Surrogate Anatomic/Functional Sites for Evaluating Cancer Risk: An Extension of the Field Effect

The goal of cancer research is to reduce the morbidity and mortality associated with this disease. This end point can be achieved either by cure of existing cancer or prevention of new cases of cancer. The enormity of the problem is evident in recent estimates, which suggest that in 1999 about 1,221,800 new cases of invasive cancer will be diagnosed in the United States, with 563,100 deaths resulting from this disease. Although there is an apparent overall downward trend in cancer incidence rates and some progress has been achieved in the realm of treatment of existing cancer, advances in the management of the most common forms of epithelial cancer (lung, colon, breast, and prostate) have been slower than hoped for. An alternative approach therefore seems reasonable. The preventive strategies that have been so successfully implemented in cardiovascular disease offer a model for such an alternative approach, i.e., a greater emphasis on research directed at the prevention of cancer.

Cancer progression is the result of a multistep process in which the cumulative effect of successive discrete genetic alterations leads to a gradual transition from normal through premalignant to frankly malignant tissue and ultimately to metastasis. This multistep process, which is currently espoused as the underlying mechanism of carcinogenesis, lends itself to prevention interventions aimed at halting the progression or stimulating the regression of existing premalignant lesions. Methods for preventing cancer draw on this multistep model directly by relying on the ability to intervene in tumorigenic progression at some point prior to the transition from a premalignant lesion to a fully evolved invasive cancer.

Chemoprevention is prominent among the prevention interventions used for arresting the carcinogenic process. The chemopreventive approach to prevention involves the administration of a drug, or chemical agent, to individuals designated as being at high, or increased, risk for a given cancer. A key to the successful design of a chemoprevention study is to distinguish ostensibly “healthy” individuals who are at an increased risk for cancer from those who are at an average risk for the disease. Because chemopreventive interventions often target the former population, methods to identify potential subjects based on risk evaluation addressing cancer in a given target organ(s) must be developed. The traditional methods that target “at-risk” groups by addressing such characteristics as family history of cancer, exogenous carcinogen exposure (e.g., smoking and occupational chemical exposures), and dietary history could benefit from the incorporation of a more precise approach to defining individuals at increased risk for organ-specific cancers.

In this editorial, we propose the hypothesis that cancer risk in specific target organs can be determined through genetic changes that occur in accessible surrogate anatomical/functional sites, suggesting the existence of an “extended field effect.” We define a surrogate organ site as a site that mirrors changes or predicts cancer risk in a target organ or tissue with which the surrogate site is contiguous or noncontiguous. Because surrogate organs may be more accessible than the target organ(s), sampling the surrogate organs as part of a “field” is likely to provide important clinical information on cancer risk in the target organ(s).

In the past, the term “field effect” has been used to explain the increase in cancer incidence in a continuous field of epithelium that may include only one site or contiguous sites (1). For example, the development of primary head and neck cancer predicts a greater risk of clinical cancer occurring in the contiguous epithelium of the respiratory tract or esophagus (1, 2). In another example, the presence of adenomas or carcinoma of the colon may indicate a higher probability of a metachronous cancer developing in the adjacent colonic epithelium (3, 4).

Here, we extend the conventional use of the term “field effect” in two major ways:

(a) In contrast to its usual application (1), the field effect in the context of this commentary should not necessarily be considered as a field “cancerization” effect wherein the ultimate expression in the surrogate organ sites is cancer. The surrogate organ may indicate cancer risk in the target organ(s), but the surrogate organ itself may or may not be at risk for cancer. Thus, we extend the notion of a field, comprising contiguous organs affected by a common insult, to the clinically asymptomatic or genetic setting.

(b) We broaden the notion of a “field effect” from a purely anatomical concept to a more general, functional level, i.e., extending the idea of a field effect beyond physically contiguous organs. Thus, cells of noncontiguous organs that are subject to similar molecular controls may constitute a physiological or “functional field,” in that they may experience similar cellular changes in response to a systemic insult.

We use “field effect” to denote the sharing, or systemic expression, of abnormal biomarkers in both the specific target organs and the more readily accessible surrogate organ sites. These abnormal biomarkers, whether at the genetic or at a higher phenotypic level, may or may not themselves be precursors of cancer. When present in the surrogate site, these abnormal biomarkers may indicate a higher risk for cancer in the less accessible target organ. Such biomarkers may be found long before histologically evident cancer develops in the target organ (5, 6). Once a field effect has been identified, one is able to estimate the likelihood of a subsequent primary cancer in one component of the field by identifying a surrogate marker in another component of the field. The lead time provided by the identification of such a field effect allows the timely institution of appropriate chemopreventive or early detection measures (7).

