Appendectomy and cancer risk

J. &on. Dis. 1968, Vol. 21, pp. 391-415. Pergamon Press Ltd. Printed in Great Britain

APPENDECTOMY

AND CANCER RISK*

AN EPIDEMIOLOGICAL

EVALUATION

LYONHYAMS,M.D., M.S.t and ERNESTL. WYNDER,M.D. Division

of Environmental (Received

Cancerigenesis, Sloan-Kettering Institute for Cancer Research, New York, N.Y. 10021, U.S.A.

26 September

1967; in fitraI revised form 1 July 1968)

Two STUDIEShave reported significantly higher past appendectomy rates in patients with some types of cancer when contrasted to noncancerous control groups [l, 21. From these findings, certain relationships-how the lack of appendicial lymphoid apparatus may increase the risk for cancer-have ‘been described. Such a potentially important result requires re-exploration. Appendectomy should be viewed as a compound epidemiological event. Its occurrence is probably influenced by a host of factors such as age, race, sex, socio-economic level, and place of birth. Studies which do not consider these variables simultaneously with risk evaluation run the danger of mistaking a secondary association for a direct one. The purpose of this study is to contribute to the accumulating data on appendectomy prevalence in various diagnostic groups, to report on the epidemiological concomitants of appendectomy and to evaluate the relationship between previous appendectomy and cancer in man. PREVIOUS STUDIES MCVAY [Il. in an autopsy study, reported a significant relationship betweeen previous appendectomy and the subsequent development of colon cancer. He also found significant differences in appendectomy rates between a control group consisting of patients with cardiovascular disease and a study group composed of patients with various malignant neoplasms. GR,OSS [3], in an interview study, could not relate appendectomy or tonsillectomy prevalence to cancer. He claimed, however, that cancer patients had had their procedure significantly later than the control groups. BIERMAN[ 21, utilizing both autopsy and interview material, found a higher rate of previous appendectomy in a mixed cancer group. Separate analysis implicated cancer of colon, ovary and breast as being particularly affected by the relationship. He also reported significantly higher rates in granulocytic leukemia and Hodgkin’s disease. HOWIE [4] conducted a British study utilizing mailed questionnaires. He could not demonstrate a higher rate of appendectomy in any *This study was supported by the American Cancer Society Grant E-87E and in part by NC1 Grant CA-08748. t Present address: Department of Community Medicine, Rutgers Medical School, Rutger, The State University, New Brunswick, N.J. 08903, U.S.A. 391

LYON HYAMS and ERNEST L. WYNDER

392 TABLE 1.

COMPARISJN 0~ CRUDE APPENDECTOMYPERcE~VMGES

Study

Type of sample and total effective number in the study

AMONG VARIOUS STIJDI~~

All All noncancers cancers

Colonrectum Breast

McVay 10.4 22.6 15.7

4.0 15.5 7.7

13.6 24.8 18.3

6.0 13.0 11.9

19.0 13.0 14.4

6.0 8.0 6.9

13.0

Interview (122) and autopsy (1287)

35.0

24.3

47.9

61.7

Interview

19.0

21.0

20.0

21.8

Interview and physical inspection (448)

9.0 27.8 16.0

15.0 26.6 19.3

4.2 33.0 12.1

31.6

Interview (1582) and questionnaire (700)*

15.3t 36.3 25.4

17.5 33.3 23.9

21.2 43.7 30.6

38.5

Autopsy : combined Howie and Timperley M F combined Bierman combined Gross combined Kessler M F combined Hyams and Wynder M F combined$

(820)

Mailed questionnaire

(1642)

(500)

*These 700 cases were controls used only as a contrast for men with Hodgkin’s disease. tThese rates exclude the lymphoma’s and the leukemia’s $The combined percentages represent a weighted average of the sex specific rates.

specific cancer group, nor in the combination of them. KESSLER [S], in a study based on physical examination, interview and record review, was unable to reject equality of appendectomy prevalence in the cancer and control groups. Table 1 presents base line data from the various investigations. The diversity of the above findings is undoubtedly a result of many factors, several of which can be explained in light of the design and methodology of each given study. It is advisable to explore such considerations. Hopefully, this will permit a more effective consideration of our methodology and will point to the obvious difficulties presenting themselves to any investigator. METHODOLOGICAL

AND

THEORETICAL

CONSIDERATIONS

Selection of comparisongroups

Problems associated with the choice of a suitable control group are well known. Does a difference in the proportion between two groups, the case being higher than the control, reflect an unusually high rate in the case group or low one in the control group? Do heart patients have a lower rate of appendectomy than cancer patients who represent a population value or vice versa? A correct appraisal of the group most representative of the standard population value seems essential in order to answer the above question. Unfortunately, the ideal control in this study is neither easily defined nor sampled. In this situation it seems sensible to adopt the following orientation. Whenever little is known about the relationship between the variable in question, in this case appendectomy prevalence and its distribution in diverse populations, the investigator should not restrict himself to a single control. Groups of varying composition should be subject to

Appendectomy

and Cancer

Risk

393

analysis. Consistency in outcome would strengthen the conclusions arising from any observed difference in proportions, whereas inconsistencies would detract from them. There are also problems associated with choosing a case group. Comparing a mixed cancer group to some control presupposes a relationship between past appendectomy and the basic cancer processes. It further assumes that there is no differential risk for specific cancers. Most investigators would deem such a hypothesis as unlikely. Hence, one must adjust for the influence of the differential weights of specific cancers in a mixed cancer group or handle distinct entities separately. The latter is preferred, especially in view of previous publications. However, this subdivision demands a reasonable number of patients in each group. A compromize may be found in some sensible arrangement of the neoplastic categories. Cancers, such as colon cancers, breast cancer, and the lymphoma-leukemia group, should be reviewed separately while others, especially if their observed rates are smaller and homogeneous, may be fused. Source

of data

An autopsy report in conjunction with other materials found in the medical record of the patient has the advantage of virtually eliminating errors in diagnosis. In addition, since this terminal event precludes the possibility of developing a new disease, errors involving multiple categorization are limited. The prime objection to the use of autopsy and recordroom material is that appendectomy history may be reported differently in autopsy and the medicalsurgical history, depending on the given disease of a patient. Pathologists, for example, may be less prone to investigate and report appendicial presence or absence in a patient succumbing from a heart disease than from abdominal cancer. Histories are notably unreliable in this regard. This situation could result in an important bias---namely, an underestimation of the percentage of previous appendectomy in the non-cancerous groups. Secondly, it would seem most difficult to separate appendectomies performed prior to the onset of the pertinent disease from those performed concomitant with the illness. Patients with colon cancer, while undergoing some operative procedure connected with the disease, may simultaneously have their appendices removed. This condition would lead to a magnification of the appendectomy rate in the cancer group. Furthermore, autopsy populations, at least in the United States, are poor indicators of general population characteristics. Additional drawbacks to record study are that they do not allow us to obtain a consistent base line of epidemiological data and often do not afford the opportunity of differentiating between “primary” and “secondary” appendectomy. Studies based on personal interviews with patients correct the weaknesses evident in autopsy design. Their chief disadvantages are related to the number of unconfirmed diagnoses and the reliability in which appendectomy status is elicited. The first source of error can be effectively limited through insistence on pathologically confirmed diagnosis, the latter through careful questioning and analysis of appendectomy information. Designs which rely primarily on information collected from mailed questionnaires

394

LYON HYAMSand ERNESTL. WYN~ER

potentially share the properties of the personal interview. Certain types of data, however, are difficult to obtain, such as past disease and operative experience, and the differentiation between primary and secondary appendectomy. Eflect of certain core factors on appendectomy:

sex and the form of appendectomy

Secondary or incidental appendectomies are more frequent in women than men, since they are subject to a more diverse series of abdominal operations. Unequal representation of the sexes in contrast groups must be either adjusted statistically in testing for differences in total appendectomy rates, or preferably handled separately. Furthermore, there are disadvantages in restricting comparisons to total appendectomy prevalence. A subdivision into primary and secondary procedures allows for the more distinct analysis of possible epidemiological relationship.

