Path model of multidimensional compliance with cancer therapy

HEALTH PSYCHOLOGY, 1987, (5(3), 183-207 Copyright © 1987, Lawrence Erlbaum Associates, Inc. This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. EMPIRICAL CONTRIBUTIONS Path Model of Multidimensional Compliance W i t h Cancer Therapy Jean L. Richardson, Gary Marks, C. Anderson Johnson, and John W. Graham Department o f Preventive Medicine University o f Southern California Kenneth K. Chan Pharmacokinetics Laboratory University o f Southern California Judith N. Selser, Connie Kishbaugh, and Yvonne Barranday Comprehensive Cancer Center University o f Southern California Alexandra M . Levine Department o f Medicine University o f Southern California Patients newly diagnosed with hematologic malignancies were followed for a 6-month treatment period to assess compliance with three regimen requirements for cancer therapy: anti-neoplastic medication self-administered intermittently, supportive medication self-administered daily, and monthly clinic appointments. The effect on compliance of three intervention "packages" (some combination of education, shaping of pill-taking behavior, and home restructuring) and the extent that patient satisfaction, knowledge, and uncertainty about illness-related events mediated the effects of the interventions were also examined. Blood levels of the drugs and self-report measures indicated that compliance with daily pill taking was higher for each intervention group compared to a control group. Similar results were obtained for compliance with Requests for reprints should be sent to Jean L. Richardson, Department of Preventive Medicine, University of Southern California, Suite A-301, 1420 San Pablo Street, Los Angeles, CA 90033. This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 184 RICHARDSON ET AL. clinic appointments. No improvement in intermittent self-medication was found. Although each intervention package increased patient knowledge and satisfaction, path analyses demonstrated that knowledge did not affect any aspect of compliance, whereas satisfaction was associated with increased appointment keeping only. Daily pill taking was influenced directly by the behavioral components of the interventions. Uncertainty did not influence compliance but was associated with depression, which was negatively correlated with intermittent self-medication. Key words: cancer, compliance, hematologic malignancies, path analysis, intervention, medication It has been estimated that approximately 50% of all patients, across a wide range of diseases, therapies, and individual patient characteristics, are noncompliant with prescribed treatment regimens (e.g. Bergman & Werner, 1963; Marston, 1970; Sackett, 1980; Sackett & Snow, 1979). Compliance with oral cancer therapy, however, has received very little attention (e.g., Barofsky, 1984). It is unclear whether compliance with cancer treatment would be appreciably different from compliance with other disease regimens that have been examined in previous reports. On one hand, one might expect the life-threatening nature of cancer to elevate compliance rates. Fear of treatment side effects, however, with no guarantee of treatment efficacy, may produce avoidance of specific medications. Noncompliance with prescribed medication, if occurring in a significant number of cancer patients, could invalidate the results of clinical trials, prevent proper evaluation of a given treatment regimen, and adversely affect patient survival (e.g., Bonadonna & Valagussa, 1981; Feinstein, 1979; Feinstein & Ransohoff, 1976). Treatment regimens for most malignant diseases are generally quite complex and involve several medications given simultaneously and sequentially. Anti-neoplastic (i.e., active chemotherapeutic) agents are given intravenously at scheduled clinic appointments, and other active agents, such as prednisone, are self-administered orally for short periods of time. Some chemotherapeutic agents have nonspecific cytotoxic effects that can damage proliferating normal tissue or cause side effects that may require the physician to limit their use. Supportive therapy (e.g., allopurinol) is often used in conjunction with anti-neoplastic agents to reduce side effects so that the patient can tolerate vigorous therapy (e.g., Haskell, 1980). Given these characteristics of cancer treatment, it is important that studies of patient compliance include attention to both anti-neoplastic and supportive therapy. The purpose of the present study was to examine compliance with three regimen requirements for cancer therapy: anti-neoplastic medication self- This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. PATH OF COMPLIANCE WITH CANCER THERAPY 185 administered intermittently, supportive medication self-administered daily, and scheduled monthly clinic appointments for infused chemotherapy. We also examined the effect on compliance of three intervention "packages" that included patient education, shaping of pill-taking behavior, and home restructuring and "cueing." Further, the study assessed the extent to which three cognitive-emotional variables (patient satisfaction with health care, knowledge about one's disease and treatment, and uncertainty about illnessrelated events) mediated the effects of the interventions on compliance. BACKGROUND Using random urinary levels of prednisone as an indicator of compliance in patients with acute leukemia, Smith, Rosen, Trueworth, and Lowman (1979) found that 33% of the children and 59% of the adolescent patients were noncompliant with the prescribed drug. In contrast to these compliance figures, studies using medical record abstracts as well as patient self-reports have demonstrated that compliance with clinic appointments to receive intravenously administered chemotherapy may be as high as 98% among adult breast cancer cases (Carey et al., 1979; Glass et al., 1981; Taylor, Lichtman, & Wood, 1984). Although these latter results suggest that compliance with clinic appointments to receive intravenous medication typically is high, they should not be construed to mean that patient compliance with prescribed oral self-administered chemotherapy will be equally high. Patients in Taylor et al.'s study reported that they took their prescribed oral chemotherapy irregularly. Further, Bonadonna and Valagussa (1981) cited noncompliance among patients who were supposed to self-administer cyclophosphamide as a problem in assessing the effectiveness of their clinical trials of breast cancer protocols. As these data suggest, and as found by Inui, Carter, Pecoraro, Pearlman, and Dohan (1980), compliance may differ between aspects of a treatment regimen. Different aspects may be adhered to at different levels because each has its own behavioral demands, each has its own meaning to the patient, and/or because a particular aspect is communicated or reinforced differently by medical providers. Different dimensions of compliance may be influenced by different sets of mediating variables. For example, operant-behavioral models suggest that compliance requires skills training, structuring of one's environment, repetition of information, and reinforcement of behavior (Dunbar, Marshall, & Hovell, 1979; Kasl, 1975). These approaches, however, might be expected to have a stronger impact on compliance with aspects of a regimen that require routine or habitual behaviors such as daily pill taking; they may have much less effect on irregular demands. This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 186 RICHARDSON ET AL. Along the same line, research that has examined aspects of the communication model suggests that improved patient-provider interaction can improve patients' understanding of their disease and treatment (Ley, 1979), their retention of medical information (Bertakis, 1977; Brody, 