This broader definition of the field effect is best illustrated as an experimental paradigm by the dominantly inherited cancer syndromes (8, 9). For example, cancer-associated phenotypic or genotypic expressions can be seen in clinically normal cells obtained from colorectal and breast cancer patients and from a portion of their clinically asymptomatic children (8, 9, 10, 11, 12, 13, 14). In this context, “clinically normal cells” constitute the “surrogate site,” which itself is not at increased risk for the development of cancer, whereas colon and breast compose the “target organs” at increased risk. Such cancer-associated expressions have been attributed to specific gene mutations in the heritable cancers (8, 9, 10, 11, 12, 13, 14). The sampling of surrogate sites for genetic or phenotypic markers that portend an increased risk for cancer at a second, target, site is implicit in the clinical diagnosis of such inherited cancer syndromes.

The evaluation of risk for sporadic cancers, however, poses a considerably greater challenge. This difficulty is encountered particularly in target organs that are not readily accessible. The lack of suspicion from a positive family history, the absence of a syndrome, or constellation of known coexisting clinical changes in accessible organs contribute to this obstacle. For instance, sampling the skin or the oral cavity can be a routine procedure, whereas sampling the lung or ovary can be both logistically difficult and costly. Thus, assessing the cancer risk in the less accessible organ sites by screening for genetic alterations in accessible surrogate organ sites might well change current approaches to early detection and thereby would greatly facilitate recruitment of appropriate increased-risk candidates for chemoprevention trials.

In accord with our description of an “extended field effect,” cancer risk in the less accessible target organs can be evaluated by determining the genetic changes that occur in the readily accessible, surrogate organ sites with which the less accessible target organs form an anatomical and/or functional continuum. An example in which an anatomical continuum provides the basis for the field effect is the upper aerodigestive system. As applied to this example, the specific hypothesis is that genetic changes detected in the pharynx predict cancer risk in the lung. The field effect in this case derives from a common exposure to cigarette smoke resulting from the geographic proximity of the anatomically contiguous surrogate, accessible organs, and the less accessible target organs. An example of a functional continuum serving as the basis for the field effect is seen in the prostate gland and the hair follicles of the scalp. The specific hypothesis in this example is that genetic changes that are mediated by androgenic receptors in the hair follicles reflect genetic changes that are mediated by androgenic receptors in the prostate gland.

Appropriate cohorts must be defined for application of the proposed surrogate site approach. At the present time, individuals with certain histological risk lesions in target organs of interest (e.g., dysplasia) are treated effectively by chemopreventive agents (15, 16). As a result, the direct clinical identification of such lesions in the target site a priori precludes application of the surrogate site approach.

In summary, the objective of this editorial is to introduce the hypothesis that the risk of developing cancer in specific target organs can often be estimated by determining the genetic changes occurring in readily accessible anatomical/functional sites (surrogate organ sites). Examples of surrogate organ sites for six cancers, lung, ovary, esophagus, bladder, breast, and prostate, are considered. Enough information exists with regard to these organs to use them as models for demonstrating the concept of surrogate organ sites. In the case of other major cancers present in different organ sites such as colon, documentation of risk fields associated with these sites has not yet been developed to the point of allowing their inclusion as examples in this discussion. Among the examples selected, some are more amenable than others to hypothesis testing in the immediate future. Thus, the extended surrogate field encompassing the lung, head and neck, esophagus, and bladder is already well established. On the other hand, the notion of a comparable extended field relating to the breast or prostate has been included to illustrate the biological and clinical potential of this field approach to multiple systems.

In most cases, cancer of the upper aerodigestive tract results from cigarette smoking (17). The smoke makes contact with the mucosa that lines the oropharynx, nasopharynx, larynx, and bronchial tree. Because smoking exposes the mucosa of both the head and neck and the bronchial tree to similar carcinogens, similar genetic changes are likely to occur all along the epithelium, despite potential differences in metabolism between the two tissues.

Patients with primary head and neck cancer exhibit an increased incidence of second metachronous or synchronous tumors of the head and neck as a result of a field change (1, 18, 19, 20, 21). Approximately 15% of patients with head and neck cancer will develop second primary cancers in the head and neck region (22). Furthermore, an increased incidence of second primary cancers of the head and neck occurs not only in patients with head and neck cancer but also in patients with lung cancer (2). This dual association of second primary head and neck cancers is substantiated by epidemiological studies. In fact, the association has been found to be reciprocal. According to the SEER² Program of the National Cancer Institute, patients presenting with lung cancer have a statistically significant 2.5 observed/expected ratio for second primary cancers in the head and neck, and patients with head and neck cancer have a statistically significant 3.4 observed/expected ratio of a subsequent lung cancer. Of note, these data do not distinguish between smokers and nonsmokers.