The present age of the patient, his age of diagnosis, and perhaps the year in which he was at that specitied age should be considered. Although appendicitis and subsequent primary appendectomy show the greatest incidence between the years of 15-25, primary appendectomy does not occur uncommonly after 30 [6]. Secondary rates start later in life and cumulatively increase with age. Consequently, age is directly proportional to the probability of a past procedure and must be rigorously controlled for in the analysis. One can also speak about the temporal mode in performing appendectomies, which reached its peak in the mid 1930’s and subsequently declined 171. Thus persons 50 yr old in 1965 were in prime risk age at prime risk time, i.e. at age 20 in 1935, and should evidence a relatively higher rate of appendectomy than, for example, persons who reached 50 yr of age in 1955. From the viewpoint of analysis, the temporal mode discrepancies are adjusted for simultaneously with age-if the cases and controls are collected within a few years of one another. However, it can confine the type of comparisons, intra- or interstudy, that differ markedly to the year that diagnosis is established. Such problems arise within MCVAY’S work [l] as is also stressed by HOWIE [4]. An additional criterion has to do with some epidemiological factors and their possible relationship to the risk of past appendectomy. Several studies, for instance, have shown associations between psychosocial variables and primary appendectomy [8-lo]. METHOD

General

During 1965 and 1966 a questionnaire was administered by interviewers to 1507 subjects in various hospitals and clinics in New York City*. Basic epidemiological data and information regarding appendectomy history were elicited. The data *The interviewing of add,itional lymphoma and leukemia patients extended into 1967. The data derived from these people were not used for the examination of appendectomyepidemiological factor relationships, which is based on the original 1507 subjects. Apart from the above mentioned sample sources, self-administering questionnaires were distributed to 700 healthy students to serve only as control data for patients with Hodgkin’s disease.

Appendectomy TABLE 2.

DISTRIBUTION

OF

sAhmE

and Cancer

POPULATION

AMONG

Risk

395

HOSPITALS AND THE DATE SAMPLED

Effective number in sample* Male Female

Hospital Memorial Cancer Centert Montefiore Hospital Brooklyn and Manhattan Veterans Administration Hospitals New York University, Cornell Division Manufacturing Corporation Preventive Medical Strang Clinic Total

421 99

374 130

146

6

37 118

27 44

79 900

101 682

Date sampled 1965, 1966, 1967 Summer 1965 1965

Short

1965 interval 1965

in

Institute-

*Patients interviewed with benign neoplasms, equivocal diagnosis, tomy status was uncertain are not included. iOf the 795 patients sampled at Memorial Cancer Center, 1967. These data did not enter the epidemiological tabulations.

1965

or those whose appendec75

were

interviewed

in

obtained was coded, punched, and sorted on standard IBM apparatus and was subsequently analyzed in accordance with a loose a priori determined plan. Sampling

The institutions sampled, the dates of such activity, and the number of subjects entering the study from each place is given in Table 2. There are four broad groups of subjects: neoplastic; cardiovascular; patients with other diagnosed diseases; and a group of healthy individuals. In each place, special care was taken to select a representative sample. There Patients considered too ill to be were no refusals to submit to the questioning. questioned on one day were interviewed the following week. Interview

and questionnaire

The interviewers were experienced and trained for the special requirement of this study, which was an intelligent evaluation of appendectomy status. Primary appendectomies are defined as resulting from appendicial symptomatology and are not accompanied by the removal of another organ. Secondary or incidental removal of the appendix results from other organ-directed operative procedures. The reliability of past appendectomy status, as reported by the patient, was investigated through the examination of the abdomen for scars and by careful questioning. Our impression was that errors, whether positive or negative, in reporting primary appendectomy were negligible. With regard to incidental appendectomies, our impression was that when the patient said the appendix had been removed concomitantly with another procedure, this answer was reliable. There were two classifications of cases in secondary procedures, however, that may have cause3 some error. One group had some type of abdominal and/or pelvic procedure performed in the past and did not know their appendectomy status. These cases Another group, with a history of such were noted but not used in analysis. abdominal operative procedures, denied any concomitant appendectomy. Not all of

396

LYON HYAMS and ERNESTL. WYNDER

these negative responses were likely to be accurate. to evaluate.

This source of error is difficult

Questions

On the epidemiological form, standard questions were asked. The place of birth was not originally included. This was an oversight, especially in view of what later seemed to be an obvious relationship between nativity and the risk of appendectomy. Fortunately, it was possible to recover such information from over 95 per cent of the records of patients seen at Montefiore Hospital and The Preventive Medical InstituteStrang Clinic. It was not feasible, however, to recover this data in sufficient quantity at the other institutions to warrant presentation. Questions were also asked in regard to constipation, hemorrhoids, and hernia in an attempt to relate these procedures to appendectomy. Careful questioning in regard to past history of cancer was also effected. Diagnosis

Effort was expended toward confirming the diagnosis given by the chart. ,In those patients where microscopic confirmation was possible, it was generally obtained. Thus, we were able to histologically confirm 89 per cent of all diagnosed cancers. In others, i.e. myocardial infarction, respiratory disease, etc., this type of assurance was not reasonably available in many instances. We have, thus, less confidence in the specificity of diagnoses in the non-neoplastic group, but we are quite confident of the dichotomy of cancer vs. non-cancer. When the diagnoses was equivocal or could represent a neoplastic process, it was not utilized in further analysis. Analytical considerations

The results are divided into a number of parts: rate comparisons, epidemiological relations, and an attempt toward integration of these two. Finally, the age of appendectomy and the lymphoma and leukemias are examined. The first section compares sex-specific, age-adjusted appendectomy rates between various cancer and control groups*. Our purpose in varying the composition of the control comparison group is to show that outcome can be strongly dependent on this choice. Special care was taken, in some instances, to maximize the possibility of detecting any differences. Since we did not originally support the reported appendectomy-cancer risk relations [l, 21 because of a lack of suitable epidemiological control and other methodological limitations, we have guarded against biasing our report in this way. Ideally, appendectomy rates in cancer patients should be compared with a more general population of healthy people. This was impractical to obtain. The second or epidemiological part of the result section evaluates the effect of *For age adjustment, we utilized the method suggested by YATESas pointed out and illustrated by FERTIG[ll] for the treatment of disproportionate frequencies in the two-way layout analysis of variance. KINCAIDhas presented a generalization of the methodology [12]. MANTEL[13] and CORNFIELD [14] have suggested alternate solutions which, in the absence of interaction, are approximately equivalent to this procedure. Mantel’s value of x 2 is simultaneously presented in Table 3. Age stratification was in four classes: Under 50; 50-59; 60-69; and 70 and above.