1980; Cassileth, Zupkis, Sutton-Smith, & March, 1980), and their satisfaction with their health care (Hulka, 1979; Korsch, Gozzi, & Francis, 1968; Ley, Bradshaw, Kincey, & Atherton, 1976). Studies have further shown that patient satisfaction correlates positively with compliance (German, Klein, McPhee, & Smith, 1982; Korsch & Negrete, 1972; Zimmerman, 1982). Although these studies point to the importance of patient satisfaction with providers as a mediator of compliance, satisfaction may not increase compliance with all aspects of a treatment regimen. Pleasant experiences with providers and satisfaction with the medical care one is receiving might have its strongest impact on compliance with that aspect of the regimen (viz., scheduled clinic appointments) that calls for direct contact with providers. Satisfaction may not translate as strongly into increased compliance with home self-administered medications. This possibility has not been examined to date because most studies have not investigated the satisfaction-compliance relationship under multiregimen treatments. Within the communication model, the roles of information and knowledge have also been examined as factors affecting compliance. The first requisite for achieving and sustaining compliance may be adequate understanding of the disease and its treatment (Green, Kreuter, Deeds, & Partridge, 1980). Interventions designed to increase patients' knowledge about their disease, however, have generally not been found to have a large impact on compliance (Green, 1979). Some researchers have suggested that knowledge by itself may not be an important mediator of compliance (Haynes, 1979) — a position that others think may be premature (Leventhal, Zimmerman, & Gutmann, 1984). Knowledge may have a modest effect on compliance in general but by itself may not be a powerful factor. Understanding one's disease and what is required in complying with one's treatment regimen is very different from developing skills needed to perform and sustain compliance behaviors. Thus, knowledge may need to be supported by skills training and operant-behavioral techniques to have an appreciable impact on compliance. Uncertainty about illness-related events and its relation to compliance have received very little empirical study. Uncertainty has been shown to correlate with lack of comprehension among cancer patients (Patterson, 1981). Furthermore, a high degree of uncertainty may decrease the use of direct action and information seeking as modes of coping (Mishel, 1981). Uncertainty can to some extent be minimized by providing clear information and continuing feedback to the patient. In some cases, however, even the care givers may be uncertain about the course of the illness or plans for treatment. Mishel (1981) examined medical patients' uncertainty as it related to symptomatology, diagnosis, treatment, relationship with providers, and plans for This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. PATH OF COMPLIANCE WITH CANCER THERAPY 187 the future. Her factor-analytic study isolated two dimensions of uncertainty: "multi-attributed ambiguity" and "unpredictability." These dimensions may well be important mediators between interventions designed to structure or categorize events and compliance with medications tied to that structure. Finally, the importance of emotional states in compliance has received very little attention. Estimates of the prevalence of depressive symptoms in cancer patients range from 20% to 58% (Craig & Abeloff, 1974; Plumb & Holland, 1977). Despite this prevalence, our understanding of the impact of depression on compliance is minimal. Depressed people are characterized by a negative view of the future, apathy, and an inability to identify and sustain behaviors that will help resolve problems (Beck, Rush, Shaw, & Emery, 1979). These characteristics suggest that depressed patients may be less likely to attend to interventions or to maintain compliant behaviors. PRESENT STUDY In order to examine the issues discussed above, patients newly diagnosed with hematologic malignancies were followed for a 6-month treatment period, and their compliance behavior was monitored. At the time of intake to the hospital, they were interviewed to collect information on demographics, perceived severity of their disease, and depression. At this time, they were also assigned either to a control condition or to one of three intervention conditions that included some combination of three components: shaping of daily pill-taking behavior; home restructuring and cueing; and education pertinent to one's disease, its treatment, and side effects. At the third-month time period, a questionnaire was administered to assess three variables: patient satisfaction with providers, knowledge about one's disease and treatment, and uncertainty about illness-related events. These variables were examined as links between the interventions and three dimensions of compliance: daily pill taking (supportive medication), intermittent pill taking (oral anti-neoplastic medication), and scheduled monthly clinic appointments (infused anti-neoplastic medication). Compliance with medication was assessed with biochemical as well as with self-report indices on a monthly basis throughout the 6-month treatment period. Depression was reassessed at the end of the study period. METHOD Sample Sixty men and 32 women were all diagnosed and treated at the Los Angeles County-University of Southern California Medical Center, which This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 188 RICHARDSON ET AL. serves a largely nonreferral patient population. Participation in the study was limited to persons who were 18 years of age or older, who spoke English or Spanish, and who were able to provide legal consent. Of those who met these criteria, 20% refused to participate. Participants were newly diagnosed with one of six hematologic malignancies: multiple myeloma, acute leukemia, chronic leukemia, indolent lymphoma, aggressive lymphoma, or Hodgkin's disease. The patients were grouped into one of three categories depending on the level of severity of the malignancy (i.e., prognosis). The level of severity was determined from criteria in DeVita, Hellman, and Rosenberg (1982). The high-severity group (26%) consisted of patients with aggressive tumors requiring intensive multiagent chemotherapy; these diseases include acute leukemias, small noncleaved (Burkitt) lymphoma, convoluted T lymphoblastic lymphoma, and immunoblastic sarcoma. Patients with these diseases generally show good response to chemotherapy initially, but the prognosis is short-term survival (generally less than 2 years). The moderate-severity group (31%) consisted of patients with multiple myeloma, large cleaved lymphoma, and large noncleaved lymphoma. Patients with these diseases generally show good response to therapy, and prognosis is moderate survival (2 to 5 years). The remaining 43% of the patients had relatively indolent diseases including chronic leukemias, small cleaved lymphoma, and Hodgkin's disease. Patients with these diseases generally have a good chance of long-term survival (greater than 5 years). The patients ranged in age from 18 to 86 years, with a median age of 42 years. The sample was primarily public-pay patients of lower socioeconomic status. The major ethnic groups represented were Hispanic or Latino (55%), black (27%), and non-Hispanic white (14%). Spanish was the primary language for 38%. More than half (56%) were either unemployed, disabled, or retired; 21 % were employed either full- or part-time; 7% were homemakers, and 2% were students. Currently or most recently they were laborers, service workers or operatives (62%), foremen or craftsmen (15%), managers or professionals (11%), clerks and salesmen (12%). Sixty-seven percent reported annual