Focusing on smokers with cancer, genetic changes are found in the histologically normal mucosa of the upper aerodigestive tract in cases of head and neck cancer and in the bronchial mucosa in cases of lung cancer (23, 24, 25, 26, 27, 28, 29). In addition, the smoking-related molecular changes at these contiguous sites have many similarities, with deletions of chromosomes 3p and 9p being early events at both sites, and mutations of the p53 gene being late events (27, 29, 30, 31, 32). To date, however, no correlation has been established between the occurrence of genetic changes in the upper aerodigestive tract and their occurrence in the lung. Nevertheless, the finding of such genetic changes in the head and neck is anticipated in association with lung cancer based on the higher than observed/expected ratio noted above for second primary tumors of the upper aerodigestive tract in cases of lung cancer.

If the hypothesis that related genetic changes occur all along the aerodigestive tract as a result of smoking is correct, then the oral and nasopharyngeal mucosa may serve as surrogate anatomical sites for genetic changes that occur in the lung prior to the development of invasive lung cancer. In this case, the surrogate risk markers are mechanistically related, via a field effect, to the target site for which they represent risk. Both the extent and type of these putative genetic changes, although not necessarily identical in the two organs, could function as an early warning signal for lung cancer, thereby providing justification for chemoprevention intervention ipso facto.

Smokers and ex-smokers, as well as patients with treated head and neck, esophageal, and lung cancer, could be tested for genetic changes by scraping epithelial cells from the pharyngeal mucosa. Individuals with genetic changes would be considered at higher risk than that imparted by their disease history and/or tobacco history alone. This additional risk would be especially pertinent if the nature of the suggestive changes resembled genetic alterations that are known to be associated with lung cancer. These individuals, therefore, would become candidates for chemoprevention for both head and neck as well as lung cancer.

As with the lung, the mucosa of the head and neck may also serve as a surrogate anatomical site for squamous cell carcinomas of the esophagus, especially in cases of smoking and alcohol abuse. According to the SEER Program, patients with head and neck cancer have a statistically significant observed/expected ratio of 15.9 of second primary cancers in the esophagus, and patients presenting with cancer of the esophagus have a statistically significant observed/expected ratio of 7.8 of second primary cancers of the head and neck. These data do not distinguish between smokers and nonsmokers.

As a result of common etiology, the urinary bladder may also be associated with the same field as the upper aerodigestive tract. Smoking has been implicated as a cause of bladder cancer. Thus, SEER Program data show that patients with primary bladder cancer have a statistically significant 1.5-fold increase in second cancers of the lung when compared with the general population. The converse has not been observed, however, because patients with lung cancer as a first cancer do not have a higher incidence of bladder cancer. A suggested reason for this is that lung cancer patients die prior to the opportunity for a bladder cancer diagnosis. Such an association based on common etiology (tobacco exposure) raises the question of whether specific molecular genetic changes occurring in the accessible oral cavity have the potential for identifying those patients who are at high risk for bladder cancer.

Most ovarian cancers arise from that portion of the peritoneum that covers the capsular surface of the ovary (surface epithelium). Yet, there are clinical data to suggest that ovarian cancer is also associated with a field effect throughout the peritoneum (33). Ovarian cancer can be bilateral, suggesting that a field effect exists in the surface epithelium of both ovaries. In women with bilateral oophorectomy, cancer can still develop in the surrounding peritoneum (34). Furthermore, a subset of women exists who present with extraovarian, intraabdominal carcinomatosis that is histologically similar to ovarian cancer (35, 36). Finally, some ovarian tumors arise synchronously on the surface of the ovary and in the adjacent peritoneum (34, 37, 38). Thus, the peritoneum may also be the site of origin of epithelial cancers. Accordingly, the application of the risk assessment hypothesis to ovarian cancer would entail the determination of whether those genetic changes known to be associated with ovarian cancer also occur commonly as part of a field change within the peritoneum surrounding the ovary. If the hypothesis is correct, then the adjacent peritoneum, even if normal in appearance, becomes a surrogate organ site for ovarian cancer. The peritoneum exhibits the requisite accessibility of a surrogate site in that it can be sampled invasively through the pouch of Douglas by scraping the cells that cover the pouch, albeit with some risk of peritonitis (39, 40, 41).