Appendectomy

and Cancer

Risk

397

certain epidemiological factors on the risk for appendectomy. General x2 analysis is utilized extensively in this section. For men, it was sufficient to relate total vs. no appendectomy to the epidemiological factor, since primary and total appendectomy However, for women the total tabulation was rates are practically synonymous. divided into primary and secondary versus no appendectomy for two additional separate evaluations. We initially examined appendectomy rates in various diagnostic subdivisions and confirmed that factor-rate relations were not diagnosis dependent. Subsequently, diagnostic divisions were fused for analysis. The totals on the tables vary with each epidemiological factor. This is a function of those individuals whose specific epidemiological factor was not known. In reporting the percentages of incidental appendectomy, it was considered misleading to divide the number of incidental appendectomies by the total number in the group. Included in this total were those who had already had their appendices removed and were not at risk. Therefore, it was decided to reduce the denominator by the number of primaries and call this percentage “corrected secondaries”. We recognize this as an extreme measure, since some proportion of these people underwent this procedure at a relatively late age and were, therefore, exposed to “appendiNevertheless, such a rule is convenient and citis” and the subsequent operation. expressive. All reported percentages of secondary appendectomy are corrected in this manner. Each a priori test was performed at a significance level of 5 per cent. Trends of interest were verbalized and presented in the integration of material for discussion. RESULTS

Prevalence information : crude appendectomy rates

Table 3 presents sex-specific crude past appendectomy rates in various diagnostic divisions and represents base line data on 609 cancer patients and 676 controls. The lymphoma and leukemias are reported elsewhere. Other obviously related information is given simultaneously. Although it may be misleading to utilize these crude rates for visual comparison, one is nevertheless drawn to the great variability shown in the different parameters of appendectomy prevalence among the diagnostic categories*. Its magnitude, expressed as the range, is given in the table. As one would expect, secondary appendectomy rates and thus the total figures are considerably higher in females. In addition, their primary rates are, in every instance, also somewhat higher than the corresponding male proportion. The demonstration of a significant difference in total appendectomy proportion between a cancer and a control group is dependent not only on the cancer group, but also upon the choice of the control group. This is especially true in females whose appendectomy prevalence is 20.5 per cent in the “other non-neoplastic disease” category, while in the “no present disease” it is nearly twice as large, 40.4 per cent. To illustrate this point, we tested the Null Hypothesis of equality of the proportion of past total appendectomy procedure between various groups of interest. *This brings up the statistical problem of possible underlying heterogeneity, while analysis is based on homogeneity. If heterogeneity was indeed present, then one uses an under-estimate of the variance, and differences in proportions are judged significant with greater ease.

LYON HYAMS and ERNEST L. WYNDER

398

TABLE 3. Diagnostic

CRUDE APPENDECTOMYRATES 1~ VARIOUS DIAGNOSTICc~~~xjo~lf~

No.

category

Median present age

Primary appendectomy (No.1 (%I

Secondary appendectomy (No.1 (%I

Total Corr. %t

(No.)

(%)

13

Specific cancers

Head and neck Lung and bronchi Prostate and penis Breast Cervix Colon-rectum Other cancers* All cancers except lymphomas and leukemia Specific controls No present disease Cardiovascular Other non-neoplastic disease All

controls

E F F M F M F

82 25 52 27 122 37 99 71 60 34

65.5 59.5 61.7 70.2 58.8 53.0 65.8 62.2 59.6 59.0

11 5 7 2 25 6 18 17 2 3

13.4 20.0 13.5 7.4 20.5 16.2 18.2 23.9 3.3 8.8

2 3 0 1 22 4 3 14 3 6

2.4 12.0 0.0 3.7 18.0 10.8 3.0 19.7 5.0 17.6

2.8 15.0 0.0 4.0 22.7 12.9 3.7 25.9 5.2 19.4

f 3 47 10 21 31 5 9

15.9 32.0 13.5 11.1 38.5 27.0 21.2 43.7 8.3 265

M F

320 289

64.5 59.3

40 56

12.5 19.4

9 49

2.8 16.9

3.2 21.0

49 105

15.3 36.3

M F M F M F

1% 141 101 62 103 73

47.1 52.6 60.7 63.0 53.0 49.0

37 33 12 9 14 12

18.9 23.4 11.9 14.5 13.6 16.4

1 24 3 11 3 3

0.5 17.0 3.0 17.7 2.9 4.1

0.6 22.2 3.4 20.8 3.4 4.9

38 57 15 20 17 15

19.4 40.4 14.9 32.3 16.5 20.5

M F M F

400 276

52.2 54.1

63 54 3.3 8.8 -

15.8 19.6 18.9 29.4

7 38

M F

1.8 2.1 13.8 17.1 O.O- 5.2 4.9 - 25.9

70 17.5 92 33.3 8.3 -21.2 20.5 -44.1

*The category “other cancers” in males corresponds to those cancers of the digestive and excretory system not previously considered and in addition, a small number of patients with cancer of the thyroid, skin, etc. The above description applies to the female category but, in addition, includes lung and bronchial cancers. tin future tabulattions the percentage of secondary procedures given are the corrected percentages.

Table 4 presents the results of these age-adjusted tests for difference in total appendectomy prevalence between diverse diagnostic groups. The various tests between different control and cancer groups were performed for two reasons: to indicate that the outcome may be a function of the type of control group and to increase the sensitivity to any real differences that may exist. Some obvious conclusions may be obtained from this analysis. Equality of total appendectomy prevalence was accepted in all male contrasts. Since some of these comparisons were chosen to maximize the possibility of obtaining a significant difference in rates, this increases our confidence in the absence of a ditference in rates. The difference between female combined controls and combined cancers was nonsignificant. Colon-rectum and breast cancer did not show rate differences when compared with the female control group without present disease or with the cardiovascular group alone. They did show significant differences when the control designations were fused or changed. Colon-rectum and breast cancers exhibitA significantly higher rates when compared to other non-neoplastic diseases or the

Appendectomy TABLE 4.

Contrasting

and Cancer

399

Risk

RESULTSOF AGE-ADJUSTED TESTS OF DIFFERENCES IN TOTAL APPFXDECTOMY PREVALENCE BETWEEN VARIOUS DIAGNOSTIC GROUPS

groups

Male AI1 cancers t vs. all controls All cancers vs. cardiovascular Colon-rectum vs. all controls Colon-rectum vs. no present disease Colon-rectum vs. cardiovascular Colon-rectum vs. cardiovascular and other disease Female All cancers vs. all controls Colon-rectum vs. all controls Colon-rectum vs. no present disease Colon-rectum vs. cardiovascular plus other non-neoplastic disease Primary rates alone Secondary rates alone Colon-rectum vs. other non-neoplastic Breast cancers vs. all controls Breast cancers vs. no present disease Breast cancers vs. cardiovascular disease Breast cancers vs. cardiovascular and other non-neoplastic disease Primary rates afone Secondary rates alone Breast cancers vs. other nonneoplastic disease

Observed

x1%:

Probability

levels

0.70 0.23 0.24 0.01 2.05 1.21

(0.57) 9: (0.01) (0.32) (0.05) (1.41) (1.15)

0.50 0.70 0.70 0.80 0.20 0.30

> > > > > >

p > 0.40 p > 0.60 p > 0.60 p > 0.70

0.97 3.43 1.37 6.21**

(0.75-I (3.09) (1.10) (4.99)

0.40 0.10 0.30 0.025

> > > >

p > 0.30 p > 0.05 p > 0.20

3.57 5.992’ 8.23** 1.77

(6.60) (LS9)

0.10 >p>o.o5 0.025 > p > 0.01 0.005 > p > 0.001 0.20 > p > 0.10

0.49

(0.08)

0.5

> p > 0.40

1.58

(1.62)

0.5

> p > 0.25

4.57**

(3.13)

0.05 > p > 0.025 0.20 > p > 0.10 0.025 > p > 0.01

(5.73)

0.025 > p > 0.01

1.97 5.56** 6.38**

p > 0.10

p > 0.20

p > 0.01

tNone of these diagnostic divisions include the lymphoma or leukemic group of diseases. $A double star, **, ind,icates that a statistically significant difference exists, the magnitude of which is given in the next column. §The Mantel-Haenszel x12 statistic [13] is given in these parentheses to serve as a methodological comparison. In general, it gives a similar probabilistic outcome. When there are discrepancies, the Mantel-Haenszel statistic gives a more conservative value.

fused cardiovascular plus other non-neoplastic disease categories. These findings illustrate the problems of choosing a suitable standard. In both the colon-rectum and the breast cancer groups, it was the difference in secondary appendectomies that contributed primarily to the difference in the total rates (x2 values are presented in Table 4). These findings, as well as the extremely variable appendectomy rates in all dimensions, give the impression that other epidemiologic considerations must play a role in the past history of appendectomy. Age- and sex-specific rates Of course, the crude appendectomy rates are not comparable. The basic data for contrasting purposes can be given in age- and sex-specific tabulations. The examination of such data is of interest. These tabulations for condensed and control groups are given in Tables 5 and 6. Next we tested the differences in total appendectomy rates between the total cancer and the total control group in each age category by usual a priori tests.