incomes of less than $10,000. The median level of education was high school experience short of a diploma. The scores on the measures of occupation, income, and education were separately standardized and then averaged to create a socioeconomic (SES) index for each patient. Treatment Protocols Compliance with two drugs, allopurinol and prednisone, was monitored. All patients received 300 mg of oral allopurinol per day; they were to selfadminister every 24 hr. This medication plays a supportive role in cancer This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. PATH OF COMPLIANCE WITH CANCER THERAPY 189 therapy.1 Prednisone was also to be self-administered orally; it was prescribed for approximately 60% of the patients. This drug is an antineoplastic agent in the treatment of hematologic malignancies and was to be taken consecutively for a 5-day period each month.2 Most clinical protocols for the treatment of hematologic malignancies use multidrug regimens. In the present study, the regimens were categorized as either "easy" or "complex" based on the difficulty of tolerating the treatment and on the number of chemotherapeutic agents involved. Thirty-one percent of the patients had easy regimens, and 69% had complex regimens.3 Study Design The study used a sequential cohort design in which successive cohorts of newly diagnosed patients were entered into different conditions (Cook & 1 Allopurinol is taken to prevent uric acid nephropathy due to tumor lysis that may complicate the vigorous treatment of acute leukemia, myeloma, and lymphoma. Clinically, allopurinol is used continuously or during the phase of vigorous treatment. The most common adverse side effect of this drug involves occasional hypersensitivity reactions of the skin or blood (see Haskell, 1980). Patients who were allergic to allopurinol had the drug withdrawn and were dropped from the study. Patients were informed when they entered the study that allopurinol does not produce any of the side effects commonly associated with cancer chemotherapy (e.g., nausea, vomiting, hair loss, unusual fatigue). Because allopurinol does not produce difficult or extreme side effects and because patients were informed of this at the start of their participation in the study, it is unlikely that they were consciously withholding self-medication to eliminate physical effects of the drug. Corticosteroids, such as prednisone, are commonly employed as primary cytotoxic drugs in the treatment of malignant hematopoietic diseases such as acute lymphoblastic leukemia, chronic lymphocytic leukemia, lymphomas, and multiple myeloma. Short intensive courses of a short-acting corticosteroid such as prednisone are usually employed, although for some chronic conditions, alternate-day prednisone may be used to maintain a beneficial response. Prolonged daily high-dose corticosteroid therapy may be associated with a host of complications and side effects that are not usually seen with the short, intensive courses used to treat hematologic malignancy. Sodium and water retention with development of significant weight gain, potassium loss, mood change, and exacerbation of diabetes mellitus are reactions that may be seen more commonly with short-term prednisone administration (see Haskell, 1980). These potential side effects were carefully monitored in our patients. 3Easy regimens were considered to be (a) melphelan and prednisone; (b) cyclophosphamide, vincristine, and prednisone (CVP; e.g., Bagley, DeVita, Berard, & Canellos, 1972); and (c) daily oral chlorambucil, with or without prednisone. The complex regimens included (a) nitrogen mustard, vincristine, procarbazine, and prednisone (MOPP; e.g., DeVita, Serpick, & Carbone, 1970); (b) hydroxyldaunamycin, bleomycin, vincristine, and DTIC (ABVD; e.g., Santoro, Bonfante, & Bonadonna, 1982); (c) hydroxyldaunamycin, cyclophosphamide, vincristine, and prednisone (CHOP; e.g., McKelvy et al., 1976); (d) bleomycin, hydroxyyldaunamycin, cyclophosphamide, vincristine, and prednisone (BACOP; e.g., Schein et al., 1976); and (e) methotrexate, bleomycin, hydroxyldaunamycin, cyclophosphamide, vincristine, and dexamethasone (M-BACOD; e.g., Skarin et al., 1983). These agents have different mechanisms of action, pharmacologic characteristics, standard dosages, side effects, and clinical purposes. For an introduction to cancer chemotherapeutic agents, see DeVita et al. (1982) and Haskell (1980). This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 190 RICHARDSON ET AL. Campbell, 1979; Haynes, Taylor, & Sackett, 1979). For each 2- to 3-month period, all newly diagnosed patients received the same intervention package. The control and intervention conditions were implemented four times each during the 3-year course of the study. The conditions were counterbalanced to avoid seasonal effects, practice effects on the part of the staff, and contamination by other factors. In hospital-based studies, contamination (spillover) can easily occur if patients become aware that other patients are receiving different treatment from health care providers. In such settings, a successive cohort design decreases potential spillover effects, whereas a design with random assignment of participants to conditions does not. Intervention Design The study consisted of four conditions: no-intervention control (n = 23), education and home restructuring (n = 18), education and shaping (n = 26), and education, shaping, and home restructuring (n = 25). Patients were assigned to one of these four conditions based solely on the sequential cohort that was in effect on the date they entered the study. Educational component. This component consisted of an interactive slide-tape presentation administered in Spanish or English by the project nurse prior to the onset of treatment. Six similar slide-tape presentations were prepared, each specific to a particular disease category. Each presentation contained three sections: (a) What is the disease? (b) What is the treatment, how does it work, what are the side effects, and why do they occur? (c) What is the patient's responsibility regarding his or her health care? The importance of complying with active and supportive care to achieve a positive treatment response was stressed. After each section, the nurse stopped the tape to ascertain whether the patient could recall the information and answered any questions. Home visit. The second intervention component consisted of a home visit made by the project nurse within 1 week of patient discharge from the hospital. The nurse ascertained the patient's daily habits so that a "cueing" system for pill taking (e.g., positioning pills next to a toothbrush or coffee pot) could be tied to those habits. A discharge poster, designed to reinforce the educational component, was placed in the home. A family member was asked to be present during the home visit, at which time a written contract was agreed upon and signed by the patient and relative. The contract was used primarily to formalize a role in compliance for the family member, who agreed to help the patient with pill taking, with problem solving related to the illness and its treatment, and with keeping clinic appointments. A similar PATH OF COMPLIANCE WITH CANCER THERAPY 191 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. program has been shown to be effective with hypertension patients (Morisky etal., 1983). Shaping component. This component consisted of a four-level pillshaping procedure implemented while the patients were in the hospital. During the first level, the name, purpose, side effects, and proper schedule of the drugs to be taken were explained. During the second level, the nurses returned with the pills and asked the patient to identify the pill and dosage. After this level was successfully completed, the third level was introduced. The