In addition, several lines of evidence point to the endometrium as part of a “field” that would allow it to serve as an alternative surrogate organ site for ovarian cancer. Histological similarity exists between at least one pathological subtype of ovarian cancer, endometrioid carcinoma, and certain tumors that arise from the Mullerian duct (fallopian tubes, cervix, and upper vagina; Ref. 42). Additional support for a field effect encompassing both ovary and endometrium comes from the many reports of synchronous and metachronous ovarian and endometrial cancers. In some cases, these multiple tumors have been suggested by survival data and molecular analyses as constituting independent primary lesions (43, 44, 45, 46). Furthermore, certain endometrioid ovarian cancers share molecular markers with endometriosis (47), a common gynecological condition in which normal-appearing endometrial tissue proliferates ectopically, outside the uterine cavity. Although the malignant potential of endometriosis is very low, a transition from benign to malignant has occasionally been demonstrated. In some cases, the malignant transformation has been shown by histological observations and molecular studies to involve the development of ovarian carcinomas of the endometrioid or clear cell type from lesions of endometriosis (47, 48). Taken together, these observations support the existence of a “field effect” encompassing the ovaries and the endometrium, which is reflected in multifocal tumorigenesis in the upper genital tract and the ovaries (44). Thus, in addition to the pelvic peritoneum, the more accessible endometrium may also serve as a surrogate organ for the ovary.

The value of appropriate surrogate site (peritoneum or endometrium) testing for incipient genetic changes in ovarian cancer, especially in women at increased risk, cannot be overestimated, because effective screening tools for this disease do not exist.

Several studies suggest that the entire mammary epithelium is at increased risk for malignant transformation in breast cancer patients (49). Included among these are studies of inherited susceptibility, generalized genomic instability, increased risk of subsequent cancer in both breasts in the setting of biopsy-proven atypical ductal hyperplasia, multicentricity of lobular carcinoma in situ, and increased risk of contralateral cancer after cancer in one breast. Although true multicentricity is rare in ductal carcinoma in situ (50), the presence of foci of ductal carcinoma in situ have been noted in >40% of breast specimens removed for invasive carcinoma (51), supporting the notion of a field effect encompassing the entire mammary epithelium. Thus, by histological criteria alone, at least 40% of all breast cancer patients potentially have a field predisposition to cancer within the breast. Molecular evidence for the field effect in breast cancer pathogenesis has been published recently (52, 53, 54, 55, 56).

This field predisposition may not be limited to mammary tissue, but it may also include the embryological counterparts of the mammary gland. Accordingly, potential surrogate anatomical sites that are likely to express genetic changes predictive of breast cancer include the skin appendages located along the mammary line, given that the breast is an apocrine sweat gland derivative. Specifically, the hypothesis should be tested that the apocrine sweat glands of the axilla and groin along the mammary line constitute accessible surrogate sites in which genetic changes can be assayed for purposes of assessing risk for breast cancer. Although genetic changes are often found in the normal ductal epithelium and stroma adjacent to a tumor (13, 14), indicating a field effect in the breast, no information exists as to whether this field effect extends beyond the breast to the apocrine sweat glands located along the mammary line. The mammary line would represent an “extended field effect,” not only because it results from a functional as opposed to a purely anatomical continuum but also because the accessible end points are biomarkers rather than cancer.

Those genetic changes that are found by sampling the apocrine sweat glands located in the axilla or groin are likely to be caused by hormones that also affect the mammary epithelium. The objective, therefore, is to estimate the risk of breast cancer based on putative genetic changes that result from the activity of systemically acting hormones (e.g., estrogen) in the apocrine glands located along the mammary line. Sampling of these sweat glands, which can be done by the simple procedure of punch biopsy, would be considered based on their accessible anatomical location as well as on the functional field effect that encompasses them.

Clinical and laboratory observations support a possible functional field effect along the mammary line. Indirect support for such an effect comes from the histological evidence of apocrine metaplasia that is frequently found in the breast in association with fibrocystic change as well as in association with cancer. This apocrine metaplasia resembles the apocrine sweat glands that are found in the axilla, the most cephalad component of the mammary line. Further indirect evidence for the existence of a mammary apocrine functional field is provided by reports of apocrine carcinomas of the breast (57). Support for a functional field that extends beyond the breast comes from the histological similarity between Paget’s disease of the nipple and extramammary Paget’s disease, which arises from epidermally derived adnexal structures in the vulvar area. Furthermore, extramammary Paget’s disease has been observed in a supernumerary nipple (58), providing direct evidence for a field along the mammary line. In fact, extramammary Paget’s disease has occurred concurrently in both the axilla and the groin, sites that are both part of the mammary line (59, 60). A possible biological basis for this association between breast and extramammary abnormalities is provided by the demonstration of estrogen receptor-related proteins in apocrine sweat glands (61). Nevertheless, to date there are no reported studies on the effect of estrogens in glandular structures along the mammary line nor any data as to whether such effects reflect what is occurring in the breast.