400

LYON HYAMS and ERNEST L. WYNDER

TABLE 5. DISTRIBUTIONOF

-IHE TYPE OF APPENDECTOMY BY AGE IN MALECONTFLOL

AND

CANCER

GROUPS Present

age

Primary appendectomy (No.) (%)

Secondary appendectomy (No.) (corr. %)

All appendectomy (No.) (%)

:

16.3 14.3 19.6 16.7 16.4 15.5 20.5 11.3 12.5 21.7

184 42 102 66 67 116 39 71 8 23

70 48

17.5 15.1

400 318*

Under 50 Control Cancer 50 - 59 Control Cancer 60 - 69 Control Cancer 70 - 79 Control Cancer 80 + Control Cancer

27 5 20 8 11 17 5 7 0 3

14.7 11.9 19.6 12.1 16.4 14.7 12.8 9.9 0.0 13.0

3 1 0 3 0 1 3 1 1 2

1.9 2.7 0.0 5.2 0.0 1.0 8.8 1.6 12.5 10.0

30 6 20 11 11 18 8 8

Total

63 40

15.8 12.6

7 8

2.1 2.9

Control Cancer

Total number in group

This fused cancer group does not include the lymphoma and leukemia group of neoplasms. The controls represent the fusion of cardiovascular disease, no present disease, and other non-neoplastic dmisease groups. *Two cases could not be classified by age.

Although this is a good example of “over-testing”, it seemed fair to use procedures that were sensitive to differences. If we could not find them in this fashion, they were viewed as nonexistent. However, if and when differences were found, we could explore their meaning in terms of the study of their components, i.e. primary and secondary procedures. Females. Three of the four major age-specific total appendectomy rates are higher in the mixed cancer group (Table 6). None of the differences are statistically signiTABLE

6.

DISTRIBUTIONOF THE TYPE OF APPENDECTOMY BY AGE

IN FEMALE

CONTROL

AND CANCER

GROUPS

Present

age

Under 50 50 60 70 80

Control Cancer - 59 Control Cancer - 69 Control Cancer - 79 Control Cancer + Control Cancer

Total

Control Cancer

Primary appendectomy (%) (No.)

Secondary appendectomy (No.) (corr. %)

All appendectomy (No.) (%)

Total number in group

19 21 21 15 7 14 5 6 2 0

18.8 28.8 25.6 19.2 13.0 16.1 16.1 13.3 25.0 0.0

10 9 18 15 8 14 2 11 0 0

12.2 17.3 29.5 23.8 17.0 19.2 7.7 28.2 0.0 0.0

29 30 39 30 15 28 7 17 2 0

28.7 41.1 47.6 38.5 27.8 32.2 22.6 37.8 25.0 0.0

101 73 82 78 54 87 31 45 8 6

54 56

19.6 19.4

38 49

17.1 21.0

92 105

33.3 36.3

276 289

The controls represent the fusion of cardiovascular disease, no present disease, and other non-neoplastic disease groups. The fused cancer groups do not include the lymphoma and leukemia group of neoplasms.

Appendectomy

and Cancer

401

Risk

ficant when each age level is compared separately in a x2 procedure for the testing of proportions. The trend is also reversed in the 50-59 age category. Simultaneous age-adjusted analysis between these two large groups has been given previously. Inspection of the temporal trends within the primary and secondary appendectomy groups showed no consistent relationship between risk for primary and for secondary procedures. In addition, the magnitude of female secondary rates across age groups manifests random arrangement. However, in all ages except 50-59, secondary rate proportions are higher in the cancer group, although the magnitude of the difference is not striking. Such a trend brings to mind problems associated with the reliability of eliciting a past history of incidental appendectomy previous to present disease classification. Males. A visual comparison of the total appendectomy rates in men make the formality of testing unnecessary for the purpose described (Table 5). In each of the four major age-specific divisions (excluding 80+) these proportions are higher in the control group. Thus, we could not possibly demonstrate that rates are greater in the cancer group. Comparisons of rates between age within diagnostic grouping impresses one with their homogeneity, especially when a comparison is made to the female data. This can be interpreted as their being less at&ted by other influencing factors, excluding appendicitis, than the respective female groups. Male-female contrasts of past primary appendectomy. A general agecomparison of primary appendectomy between the sexes was adjusted performed separately for the two main diagnostic groupings, cancer and control. In both instances, women evidenced a significantly higher rate of such procedures than did men (xl” Control=4.47, p < 0.05 and ~12Cancer=7.27, p < 0.01). Epidemiology Race. The distribution of race by appendectomy

form is given in Table 7. Negro men have a significantly lower proportion of past appendectomy than white men (x:=6.60, p < 0.025). The very low percentage of 5.8 in this group is impressive. A similar but less pronounced disparity in appendectomy history between races is observed in women. However, these differences are nonsignificant. The somewhat increased proportion of total appendectomy for white females is reflected equally between the primary and incidental procedures. TABLE

Negro Other Total

DISTRIBUTION OF THE TYPE OF APPENDECTOMYBY RACEAND SEX Total number in group

Primary appendectomy (No.) (%)

Secondary appendectomy (No.) (corr. %)

All appendectomy (No.) (%)

M F M F M F

116 121 4 : 0

15.4 19.5 5.8 12.5 0.0 0.0

19 95 0 5 0 0

3.0 19.0 0.0 11.9 0.0 0.0

135 216 4 11 0 0

17.9 34.8 5.8 22.9 0.0 0.0

754 621 69 48 6 7

M F

120 127

14.5 18.8

19 100

2.7 18.2

139 227

16.8 33.6

829 676

Race White

7.

The x12 for men and women are calculated Negro vs. all and no appendectomy.

independently

for the four-fold

table:

White and

LYON HYAMS and ERNEST L. WYNDER

402 TABLE 8.

DISTRIBUTION OF ?HE TYPE OF APPENDECTOMYBY RELIGIONAND SEX

Primary appendectomy (%) (No.)

Religion Hebrew (non-dietary) Hebrew (dietary) Protestant Roman

Catholic

Other-none Total

Secondary appendectomy (No.) (corr. %)

All appendectomy (No.) t%)

Total number in group

M F M F M F M F

20 36 8 15 52 31 36 39 3 6

14.5 21.8 11.0 15.6 18.1 16.9 11.9 19.3 15.0 22.2

2 23 5 14 2 28 5 31 4 4

1.7 17.8 7.7 17.3 0.8 18.4 1.9 19.0 23.5 19.0

22 59 13 29 54 59 41 70 7 10

15.9 35.8 17.8 30.2 18.8 32.2 13.6 34.7 35.0 37.0

138 165 73 96 288 183 302 202 20 27

M F

119 127

14.5 18.9

18 100

2.6 18.3

137 227

16.7 33.7

821 673

M

F

The x; for men and women are calculated independently for the 3 by 2 table: all Hebrew, Protestant and Roman Catholic vs. all and no appendectomy.

Religion. Equality of rates in appendicial removal by the three main religious groupings in men, i.e. Jews, Roman Catholics and Protestants, is accepted when testing through 2 (Table 8). We decided to test differences in rates between Hebrew dietary and non-dietary groups in a four-fold table of primary procedures. Although this comparison was statistically nonsignificant, the trend toward a higher proportion of primary appendectomy in non-dietetic Jews is maintained in varied subdivisions of the data. This applies to females as well as males. Such trends are probably attached to the place of birth rather than the diet. Women, as compared to men, evidence a similar outcome in all major subdivisions and there is homogeneity of rate in the various appendectomy forms. Education. In men, there are no sign&ant differences in past appendectomy procedure among the four educational classifications (x,2=4.16, p > .OS, Table 9). However, there is a seemingly positive regression with increasing educational attainment. Another form of test more sensitive to specific trends might have yielded a “significant” result. For women, there is a signilkant disparity between total appendectomy rates among the various educational categories (Table 9, x32= 9.01 ,p < 0.05). The greatest percentage of past total appendectomy is found within the college l-4 yr category; not at the graduate level. The association between educational level and total appendectomy prevalence in women is mainly a function of the primary procedure. The fiz of primary vs. no appendectomy is significant (p (0.005); whereas the ~3” of secondary vs. no appendectomy is nonsignificant. As in primaries, the highest percentage of past secondary appendectomies is found in the college l-4 group. Such a tendency is of special interest. Place of birth-U.S. vs. non-U.S. We introduced this question after our study was in progress, Since the practice of primary appendectomy as a palliative measure is influenced by the attitude of the patient as well as that of the physician, one may hypothesize a lower rate of previous appendectomy in the foreign-born. This is

Appendectomy

TABLE 9.