patient was told to ring for the nurse to bring each pill at the appropriate time. At the fourth level, medications were brought to the bedside, and the patient was responsible for self-medication under the supervision of the project nurse. A similar program has been shown to be effective with hypertension patients (Johnson & Beardsley, 1978). The control patients did not receive any of the specially developed educational materials, they did not receive a home visit by a project nurse, and they were not trained in details of shaping pill-taking behavior. All the intervention materials were carefully stored so that they would not be inadvertently used with control patients. At the time of diagnosis and protocol work-up, physicians and nurses gave control patients standard (routine) explanations of their illness, of its treatment, and of its side effects; also, the patients' questions were answered. During the 6-month study period, control and intervention patients received similar behavioral treatment from the health care providers at the clinic. Questionnaire Measures Each patient was followed for a 6-month treatment period. During this time, three waves of confidential questionnaires were administered individually to each patient. They were administered in either English or Spanish by a bilingual woman trained in social work and interviewing. Only the questionnaire items pertinent to the concerns of this article are discussed below. Intake questionnaire. This questionnaire was administered within 1 week of diagnosis. With respect to perceived severity of the illness, patients were asked, "How ill do you think you are now?" They could respond either extremely, somewhat, or not very ill. Psychological adjustment was assessed with the Zung Depression Scale (Zung, 1965). Only the 10 items that related to psychological functioning were used in our analysis. The items that related to physical functioning were omitted because responses to them would likely be confounded with the physical effects of one's cancer. In addition to filling out the Zung scale, which appeared at the end of the questionnaire, partici- 192 RICHARDSON ET AL. This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. pants were asked at the beginning to indicate how sad or depressed they were on a 4-point Likert-type scale. Their responses to this question correlated (.45) with their average Zung rating, and alpha reliability for the 10 Zung items and the "sadness" item was .78. A single depression score for each patient was created by standardizing and then averaging the responses to these items. Third-month questionnaire. The first section of this questionnaire contained 5 items related to patients' satisfaction with the medical care they were receiving. These questions were similar to those used by Atkisson and Nguyen (1979). The items dealt with the quantity and adequacy of healthrelated information received, perceived concern on the part of the staff, and general satisfaction with care. Patients responded on 4-point Likert-type scales. Alpha reliability for the set was .75. The items were separately standardized and then averaged to create an overall satisfaction score. The next 15 items dealt with patients' knowledge about their illness, treatment protocol, and side effects. The content of the questions was based largely on the educational intervention. Two thirds of the questions were multiple-choice, and one third was of matching format. The number of correct responses comprised the knowledge score. Patients were then administered the Uncertainty in Illness Scale (Mishel, 1980, 1981). This is a 34-item scale tapping uncertainty in symptomatology, diagnosis, treatment, relationship with care givers, and planning for the future. Although Mishel (1981) isolated two distinct dimensions of uncertainty ("multi-attributed ambiguity" and "unpredictability"), a factor analysis of our patients' responses did not yield any clear factor structure. However, because the alpha reliability for the entire set of items was .87, a single uncertainty score was computed for each patient. Sixth-month questionnaire. Psychological adjustment was reassessed with the Zung scale. The patients' average rating correlated .39 with the single-item "sadness" rating made toward the beginning of the questionnaire. The alpha reliability for these items was .79; thus, a single depression score was calculated. Compliance Measures Drug profile. Shortly after diagnosis, serum samples were obtained from 19 patients for detection of allopurinol and its metabolite and from 13 patients for prednisone and its metabolite. Serial serum samples were taken at the start and at lA hr, Vi hr, 1 hr, 1 Vi hr, 2 hr, 3 hr, 4 hr, 5 hr, 6 hr, 8 hr, 12 hr, and 24 hr and were stored at - 20°C for subsequent assay. To construct the drug profiles, each patient's blood sample at each time period was divided This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. PATH OF COMPLIANCE WITH CANCER THERAPY 193 in half and was independently analyzed. The mean value from the two analyses was then computed for each time period for each individual tested. Plasma prednisone levels were assayed by high-pressure liquid chromatography slightly modified from Frey, Frey, and Benet (1979). This method separates prednisone from its major metabolite, prednisolone, and provides a limit of detection of 10 ng/ml. Plasma allopurinol levels were assayed by the high-pressure method of Breithaupt and Goebel (1981), which allows separation of allopurinol from its major metabolite, oxipurinol, and which can detect allopurinol in plasma at levels as low as 100 ng/ml. Because the treatment protocols employed often consisted of multidrug regimens such as MOPP, the assay procedures were evaluated for possible interference by the concomitant drugs and their metabolites. We found that nitrogen mustard, vincristine, procarbazine, cyclophosphamide, adriamycin, bleomycin, melphelan, chlorambucil, myleran, or their metabolites did not interfere with the assay, and allopurinol and prednisone did not cross-interfere. The profiles were then examined to determine the minimum and maximum blood level of the drugs and their metabolites for the period from 1 to 4 hr after ingestion. This 4-hr "window" was selected for study because patients were instructed to take their medications in the morning, with clinic appointments (at which blood was drawn) scheduled for late morning hours. Thus, patients who took their medications that day should have taken them within the previous 4 hr. The profiles for the two drugs and their metabolites were analyzed to obtain 95% confidence intervals (CI) for detection of the medications within the window. The CI for prednisone levels was .028 - .518 fig/m\ (100 ng = 0.1 jag), and the CI for prednisolone was .243 - 1.204 ^g/ml. The CI for allopurinol was .051 - 2.086 ^g/ml, and the CI for oxipurinol was 1.616 7.95 ^g/ml. Although the CIs were computed based on the drug levels obtained 1 to 4 hr after drug ingestion, the CIs were sensitive to drug ingestion occurring many hours before the drawing of blood. For example, analysis of the profiles at the 24-hr period indicated that allopurinol was present at a mean level of .25 fig/ml, and oxipurinol was present at a mean level of 3.23 jag/ml. Each of these levels are included within their respective CI. In fact, the CI for oxipurinol indicated the metabolite could be detected even when the allopurinol had been ingested up to 48 hr prior to the drawing of blood. The mean prednisone level at 24 hr (.05 ^g/ml) was included in the respective CI, but the mean prednisolone level (.06 fig/ml) was slightly below the CI. These ranges of observed drug concentrations are fully comparable to those reported in previous studies (e.g., Breithaupt & Goebel, 1981; Frey et al., 