I . Early coronaiy hean disease. Am J Epidemiol83: (3) PAFFENBARGR~SR J R, ASNES DP: Chronic college students. 111. Precursors of suicide i (5) FEARS TR. S c o r r o J , SCHNEIDERMAMNA: Mathemau ( 8 ) ROTHMAN KJ: Computer individual matching. In (9)VIANNANJ, POUN AK: E (12) VIANNANJ, POUN AK: Childhood lymphatic leukemia: seasonality and possible association with congenital Am J Epidemiol 103321-332, 1976 (f3)ADELSTEINA M . DONOVAN JW: Malignant disease in nancy. Br Med J 43629-631, 1972 (14) FEDRICKJ. ALBERMANED: Reported influenza in p and subsequent cancer in the child. Br Med J 2:485- J NATL CANCER INST a VOL. 60, NO. I , J ANUARY