Total

403

Risk

DISTRIBUTION OF THE TYPE OF APPENDECTOMYBY EDUCATIONAND SEX

Secondary appendectomy (No.) (corr. %)

Primary appendectomy (%) (No.)

Education None, Grammar School, and religious only High School and Trade vocational College l-4 yr College graduate and post-graduate

and Cancer

All appendectomy (%) (No.)

Total number in group

M F

37 33

11.7 14.7

35

3.2 18.2

46 68

14.6 30.2

316 225

M F M F M F

43 57 11 20 28 16

14.4 18.3 13.8 33.9 21.7 23.2

7 47 2 10 1 9

2.7 18.5 2.9 25.6 1.0 17.0

50

104 13 30 29 25

16.8 33.4 16.3 50.8 22.5 36.2

298 311 80 59 129 69

M F

119 126

14.5 19.0

19 101

2.7 18.8

138 227

16.8 34.2

823 664

The xs2 for men and women all and no appendectomy.

are calculated

9

independently

for all educational

divisions

vs.

especially true of those of sufficient age to reflect a fear of hospitalization and surgery. Patients interviewed at Montefiore Hospital are located, for the most part, in the cardiovascular and other non-neoplastic disease category. They represent a mixed socio-economic group the bulk of which is probably middle class. Jews form a high proportion of the cases. Table 10 shows that we cannot demonstrate a significant association between place of birth and appendectomy risk in males. There is, however, an obvious trend for more primary appendectomies in the American-born woman. We note the extremely low rate of 5.1 per cent in the foreign-born female. TABLE 10.

PLACE OF

BIRTH, SEX AND

APPENDECTOMY;

MONTEFIORE

HOSPITAL,

All appendectomy (No.) (%)

1965

No appenTotal dectomy number (No.) in group

Primary appendectomy (No.) (%)

Secondary appendectomy (No.) (corr. %)

Male U.S.A.

(330)

6.7

(7:)

2.4

4

8.9

45

Non-U.S.A.

(647)

8.7

(628.5)

4.8

6

13.0

46

7

7.7

3.6

10

11.0

81

91

16.1

23

32.9

$5)

70

(698.5) 16.1

12

20.3

35

27.1

~Total Female U.S.A.

t545.6)

20.0

Non-U.S.A.

(539.7)

5.1

Total

17

13.2

3 (593.8)

18

16.1

59 94

129

The xla values for men and women are calculated independently for place of birth vs. primary and no appendectomy. Numbers enclosed by parenthesis give the mean age of the group.

404

LYONHYAMS and ERNESTL. WSNDER

These differences in the primary procedure are statistically significant (x:=6.09, p < 0.025). In contrast, secondary appendectomies evidence similarity of observed It is of value to examine this relationship in respect to age and a proportions. possible religious influence. The average present age of a specific group is given within the cells of Table 10. Since the population which was at risk is younger in the U.S.-born, one can argue that a greater proportion of the foreign-born females are now in age category 70-80. In this group, one expects decreased prevalence according to our age-year risk analytic remarks. Our data belies the validity of this hypothesis. Another point to be considered in relation to age is the influence of possible future primary appendectomies arising out of the no appendectomy division. Since the average age of those born in the United States is younger, they are subject to higher future rates, magnifying the observed primary appendectomy rate difference. Similar tabulations were prepared from the information retrieved from The Strang Clinic. The differences in proportions of primary appendectomy for place of birth are nonsignificant in both sexes. This may be related to the relatively small number of foreign-born found in this sample (no.=28). However, a trend toward more appendectomies among native-born was present. This increases our confidence in the association found at Montefiore Hospital. So&-economic S~&LY. There are a number of parameters of socio-economic level. Education, for example, has already been described. We employed two other approaches: Hospital rating dividing the services into wards, semiprivate and private divisions. The other rating, low, middle or high, was given to each person, predominantly through an impression of the patient’s occupation or husband’s occupation. This also included such factors as education, age, religion and number of children. Table 11 presents the distribution of the form of appendectomy by socio-economic group for both sexes. In men, the regressive aspects of the primary procedure with level are apparent. Chi-square testing, however, did not show differences between rates in the low and high levels &;3=0.42, p> 0.05). In women, there is a greater separation of primary and total rates between socioeconomic levels as compared with males. Indeed, there is a significantly TABLE 11. DISTRIBUTION OF

Medium High Total

TYPE

OF APPENDECTOMY

Primary appendectomy (No.) (%)

Socioeconomic status LOW

‘IHE

BY

SOCIO-ECONOMIC

Secondary appendectomy (No.) (cot-r. %)

STATUS

All appendectomy (No.) (%1

AND

SM

Total number in groun

F

39 28

13.1 14.4

10 32

3.9 19.2

49 60

16.5 30.8

297 195

M F M F

68 86 13 11

14.8 20.4 18.3 37.9

7 59 1 5

1.8 17.6 1.7

7.5 145 14

16.4 34.4 19.7

458 421 71

27.8

16

55.2

29

M F

120 125

14.5 19.4

18 96

2.5 18.5

138 221

16.7 34.3

826 645

M

The XT values for men and women are calculated independently all and no appendeotomy.

for low and high status vs.

Appendectomy and Cancer Risk

40s

higher proportion of total appendectomies in the history of patients classified as high &,2=6.71, p < 0.01). This difference is predominantly a function of the discrepancies in the prevalence of the primary procedure. (xl” primary= 10.41, p < 0.005 and x: secondary=2.75, p > 0.05). Hospital and rating. In men, there are nonsignificant differences in total appendectomy prevalence by the hospital ratings: private, semi-private, and ward &I’= 1.09, p > 0.05). Women present highly significant differences in these rates. Their direction and magnitude are proportional to the changes in classification, i.e., they are regressive (x22= 17.01, p < 0.001). Occupation. We classified individuals into specific occupations or groups of occupations. What became obvious from the examination of this data was the great heterogeneity in the prevalence of appendectomy in differing occupations. Sixtyeight laborers, for example, evidence a primary appendectomy rate of 8.8 per cent, while 25 executives showed a rate of 24 per cent. Thirty-four unskilled working women had a previous appendectomy rate of 2.9 per cent, while 33.7 per cent of 261 housewives had undergone such a procedure. Often, but not invariably, such differences seem to be associated with socio-economic factors and age. Other abdominal-pelvic operations. This variable refers to the specific history of any major surgical intervention into the abdominal or pelvic cavities not including the procedure for primary appendectomy, but including the operation type in which incidentals were performed. Frequency distributions of such procedures by form of appendectomy were examined for males and females. We were interested in this parameter to check on the reliability of appendectomy history, to explore the question of which kinds of surgical intervention occasion incidentals and to examine the associations between primary appendectomy and a number of other operative procedures. Of 708 men without other reported abdominal or pelvic operations, 13.4 per cent had a primary appendectomy. Of the 107 men who had at least one other procedure of this sort, 20.6 per cent also underwent appendectomy. This difference in observed proportions is statistically significant (xl”=3.86. p < 0.05). The female association between a history of primary appendectomy and an increased number of other abdominal pelvic operations is more dramatic. Of 441 women without other abdominal operative history, 18.6 per cent have undergone primary appendectomy. Yet 38.6 per cent of the 143 women with other abdominal This difference in proportions is surgery have also experienced appendectomy. highly significant (x:=23.7, p < 0.001). Individuals then, of both sexes, who have had surgical removal for appendicitis have a greater number of other abdominal-pelvic procedures. We do not evaluate the temporal relationship of these various operations to primary appendectomy, but appendicial removal will generally antecede the rest. In addition, we found that women with large bowel and breast cancer have undergone a significantly greater number of previous abdominal operations than the female combined control group (x:=6.30, p < 0.025). Smoking.