1979). Monthly blood samples. After the profiles were analyzed and confidence intervals constructed, all patients were followed for a 6-month treat- This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 194 RICHARDSON ET AL. ment period. A sample of blood was obtained at each scheduled monthly clinic visit. In addition, because prednisone was given on an intermittent basis, we arranged that the drawing of blood at the patients' homes would coincide with days that the medication should have been taken. More specifically, most patients were to take prednisone consecutively for a 5-day period each month. Thus, we arranged to collect blood on the fourth day. If a patient was told by his or her doctor not to take prednisone (due to low white blood count, low platelet count, or presence of severe side effects) or allopurinol (due to allergic reaction) for a period of time, the data reflecting that period were not entered into our analyses. That is, physiological toxicity necessitating discontinuation of medication was not classified as noncompliance. Few of our patients were instructed to withhold medication at any point in time. Monthly blood samples from each patient were compared to the individual's profile standard. For patients for whom no initial profile was obtained, the monthly drug and metabolite levels were compared to the CI, which had been constructed from other samples. Each monthly level was classified either as fully compliant (drug levels within the individual's profile range or CI), undercompliant (drug levels lower than the individual's range or CI), or noncompliant (no drug present). For each patient, the proportion of time across the 6-month period that the samples fell into each of these three categories was computed. Thus, these proportions calculated separately for the parent drug and metabolite constituted the primary measures of compliance. Each patient should have had six monthly serum samples for drug analysis. If a given sample was missing due to persistent failure to attend the clinic that month (three or more broken appointments within 1 month), the patient was considered noncompliant for that month. If the patient missed the appointment for a valid reason, we assumed compliance with self-medication at the same levels observed when blood samples were provided. For example, assume that a patient provided five serum samples and that he or she was categorized as fully compliant, then undercompliant, then undercompliant, then fully compliant, and then noncompliant. If the missed saample was due to an invalid reason, then the missed sample would be coded noncompliant, and the scores for that patient would be 33.33% for each of the three compliance categories. Alternatively, if the patient missed the sample for a valid reason (or was taken off the drug for that time period), then the scores would be 40% fully compliant, 40% undercompliant, and 20% noncompliant. Self-report of allopurinol compliance. A self-report of medication taking was collected monthly. At each clinic visit, the patients were asked how many days in the preceding week they had taken the prescribed allopurinol. A response of 7 days was considered 100% compliant. The responses over the 6-month time period were averaged, and an overall percentage was computed. PATH OF COMPLIANCE WITH CANCER THERAPY 195 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Appointment keeping. A score was recorded for each patient indicating the number of monthly clinic visits kept on the first or second scheduled appointment. An overall percentage was then computed across the 6-month period. RESULTS The results are divided into three sections. First, preliminary analyses were performed (a) to examine whether our conditions differed with respect to patient characteristics, disease type, and treatment complexity, and (b) to assess whether these variables related to any of the critical dependent measures (e.g. uncertainty, knowledge, satisfaction, compliance). Second, the effects of the interventions were evaluated with an analysis of variance (ANOVA) or with an analysis of covariance (ANCOVA). Finally, an overall path model that included intake, third-month, and sixth-month variables was examined with multiple-regression analyses. Preliminary Analyses The experimental and control conditions did not differ significantly on any of the patient characteristic variables, on the severity of patients' illness (actual or perceived), or on the complexity of treatment regimen. Patients' levels of depression at intake were also equivalent across conditions. Some of these variables, however, did correlate with one or more of the dependent variables. Depression at intake correlated negatively with patient satisfaction (r = - .39, p < .01) and positively with uncertainty about illness (r = + .30, p < .01). Complexity of treatment regimen correlated marginally with uncertainty (r = - . 19, p = . 10) and with prednisone compliance (r = - .23, p = .11). Finally, SES and knowledge about one's illness correlated (r = + .57, p < .01). Given these findings, depression at intake, treatment complexity, and SES were entered as covariates in several analyses of the intervention effects. Compliance With Allopurinol Biochemical assessment. Our first set of analyses were performed with compliance summary scores. The proportion of time a patient was fully compliant was weighted as 2, the proportion of time undercompliant was weighted as 1, and the proportion of time noncompliant was weighted as 0. Separately for allopurinol and oxipurinol, these scores were summed to form single compliance indices for each patient. ANOVAs revealed significant differences between conditions in allopurinol compliance as indicated by serum This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 196 RICHARDSON ET AL. levels of allopurinol, F(3, 88) = 5.02,/? < .01, and of oxipurinol, F(3, 88) = 6.44, p < .01. To provide a more detailed analysis of compliance rates, the three categories of compliance were considered separately. As seen in Table 1, analysis of serum levels of allopurinol indicated that control patients were noncompliant an average of 77% of the time. Under compliance was found an average of 6.9% of the time, and full compliance was observed only 16.1 % of the time. Analysis of serum levels of oxipurinol indicated that control patients were noncompliant with daily allopurinol 43.3% of the time, undercompliant 7.9% of the time, and fully compliant 48.8% of the time. Compliance levels were higher for the metabolite because, as previously discussed, it has a longer half-life (up to 48 hr) and is thus detectable in the blood for a longer period of time. For example, if a patient did not take allopurinol on the day of blood sampling, but did take it a day or two before, allopurinol would not be readily detectable, whereas oxipurinol would be present, reflecting the past dose. This situation in which the parent drug is nondetectable but the metabolite is detectable may indicate sporadic or inconsistent compliance behavior. Significant increases in compliance with allopurinol were found for patients who received an intervention (see Table 1). Analysis for allopurinol indicated that full compliance increased to 45% to 50% for the intervention groups. Analysis for oxipurinol revealed that full compliance increased to 76% to 84% for intervention patients. Duncan multiple-comparison tests demonstrated that each measure of compliance was significantly higher in each intervention group compared to the control group (ps < .01); however, the three intervention conditions did not differ from one another. The summary measures of allopurinol and oxipurinol correlated .47 across all patients (p < .01). Self-report of allopurinol compliance. An ANOVA performed on patient self-report demonstrated a significant