In the case of prostate cancer, no anatomical field effect is known to extend outside the prostate. For this reason, the possible existence of a functional field effect encompassing accessible surrogate markers of risk should be explored. The objective here would be to estimate the risk of cancer based on the extraprostatic appearance of genetic alterations resulting from androgenic activity, especially in tissues known to be susceptible to hormonal activity. Such extraprostatic activities may mirror the occurrence of specific molecular genetic effects of androgens in the prostate gland.

At the present time, risk for prostate cancer can only be assessed by direct biopsy or by prostate-specific antigen, which has low sensitivity and specificity. Autopsy studies indicate that prostate cancer starts relatively early. Prostatic intraepithelial neoplasia occurs in 0, 9, 20, and 44% and small foci of overt histological cancers are seen in 0, 0, 27, and 34% of men in the second, third, fourth, and fifth decades, respectively (62, 63). This relatively early onset of prostatic cancer correlates temporally with the development of alopecia within the male population. Although case-control studies have shown no relation between male pattern baldness and prostate cancer, one prospective study has shown such a connection (64). ³ Nonetheless, alopecia is associated with androgenic activity, which ultimately leads to atrophy and destruction of the hair follicle (65). Such destruction is likely to be associated with genetic changes occurring secondary to the presence of androgens. Hair follicle cells in the scalp, like prostatic epithelium, contain androgen receptors (65). The expanded field effect hypothesis as applied to prostate cancer risk is, therefore, that the extent and types of genetic changes seen in hair follicles, whether or not associated with alopecia, may provide information about the effect of androgens on other tissues, notably on the prostate gland. In other words, the prostate and the scalp may be subject to a common functional field effect.

Androgen-elicited changes in the hair follicle may in themselves be genetically determined. The genetic constitution of the hair follicle will influence its response to genetically determined changes in hormone expression. This concept is supported, in part, by studies involving the use of Proscar, a 5α-reductase inhibitor that has been used in clinical trials of benign prostatic hyperplasia, prostate cancer, and prostate cancer prevention. One of the effects of Proscar is reduction in the rate of hair loss, which efficacy, however, appears to be restricted to alopecia presenting in only certain genetic patterns. Insofar as the genetic constitution of the hair follicle modifies its response to androgenic activity, the hair follicle may well have another role in risk assessment. Thus, although prostate cancer is common, clinically evident metastatic disease occurs less frequently. The response to hormones by the hair follicle may distinguish those individuals who are likely to develop clinically evident disease from those with occult disease but who are slow progressors, nevertheless.

Thus, the hair follicles may serve as a surrogate functional site for estimating the risk of prostate cancer. The accessibility of hair follicles as a surrogate site is evident in the ease with which they can be examined, i.e., by means of a small punch biopsy.

A major component of preventive clinical oncology is the assessment of cancer risk with an eye to providing an intervention that will intercept the carcinogenic process. Yet, many of the major cancer target organs are physically inaccessible to sampling biopsy. To address this problem, we have proposed that cancer risk in specific inaccessible target organs can be determined through the identification of genetic changes that occur in accessible surrogate organ sites. The premise of this hypothesis is that there exists an “extended field effect” that incorporates sites as surrogates for target organs based on their functional commonality with the target organ, even in cases where there is anatomical discontinuity. We have described such an “extended field effect” in surrogate anatomical sites such as the upper aerodigestive tract. In addition, we have shown that in the case of hormone-responsive tissues, the notion of a field effect can be extended to include surrogate functional sites such as apocrine sweat glands along the mammary line for breast cancer and hair follicles in the scalp for prostate cancer. Although the evidence for this hypothesis is relatively scant, this hypothesis needs to be tested because of its biological and clinical relevance. The accessibility of these surrogate sites allows the use of relatively noninvasive procedures for risk assessment and earlier detection. Such readily available procedures, together with longer follow-up periods provided by risk assessment, should facilitate the institution of effective chemoprevention trials for individuals at increased risk of cancer.

We gratefully acknowledge the constructive comments of Dr. Alfred G. Knudson, Fox Chase Cancer Center, Philadelphia, PA. We also acknowledge the stimulating discussions with many colleagues, especially Drs. Ernest Hawk, National Cancer Institute, Bethesda, MD, and David Sidransky, Johns Hopkins University, Baltimore, MD.