Various indices of the smoking habit were examined with respect to

LYON HYAMS and ERNEST L. WYNDER

406

appendectomy history. We could not find any consistent or significant trends in these measures. Weight. Percentage deviation from normal weight is calculated for patients and controls between the ages of 40-50 by the following formulation: observed Percentage

over- or underweight

weight -

expected

weight

=

X 100 expected

weight

This is obviously a more meaningful index than the absolute number of pounds under- or overweight. A condensed distribution of percentage deviation from normal weight in men and women in relation to appendectomy status was evaluated for differences. These distributions were remarkably similar, and we accepted the proposition that no relationship exists between these two factors. Bowel habits. We attempted to analyze three manifestations of bowel function: laxation, constipation and a history of hemorrhoids. Constipation seemed of interest because of its possible relation to impacted appendicial fecaliths and hemorrhoids because of secondary associations with constipation and increased intra-abdominal pressures. Somewhat intricate though reasonable indices of these measures were established. They were tested in their relation to past appendectomy. No associations could be found between any of these measures and appendectomy status. Summary of epidemiological results

A summary of significant associations is given in Table 12. By virtue of the proportion of tests ruled significant and the magnitude of the probabilistic levels we concluded that women are more sensitive to epidemiologic influences then are men. TABLE 12.

ASSXIATION OF VARIOUS EP~DEMIOLUGICAL FACTORSWITH A HISTORYOF PREVIOUS TOTAL

APPENDECTOMY

Male*

Factors Race Religion Place of Birthfor primary procedures only Education Socio-economic status Hospital status Abdominal operations Weight Smoking habits Bowel Habits

Probability level

Significance S

NS

0.025 > p > 0.01 0.1 >p>o.os

p > 0.5

NS NS NS NS

S NS

NS NS

Female

0.5

> p> p> 0.50 > p > 0.05 > p > p>

0.25 0.5 0.25 0.02 0.5 Various parameters conN o n e showed trasted. significance. Various parameters contrasted. None showed significance.

Significance NS

Probability level

NS

0.25 >p>O.l 0.1 > p > 0.05

S

0.025 > p > 0.01

S

0.05 > p > 0.02 0.01 > p > 0.005

S S S NS NS

NS

0.001 > p 0.001 > p

0.5 > p > 0.25 Various parameters contrasted. N one showed significance Various parameters contrasted. N one showed significance

*We must remind the reader that a verdict of nonsignificance does not mean that differences do not exist, only that we cannot demonstrate them. It is conceivable, that given a larger sample of males, more contrasts would have been found statistically significant.

Appendectomy

and Cancer

Risk

407

The fact that the described relationships were generally of the same type, viewing cancer and non-cancers separately, strengthens the confidence we have in the epidemiological effect. Finally our impression is that those factors that relate to appendicial removal are, with some exceptions, mainly involved with the procedure of primary appendectomy. Epidemiological

influence on appendectomy

rate diflerences

Ideally, one should simultaneously adjust for the pertinent epidemiological variables in the comparison of the proportions of appendectomy prevalence between cancer and control groups. This appears, however, unreasonable. Even when separating factors such as specific disease state, sex, age, and educational level, one is inhibited from drawing any conclusions from the subsequent test because of the relatively small number within each joint classification. We were forced to view the distribution of these epidemiological factors among specific and select diagnostic groups to determine whether the differences in rates between groups bear a consistent relationship to these variables. Do the breast cancer and colon-rectal cancer cases have a greater proportion of people from higher socioeconomic groups and better education than the controls? If they do, the argument of a direct association between cancer and previous appendectomy is weakened. If not, it is strengthened. The analysis of the data along these lines did not consistently support the epidemiological explanation for high appendectomy rates in the female breast and large bowel cancers. Another approach to this information is to view appendectomy prevalence in specific groups stratified for the pertinent epidemiological factors given. This was done using the combined control group as one division, large bowel and breast cancer as the other. In both men and women there is some Again, the results were inconclusive. For continuance of factor appendectomy trends between diagnostic groupings. example, women with breast cancer show the same increase of appendectomy rate with educational gradient as do colon-rectal cancers and controls. In other categories, however, such as socio-economic status and hospital accommodation, the control trends are opposed or are not manifested in the cancer group. This is obviously an equivocal finding and we cannot explain it fully. Some of the confusion may be due to the special characteristics of our breast cancer patients who are drawn primarily from the Memorial Cancer Center. Age at which appendectomy is performed. Obviously, the variable, time between appendectomy and present disease, is a function of the age of appendectomy and the age at which the present disease process was diagnosed. As such, it is less suitable for direct analysis than the age of appendectomy alone. Comparisons between groups in which age of appendectomy is of interest must be adjusted for present age in order to have meaning, since the older the diagnostic group, the older is the mean age of appendectomy. The converse is also true. An additional advantage when standardizing for present age is that such arrangement simultaneously compares the average time between appendectomy and present disease. Figures 1 and 2 present such contrasts between cancer and control groups in the

408

LYON HYAMS and ERNEST L. WYNDER

All controls All cancers lymphomas

0

excluding and leukemias

I

I

1

1

I

I

I

M

30

40

50

60

70

80

Age

of diagnosis

FIG. 1. Relationship between age of diagnosis (or present age) and age at which appendectomy was performed in men. The lines are visually fitted estimates. The spots represent average values for the cell; the numbers adjunct to them indicate the number of observations included in this calculation.

form of graphs. A visual fit is drawn to the suggested pattern to aid in interpretation. Statistical tests (analysis of co-variance) were employed in the .evaluation of the female data. In males, if any differences exist in the age at which appendectomy is performed, it is not consistent over the present age (Fig. 1). This is a finding which is physiologically uninterpretable. In women, statistical parallelism exists (Fig. 2, t < 1). The cancer group reveals a trend toward appendectomy at a consistently later age than do their “age matched” controls. However, an analysis of covariance did not show significant differences in the mean age of appendectomy between cancer and control groups (t < 1). Breast cancers and large bowel cancers produce the same result when contrasted to the control. They are obviously parallel to the mixed cancer group. The lymphomas and the leukemias We report on the lymphomas and leukemias separately.

Not only do they represent patients from a younger age classification, but their mesodermal origin and their usually more diffuse manifestation segregates them. When it became apparent that we lacked a suitable age matched contrast group to the younger male Hodgkin’s disease patients, we collected appropriate information from private and public high school students, junior college students, and college men and women in New York City and California. This was accomplished by

Appendectomy

-

and Cancer

Risk

409

All controls All cancers excluding lymphomas and leukemias o Breast cancer A Colon-rectal cancers l

o

-

Xl

I

I

1

I

I

I

30

40

50

60

70

80

Age

of diagnosis

FIG. 2. Relationship between age of diagnosis (or present age) and age at which appendectomy was performed in women for the ind,icated diagnostic categories. The lines are visually fitted estimates. The spots represent average values for the cell; the numbers adjunct to them indicate the number of observations included in this calculation.

handing out self-administering questionnaires to a total of 700 students in various assemblies. Appropriate instructions and supervision were mediated through a representative of our study group. The control material was added to individuals free of disease in the previous control group for contrasting purposes. Crude appendectomy rates for the main diagnostic subdivisions are given in Table 13. The male Hodgkin’s patients appear to have an unusually high rate of primary appendectomy, especially when one considers their relatively younger ages. The age-specific rates for these patients and a group of healthy control men are TABLE

13.