main effect of condition, F(3, 87) = 20.88, p < .001. Control patients reported that they were compliant with daily allopurinol an average of 48 % of the time; intervention patients reported that they were compliant between 90% and 93% of the time. Duncan tests indicated that the control group differed (p < .01) from ech of the three intervention conditions; these three intervention conditions did not differ from one another. It is of interest that the self-report overestimated compliance with allopurinol by approximately a factor of two when compared to the biochemical measure of the parent drug. The self-report and the biochemical assessment of oxipurinol, however, yielded very similar results. Consistent with this pattern, Pearson correlation analysis (performed across all patients) revealed a stronger positive relation between self-report of allopurinol compliance and the biochemical summary measure of oxipurinol (r = .53,/? < .01) than be- This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. C CO o c '»— ZJ Q. O < V) 0> CD 0 CO 0) CO OS CD 2 o.9K So CO "5. E o o k— CD "D C Z) "c co "5. E o O _>% "5 LL CD E o 0 O) OJ "c 0 o 0 Q. c CO 0 too 1 1 1 a I CO Qi si oo © . .§ 8 f t CO CO ~5 C 8 ^ Q. o < i - CO o 5 1 c p ON <7\ £ * i5 S « a god Q * ^ s a ii i <3 .a #c 1.1 < O ^ ^ ii 197 198 RICHARDSON ET AL. This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. tween self-report and the biochemical summary measure of allopurinol (r = .27, p < .05). These findings are likely due to the fact that self-report and the measure of oxipurinol are each sensitive to behavior that occurred 2 days or more before the drawing of blood, whereas the biochemical measure of allopurinol is sensitive only to a 24-hr period. Thus, the first two measures would be more sensitive to erratic or intermittent compliance. Compliance With Prednisone Compliance with intermittent prednisone was first examined with an ANCOVA (with complexity of treatment as a covariate). Analysis of the biochemical summary scores for the parent drug and its metabolite showed no significant differences between conditions (Fs < 1). The data broken down for each compliance category are presented in Table 2. Because there were so few patients undercompliant with prednisone, the undercompliant and fully compliant categories were combined to form a "compliant" category. Control patients were compliant 31.2% of the time as indicated by the presence of prednisone in the serum and 24.8% of the time as indicated by the presence of prednisolone. Similar results were obtained in each of the three intervention conditions. Analysis performed across all patients revealed that prednisone compliance—calculated as the average of the two biochemical summary measures, which were correlated .76 (p < .01)—correlated negatively with depression assessed at that the sixth-month period (r = - .37, p < .05). Appointment Keeping An ANOVA revealed significant differences between conditions in compliance with monthly clinic appointments, F(3, 89) = 8.21, p < .01. Control patients were compliant with appointments an average of 61% of the time, whereas intervention patients attended clinic between 83% and 92% of the time. The control group differed (p < .01) from each intervention group, but the three intervention groups did not differ from one another. Uncertainty, Satisfaction, and Knowledge An ANCOVA was performed on the patients' uncertainty about illness scores (with depression at intake and complexity of treatment as covariates). The analysis showed no significant treatment effects, F(3, 80) = 1.20, p > .30. An ANCOVA on the measure of patient satisfaction (with depression at intake as the covariate) indicated that the interventions had a significant impact on this variable, F(3, 86) = 2.97, p < .05. As seen in Table 3, satisfaction with medical care was higher for patients in each intervention group compared to control patients. The three intervention groups did not differ o CO •D c This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 4) t/3 6 S too '§••5 CO Qi 1 1 3 ST CO CD c o CO "c " OD 0 CD >0 a 1 <-> , .3 <L> £ »-i <u E O o h y «i OW) eg S3 »_o 03 & « 0 CO "o 0 =3 CO CO 0 CO CO CM <D _J o m .<2 u * I * O b <l> to 2 g cu s ^ £ o 3 •s § < £ ^ 0 S 2 g c CO "5. E o o c o c CO "q. E o O 0 E o 0 B 0 o 0 Q_ c CO 0 2 CS OO ft. o a od >n l 5 00 N HI ^0) s I 3a, o 'rt(> oj o c<L o> >> L 43 p o E- ^ T3 "3 C OCD^H § H3 a> O II S T3 a o03 ' <U3 es:n 1 ft 3 ft 3 8 C "§ J3 2 joa H o 13O ,; "3 C a t3 T3 S a> <u §>&£ h § jq * c 03 O +-» o R- v | 2 .§ ^ 2 I' "a "3 '3 s: S 5 a> • O ^ *j -h a 5 §« ,9 y §t 199 200 RICHARDSON ET AL. TABLE 3 Patients' Uncertainty, Satisfaction, and Knowledge Assessed on Third-Month Questionnaire This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Condition Dependent Measure Uncertainty about illness Satisfaction with medical care Knowledge about disease Control -•Ola 5.58a Education and Home Visit Education and Shaping -•05. •21b 7.54, -.07. -17b 8.35b Education, Shaping, and Home Visit •l4a 7.79b Note. The adjusted means for uncertainty and satisfaction are standardized scores; the adjusted knowledge scores are the mean number of knowledge questions answered correctly. Higher means reflect greater uncertainty, satisfaction, and knowledge. For each measure, means not sharing a common subscript differ at/? < .05 (Duncan test). f r o m one another o n satisfaction, b u t the education/home visit and educat i o n / s h a p i n g groups each differed f r o m the c o n t r o l group (ps < .05). Finally, an A N C O V A o n the measure o f patients' knowledge about their disease and its treatment ( w i t h SES as a covariate) demonstrated a highly significant treatment effect, F ( 3 , 86) = 5.13, p < . 0 1 . Each intervention group showed greater knowledge t h a n d i d the c o n t r o l g r o u p , b u t the three intervent i o n conditions d i d n o t differ f r o m one another. Correlational analysis across all patients showed that uncertainty correlated negatively w i t h satisf a c t i o n ^ = - .48,/? < .01) and w i t h knowledge ( r = - .21,/? > .05). Satisfaction and knowledge correlated positively ( r = .22, p < .05). Path Model The preceding analyses were concerned with the specific effects of the intervention conditions. It is also important to get an overall view of the predictive and mediating relationships among the intake, third-month, and sixthmonth variables. Thus, a multiple-regression path analysis was performed. The intake (or Time 1) variables included depression, treatment complexity, SES, perceived and actual disease severity, and experimental condition. With respect to experimental condition, because the three intervention groups did not differ on any of the dependent measures, patients in these groups were pooled, and a single intervention group was formed. Thus, for the path analysis, condition was a dichotomous variable (control group = 0, intervention group = 1 ) . The third-month (or Time 2) variables were uncertainty about illness-related events, patient satisfaction, and knowledge about one's disease. Finally, the sixth-month (or Time 3) variables included depression; ap- This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. PATH OF COMPLIANCE WITH CANCER THERAPY 201 pointment keeping; biochemical summary scores for prednisone, prednisolone, allopurinol, and oxipurinol; and self-report of allopurinol compliance. The model and significant paths (standardized regression coefficients) are presented in Figure 1. The path diagram does not include perceived or actual severity because these variables, although marginally correlated (r = .16), did not covary with any other variable. It is of particular interest that experimental condition predicted patient satisfaction at Time 2 and that satisfaction was predictive of compliance with