DISTRIBLJTIONOFAPPENDECTOMIESAMONGVARIO~S BY SEX

Primary appendectomy (%) (No.)

Diagnosis Hodgin’s

Lymphosarcoma

M

Reticulum cell sarcoma Leukemias, chrome and acute

F M F M F

22 3 3 8 4 3 2 2

M F

31 16

Total

disease

M F

22.0 9.1 8.8 16.0 16.0 16.7 9.5 12.5 17.2 13.7

LYMPHOMA

Secondary appendectomy (No.) (corr. %) 1

:::

t! 2 0

ANDLEUKEMI~

All appendectomy (No.) (%)

GROUPS

Total number in group

; 2

0.0 4.8 0.0 20.0 0.0 14.3

23 4 3 10 4 6 2 4

23.0 12.1 8.8 20.0 16.0 33.3 2;:;

100 33 34 50 25 18 21 16

1 8

0.7 7.9

32 24

17.8 20.5

180 117

410

LYON HYAMS and ERNEST L. WYNDER

TABLE 14.

PREVIOUS APPENDECTOMY Hodgkin’s

Age of diagnosis or present age

BY AGE IN MALE coNrIloLs*

disease

previous appendectomy (%) (No.)

lo- 15 16-20 21-30 3lf

: 8 9

HODGKIN'S

DISEASE AND

HEALTHY

Control No. in previous appendectomy (No.) (%) group

16.7 18.5 28.6 23.1

6 27 28 39

5 39 6 12

*These controls include the questionnaire samples (young healthy appropriate members of healthy interviewed subjects.

2.2 8.6 14.3 12.4 students),

No. in group 225 451 42 97

as well as age

presented in Table 14. There is an obvious trend for men with Hodgin’s disease to have more appendectomies. Statistical testing yields a significant difference (x;=8.96, 0.001 > p > .005). These men with Hodgkin’s disease were drawn from the low to medium socioeconomic level. The percentage who have had some college education is 12.3. Their religious distribution is not weighted to any particular denomination. We also examined the relationships between the age of appendectomy and the age of diagnosis in those individuals who underwent the procedure. The raw data is presented in Table 15. Various approaches to analysis were utilized, but we could not demonstrate a significant difference between Hodgkin’s and control groups. There is some tendency for men with Hodgkin’s disease to have their appendices removed at an earlier age. However, it would take a great many more cases to conmm this. The proportion of women with Hodgkin’s disease and appendectomy did not seem suggestively large. However, we cannot make a definite statement because of an insufficient number of cases. TABLE

15.

Hodgkin’s

AGE

OF DUGNOSIS

AND AGE OF APPENDECTOMY HEALTHY CONTROLS

(17,13) (21.6)

(16912) (19,4) (22,13)

(23,11) (32,4)

::1?91,“’

(24 3) (33916)

(38, 18)

(38, 18)

*(Age of diagnosis

HODGKIN'S DISEASE AND

Controls

disease

(14, 12)’ (17,7) (2297)

(3598) (54220)

IN MALE

(16,ff) (2197)

or present

(26,12) (34,24)

(14,lO)

(156)

(15,12)

(15, 10)

(16915) (1636) (1637) (1797) (1799) (17,lO) (18,lO) (18,lO) (20912) (24,12) (34928) (3694) (3837)

(1697) (16,12) (1611) (17,12) (17, 12) (17, 10) (18, 12) (18916) (20918) (25,9) (34912) (38,16) (39, 12)

(169) (16913)

(1693) (16, 10)

(17, 10) (1798) (17,16) (18,fO)

(1777) (1775)

(20, 13)

(21, 17)

(34,18) (37,26)

(39913)

age, age of appendectomy.)

08, 12)

Appendectomy

and Cancer Risk

411

DISCUSSION

The possible functional role of the vermiform appendix, especially in relation to its lymphoid tissue, has aroused recent interest. It has been called the homologue to the, Bursae of Fabricius in chickens and, as such, might be expected to play an important role in the immune-globulin defense system [15]. This has direct bearing on certain theories in carcinogenesis and has provided the theoretical frame for MCVAY’S study [I]. Indeed, if cancer in man can be viewed as the relative failure of such a defense and, if the appendix forms some vital subdivision of this system, one might expect an increased risk of cancer with a history of previous appendectomy. It was the purpose of the present study to re-examine this relationship between appendectomy prevalence and cancer in man within the epidemiological framework. Factors in this report that weaken the hypothesized relationship include: (1) Our inability to demonstrate this relationship with men in any of the contrasts, except for Hodgkin’s disease (which is discussed later), while they existed in women only for breast and large bowel cancer. It is unlikely that the appendix exerts some protective characteristic in female large bowel cancers but does not do so with men. (2) We observed differences in .outcome as a result of changing control groups in the pertinent comparison. One must ask the question: “Is a given cancer rate high or a given control rate low?” (3) Especially primary appendectomies in females are related to epidemiological variables which, in many previous reports, were not adequately controlled. Men, who are less affected by outside factors, did not show differences between cancer and control groups. (4) Women with cancer neither have their appendices out at an earlier nor at a later age than the controls. Two points of view can be advanced in regard to the time of appendectomy. GROSS [3] feels that late appendectomy, for example, is a final insult to an already failing immunological capacity. On the other hand, BIERMAN[2] points out that the appendicular tissue atrophies with age, and that one cannot expect deprivation in later life to be important. (5) The great discrepancies of appendectomy prevalence among the various studies suggest that methodological limitation and the peculiarities of specific groups overshadow the proposed relationship (Table 1). The smaller intra-study variation between various diagnostic groups in HOWIE’S British study [4]. accompanied by relatively small observed proportions of appendectomy, supports this statement. Furthermore, the figures given of 8.0 per cent and 6.0 per cent for previous appendectomy in female and male colon cancer patients respectively, seriously detract from the proposed relationship. It is of interest that breast cancers in women and large bowel cancers in both sexes manifest relatively high rates of appendectomy. We were not able to indicate that these proportions correspond to epidemiological status. Whereas the significantly higher proportions of the appendectomy prevalence in women with large bowel and breast cancers seem to be more strongly related to the incidental operation than to the primary procedure, epidemiological factors, on the other hand,

412

LYONHYAMS and ERNESTL. WYNDER

relate more strongly to primary appendectomy. However, women with breast and large bowel cancers have more abdominal operations than the controls, and are, thus, more likely to have had an incidental appendectomy. Due to our initial omission in obtaining the place of birth of some patients, it is possible we are omitting a pertinent criterion in explaining differential rates. Breast and intestinal cancers are more common among U.S.-born than certain foreign-born populations [ 161. Consequently, American women would be subject to both higher rates of breast and large bowel cancers and past appendectomies. Socio-economic status may function as a common link. It is also hard to accept the very specific and unique defense role that the appendicial lymphoid apparatus must play in order to protect only against large bowel and breast carcinoma in women and not against a mixed group of neoplasms. Until reasonable physiological explanations of these very specific relationships are advanced, such findings must be interpreted with great caution. One must also consider the possibility that since the diagnosis of breast and large bowel cancers are frequently followed by or are concomitant with abdominal surgery, the simultaneous appendectomy may be referred to an earlier procedure and serve as an important source of confusion. One would hope that primary appendectomy is invariably related to appendicitis. However, the differential diagnosis of acute appendicitis is difficult, and a number of other entities mimic it in such as pelvic inflammatory disease, spastic colon, mesenteric lymphadenitis, and pancreatitis [ 17, 181. If surgery is elected, the organ is invariably removed even though pathological changes in the appendix are not macroscopically present. T’hat many histologically normal appendices are removed is common knowledge. Various studies give estimates of its magnitude. Two such reported estimates are 16 per cent [ 171 and 33 per cent ,[19]. Such unnecessary appendectomy seems to be more common in women [lo], and the sex-specific age distribution of appendicitis shows an obvious female-male excess during the ages of 15-24 yr [9]. Our data confirms a generally higher prevalence of primary appendectomy in women. It seems reasonable to infer that these represent an excess explained by the absence of pathological confirmation. There is little evidence that true appendicitis is more frequent in women [20]. Lee has also shown a relationship between social circumstance and risk for appendectomy in men [8]. Meyer et al. have related false appendectomy to women with behavioral disorders, high socio-economic class, and subsequent sickness [lo]. Operative rates were higher in those women with normal histology than in those whose sections showed some form of appendicitis. The authors suggest that right lower quadrant pain, fever, and leukocytosis may be a conversion hysterical form. Our result supports these relationships. Males, with respect to risk for primary appendectomy, are less influenced by the measured epidemiological parameters than females. There is probably a bulk of females with psychosomatic overtones whose symptomatology mimics real disease but who are, in addition, relatively anxious to receive the attention and trauma of surgery. Considering the fact that the bulk of primary appendectomies occurred more than 20 yr ago, such women must have had the social and financial facility to avail themselves of surgery and hospitalization.