monthly clinic appointments but was not predictive of compliance with daily allopurinol or intermittent prednisone. On the other hand, experimental condition was directly predictive of allopurinol compliance (self-report as well as biochemical measures), but only marginally predictive of appointment keeping. Patients' knowledge about their disease and uncertainty about illness-related events did not predict any aspect of compliance. Uncertainty, however, was associated with depression. Depression at intake predicted uncertainty at the third month, which in turn predicted depression at the sixth month. There was also a strong relation between depression at the beginning of treatment and at the end of the study period. Time 1 (Intake) Time 2 (Third Month) Time 3 (Sixth Month) * Depression Complexity of treatment Prednisone compliance (serum prednisone) Prednisone compliance (serum prednisolone) Appointment-keeping Allopurinol compliance (self-report) Experimental condition Allopurinol compliance (serum allopurinol) Allopurinol compliance * (serum oxipurinol) FIGURE 1 Multiple-regression path model. Entries are standardized regression coefficients. *p < .10. **p < .05. ***p < .01. 202 RICHARDSON ET AL. This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. DISCUSSION Before elaborating on our obtained results, a couple of insights should be shared regarding the strengths and weaknesses of each method of measuring compliance. Biochemical analysis allows one to directly examine the extent to which drugs and metabolites are present in a patient's serum sample. Based on these levels, one then makes an inference about compliance rates. Although biochemical assessments are not subject to self-presentation biases as are self-reports, they are intrinsically linked to the time period in which the drug and metabolite are detectable in the blood. In the case of allopurinol, for example, the parent drug is detectable up to 24 hr after ingestion, whereas the metabolite is detectable up to 48 hr after ingestion. Although the metabolite provides an extended picture relative to the parent drug, blood measures basically permit an assessment of compliance at a single point in time. These measures can be improved, however, in direct proportion to the number of blood samples taken. The present study collected six samples at monthly intervals and averaged them together to yield a general compliance indicator. Self-report measures of compliance, on the other hand, can cover a longer period of time; a patient can be asked to report on the previous day, 2 days, 3 days, or further back in time. Our study requested patients to focus on the previous 7-day period. Self-reports can thus cover a longer time span than biochemical analysis are capable of covering. Self-reports, however, generally overestimate compliance due to self-presentation bias and due to the problem (especially among the elderly) of forgetting when and if medications were taken. For each of these measures, it is incumbent on the researcher to maintain on a daily basis specific information on the drugs prescribed and to not erroneously rate patients as noncompliant when the drug has been withdrawn due to allergic or toxic reactions. Our findings demonstrated that the level of compliance with each of three regimen requirements for cancer therapy among our control patients was very low. Biochemical measures revealed that these patients were fully compliant with self-administered allopurinol (daily pill taking) 16% of the time. That is, approximately one of six blood samples collected for each control patient indicated that allopurinol had been taken within the previous 24-hr period as instructed. Biochemical analysis indicated that control patients were fully compliant with self-administered prednisone (intermittent pill taking) 31% of the time. Finally, patients in the control condition kept their scheduled clinic appointments 61% of the time. These differences in baseline levels of compliance across the three dimensions of compliance suggest that researchers and practitioners should not assume that levels of compliance observed on one regimen requirement generalize to other demands of the regimen. One factor that might have contributed to noncompliance was the com- This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. PATH OF COMPLIANCE WITH CANCER THERAPY 203 plexity of the treatment regimen. The most complex regimens are generally associated with decreased levels of compliance (Haynes, 1979). Our finding of a marginally significant correlation between complexity and prednisone compliance provides support for this contention. Although our control and intervention groups were similar in composition with respect to complexity of regimen, 69% of all the patients did receive complex multi-agent treatment, which may have contributed to the low levels of compliance. It should be understood, however, that the majority of all treatment regimens for cancer are, as in this study, of a complex nature. The extent to which the low level of baseline compliance is due to the low SES of our sample is difficult to determine. For our patients, differences in SES did relate to differences in knowledge but not to differences in compliance. The range of SES in our sample, however, was narrow and low. Nevertheless, previous studies of compliance for other diseases suggest that SES is not a useful predictor of compliance behavior (Haynes, 1979); therefore, we might expect our obtained results to generalize to other SES levels. We found that the interventions significantly improved full compliance with allopurinol from 16% to about 50% (biochemical measures) and compliance with clinic appointments from 61% to between 83% and 92%. No significant improvement in compliance with intermittent pill taking was found. With respect to the first two measures above, each intervention condition showed better compliance rates than the control group, but the three intervention conditions did not differ from one another. One can speculate that the intervention effects were due largely to patient education because education was a component in each of the three intervention packages. This explanation, however, runs into trouble because if the effects were due mostly to education, then knowledge should have had an impact on compliance. Although the interventions increased knowledge, knowledge did not affect any aspect of compliance. What about the increased attention given to patients by the providers during the education sessions —might this explain the compliance effects? It is likely that attention did have an impact. Our data suggest, however, that attention may have had an impact only on appointment keeping. That is, increased attention would likely translate into increased satisfaction. Patient satisfaction, however, was predictive of compliance with clinic visits but not predictive of compliance with home self-medication. Daily pill taking was directly influenced by the interventions, apart from any of the mediating variables we measured. In each intervention group, daily allopurinol was specifically targeted with operant-behavioral techniques of shaping and/or home restructuring and cueing. Although we cannot isolate the contribution of each of these components, these techniques likely influenced allopurinol compliance directly or through mediating variables other than the cognitiveemotional variables we examined (satisfaction, knowledge, uncertainty). This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 204 RICHARDSON ET AL. These variables are not likely to be important mediators of operant-behavioral techniques. A measure of the extent to which cueing and related procedures were maintained in the home would be more appropriate to consider. We should note that our interventions were initiated at the