Appendectomy

and Cancer Risk

413

On the other hand, there is a group of men, undoubtedly stoic either by nature or environmental demand, who, even in the presence of disconcerting pain, will continue to work and will resist medical intervention. One doubts whether this type of response in men is limited to lower socio-economic groups. It can prevail in the higher levels as well. However, the presentation of very low primary appendectomy rates in Negroes could well be related to the relative inaccessibility of medical service. One must also speculate on the role of the physician and/or surgeon upon whose decision the outcome predominantly depends. Would he be influenced by a patient’s poor financial status and then be more apt to be conservative in his therapy of lower abdominal complaint. 3 Consider, as well, the resistance of a woman or man from a low socio-economic group to consulting a doctor and accepting the possibility of hospitalization when their presence in the family situation would be so acutely missed. Such a person is not only more apt to put aside hysterical manifestation, but might postpone intervention in true appendicitis and, thus, afford themselves the chance of the regression of the disease process. Background variables may influence such decisions in another even more obvious way. A person born and spending his youth and early adult life outside the United States is less apt to have any operation performed, much less an sppendectomy. This would be especially true 30-40 yr ago when the foreign-born older group of patients were at prime appendicitis age. Obviously, such a group will manifest This is due not only to a lack of earlier lower rates of primary appendectomy. opportunity, but also to the negative attitude that these people often have toward medical services, especially hospitalization and operation. Such attitudes would probably continue to exist when moving to the U.S. Our partial analysis by country of origin supports this contention. Thus, the risk for appendectomy is a complex net of interweaving epidemiological factors which should be taken into account in any evaluation of rate difference. This remark is especially true for females, but may play a less extensive role in males. However, we cannot adequately interpret the significantly greater prevalence of appendectomy in men with Hodgkin’s disease. Many of these procedures were performed at an age when epidemiological factors would not exert a direct influence. Although it is possible that parental attitudes could play a role in early appendectomy, it cannot be demonstrated, and we suspect it is not of real consequence. This finding obviously deserves further attention and verification using different patient and control groups. However, if such appendectomy relates directly to Hodgkin’s disease, it is not a major factor considering the proportion of men with Hodgkin’s disease but without appendectomy. Future studies

In view of the many concomitant variables, we suggest that in future studies the cancer and control cases be matched for the pertinent factors in the design: age, sex, place of birth, and reasonable parameters of socio-economic status. Further, in the years to come the effect of health insurance on surgery will manifest itself and must be considered. Such studies must especially concern themselves with the possible sources of error in the reportage of secondary appendectomy.

414

LYON HYhMS and ERNEST L. WYNDER

It can also be suggested that since continental rates are probably lower, they may be less influenced by epidemiological factors. Western Europe may, thus, provide a natural area of investigation. SUMMARY

One thousand five hundred and eighty-two subjects with and without cancer were interviewed. An additional seven hundred young, healthy persons completed a self-administering questionnaire. Comparisons of appendectomy prevalence in various diagnostic subdivisions and condensations were made. Subsequently, the procedure of appendectomy was epidemiologically explored, and an attempt was made to relate initial findings of the risk of cancer with appendectomy to appropriate concomitant epidemiological factors. Men with Hodgkin’s disease evidence, in this material, significantly higher rates of past appendectomy than the appropriate control group. We could not demonstrate any other statistically significant differences in appendectomy prevalence in men between diverse combinations of diagnostic groupings. In women, large bowel cancers and breast cancers showed significantly higher proportions when contrasted to certain controls but not to others. These differences were mainly related to the incidental procedure and are considered coincidental findings. A history of past appendectomy, especially of the primary variety, is linked to a number of epidemiological facets: race, place of birth, education, and socioeconomic status. These tendencies are much more obvious in women than in men. It was impossible to control for all of these factors in a final analysis of appendectomy rates between pertinent cancer and control groups. However, an exploration and discussion of the findings is presented and a future study of a more limited and controlled scope is suggested. Acknowledgements-We would like to express our appreciation to Mr. MO KATZ, Administrator of Montefiore Hospital and Dr. EMERSONDAY, Director of Preventive Medicine InstituteStrang Clinic, for their co-operation. We would also like to thank our research assistants, Mr. BERT HERMAN and Miss MARGARETSMITH,for their technical assistance in this study.

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

REFERENCES MCVAY JR., J. R.: The appendix in relation to neoplastic disease, Cancer 17, 929-937, 1964. BIERMAN,H. R. : Human appendix and neoplasia, Cancer 21, 109-l18, 1968. GROSS, L.: Inoidence of appendectomies and tonsillectomies in cancer patients, Cancer 19, 849-852, 1966. HOWIE, J. G. R. and TIMPERLEY,W. R. : Cancer and appendectomy, Cancer 19, 11381142, 1966. KESSLER, I. I.: Lymphoid tissue neoplasia study. Reported at Cornell University Medical College, New York City, 1965. ASSOCIATEDHOSPITALSERVICE: Unpublished data, 1965. LARSEN, N. P. : A twenty-six year study of appendectomies, Hawaii Med. 1. 22, 112115, 1962. LEE, J. A. H.: _4n association between social circumstances and appendicitis in young people, Br. med. 1. 1, 1217-1219, 1957. LEE, J. A. H.: “Appendicitis” in young women. An opportunity for collaborative clinical research in the National Health Service, Lancer ii, 815-817, 1961. MEYER,E., UNGER,H. T. and SLAUGHTER,R. : Investigation of a psychosocial hypothesis in appendectomies, Psychosom. Med. 26, 67lS681, 1964. FERTIG,J. : Personal communication.

Appendectomy 12.

13. 14. 15. 16.

18.

Risk

415

KINCAID,W. hl. : The combination of 2 X m contingency tables, Biometrics 18, 224-228, 1962. MANTEL, N. and HAENSZEL, W.: Statistical aspects of the analysis of data from retrospective studies of disease, J. natn. Catmer Inst. 22, 719-748, 1959. CORNFIELD,J. : Personal communication. ARCHER,0. K., SUTHERLAND,D. E. R. and GOOD, R. A. : Appendix of the rabbit: A homologue of the bursa in the chicken, Nature ZOO: 337-339, 1963. IIAENSZEL, W.: Cancer mortality among the foreign-born in the United States, J. narn. Cancer

17.

and Cancer

Inst. 26: 37-132,

1961.

COLLINS,D. C.: 71,000 human appendix specimens. A final report, summarizing forty years’ study. Am. 1. Protocol. 14, 365-381, 1963. SMITH, F. W.: Surgical significance of spastic colon. 1. mied. Ass., Alabama 28: 373377, 1959.

19. 20.

HOWIE,J. G. R.: Too few appendectomies, Lancer i, 124&1242, 1964. LEE, J. A. H.: The influence of sex and age on appendicitis in children adults, Gut 3, 8tH4, 1%2.

and young