time of intake to the hospital, but they were not regularly reinforced throughout the 6-month treatment period. This may have contributed to the lack of any treatment effect for intermittent pill taking (prednisone). Interventions designed to influence this dimension of compliance should be reinforced periodically during the treatment process. In addition to operant techniques, procedures designed to reduce the perceived complexity of intermittent self-medication might be beneficial (e.g., monthly calendars). Our results clearly demonstrated that satisfaction was a more useful predictor of compliance than was knowledge. The interventions improved both knowledge (likely due to education and patient-provider interaction about the disease) and satisfaction. These findings are consistent with experimental prospective communication studies that have shown that improved patientprovider communication leads to improved satisfaction and knowledge (Ley, 1979; Ley et al., 1976). Although these two variables were found to be positively correlated, however, only satisfaction was associated with increased compliance with scheduled clinic appointments. These results substantiate retrospective studies that have found satisfaction (German et al., 1982; Korsch & Negrete, 1972) but not knowledge (Green, 1979) to be correlated with compliance. It is important to emphasize that the type of compliance that was associated with satisfaction was appointment keeping and not either measure of oral pill taking. It is likely that patients keep appointments when they have been satisfied with their interactions with providers, but this satisfaction does not necessarily mean that they will take their self-administered pills consistently. Perhaps this is because failure to take pills can be more easily hidden from providers, whereas failure to keep appointments may harm a satisfying and necessary relationship with providers. To our surprise, the interventions did not affect patients' uncertainty about illness-related events, and uncertainty did not correlate with any aspect of compliance. Further, the form of the marginal relation between complexity of treatment regimen and uncertainty is perplexing. We anticipated that people with more complex regimens would feel more uncertain. In fact, an opposite tendency was found. Despite these null and puzzling results, we should not be too quick to dismiss the potentially important role that uncertainty may play in mediating compliance effects. Perhaps future studies that can successfully isolate specific dimensions of uncertainty may find that particular dimensions relate to specific compliance behaviors. Finally, depression appeared to be stable across the 6-month study period. Depression at intake was the best predictor of depression at 6 months and This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. PATH OF COMPLIANCE WITH CANCER THERAPY 205 was predictive of both uncertainty and satisfaction. Patients who were depressed at diagnosis were less satisfied with their health care and more uncertain about health-related matters. The realtionship with satisfaction is important because satisfaction affects appointment keeping. Depression at 6 months was also significantly correlated with failure to take intermittently prescribed pills. Patients who are more depressed may be less capable of complying with changing regimens. These findings suggest that mood may play an important role in compliance—a role that deserves further attention. ACKNOWLEDGMENT This research was supported by National Cancer Institute Grant C A 31151. REFERENCES Atkisson, C. C , & Nguyen, T. D. (1979). Service Evaluation Questionnaire. San Francisco: University of California Press. Bagley, C. M., DeVita, V. T., Berard, C. W., & Canellos, G. P. (1972). Advanced lymphosarcoma: Intensive cyclical combination chemotherapy with cyclophosphamide, vincristine, and prednisone. Annals of Internal Medicine, 76, 227-234. Barofsky, I. (1984). Therapeutic compliance and the cancer patient [Special issue]. Health Education Quarterly, 10, 43-56. Beck, A. T., Rush, A. J., Shaw, B. F., & Emery, G. (1979). Cognitive therapy of depression. New York: Guilford. Bergman, A. B., & Werner, R.J. (1963). Failure of children to receive penicillin by mouth. New England Journal of Medicine, 268, 1334-1338. Bertakis, K. D. (1977). The communication of information from physician to patient: A method for increasing retention and satisfaction. Journal of Family Practice, 5, 217-222. Bonadonna, G., & Valagussa, P. (1981). Dose-response effect of adjuvant chemotherapy in breast cancer. New England Journal of Medicine, 43, 169. Breithaupt, H., & Goebel, G. (1981). Determination of allopurinol and oxipurinol in biological fluids by high-performance liquid chromatography. Journal of Chromatography, 226, 237-242. Brody, D. S. (1980). An analysis of patient recall of their therapeutic regimens. Journal of Chronic Diseases, 33, 57-63. Carey, R. W., Sohier, W. D., Kaufman, S., Weitzman, W. A., Kelley, R. M., Lew, R. A., & Halpern, E. (1979). Five-drug adjuvant chemotherapy for breast cancer. Cancer, 44, 35-41. Cassileth, B. R., Zupkis, R. V., Sutton-Smith, K., & March, V. (1980). Informed consent-Why are its goals imperfectly realized? New England Journal of Medicine, 302, 896-900. Cook, T. D., & Campbell, D. T. (1979). Quasi-experimentation design and analysis issues for field settings. Chicago: Rand McNally College Publishing. Craig, T. J., & Abeloff, M. D. (1974). Symptomology among hospitalized cancer patients. American Journal of Psychiatry, 131, 1323-1327. DeVita, V. T., Hellman, S., & Rosenberg, S. A. (1982). Cancer: Principles and practice of oncology. Philadelphia: Lippincott. DeVita, V. T., Serpick, A. A., & Carbone, P. P. (1970). Combination chemotherapy in the treatment of advanced Hodgkin's disease. Annals of Internal Medicine, 73, 881-895. This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 206 RICHARDSON ET AL. Dunbar, J. M., Marshall, G. D., & Ho veil, M. F. (1979). Behavioral strategies for improving compliance. In R. B. Haynes, D. W. Taylor, & D. L. Sackett (Eds.), Compliance in health care (pp. 174-190). Baltimore: Johns Hopkins University Press. Feinstein, A. R. (1979). "Compliance bias" and the interpretation of therapeutic trials. In R. B. Haynes, D. W. Taylor, & D. L. Sackett (Eds.), Compliance in health care (pp. 297-308). Baltimore: Johns Hopkins University Press. Feinstein, A. R., & Ransohoff, D. F. (1976). Problems of compliance as a source of bias in data analysis. In L. Lasagna (Ed.), Patient compliance (pp. 65-76). New York: Futura. Frey, J., Frey, B. M., & Benet, L. (1979). Liquid chromatographic measurement of endogenous and exogenous glucocorticoids in plasma. Clinical Chemistry, 25, 1944-1947. German, P. S., Klein, L. E., McPhee, S. J., & Smith, C. R. (1982). Knowledge of and compliance with drug regimens in the elderly. Journal of American Geriatric Society, 30, 568-571. Glass, A., Wieand, H. S., Fisher, B., Redmon, C , Lerner, H., Wolfer, J., Shibata, H., Plotkin, D., Foster, R., Margolese, R., Wolmark, N., & other NSABP investigators. (1981). Acute toxicity during adjuvant chemotherapy for breast cancer: The National Surgical Adjuvant Breast and Bowel Project (NSABPK). Experience from 1,717 patients receiving single and multiple agents. Cancer Treatment Reports, 65, 363-376. Green, L. W. (1979). Educational strategies to improve compliance with therapeutic and preventive regimens: The recent evidence. In R. B. Haynes, D. W. Taylor, & D. L. Sackett (Eds.), Compliance in health care (pp. 157-173). Baltimore: Johns Hopkins University Press. Green, L. W., Kreuter, M. W., Deeds, S. G., & Partridge, K. B. (1980). Health education planning: A diagnostic approach. Palo Alto, CA: Mayfield. Haskell, C. M. (1980). Cancer treatment. Philadelphia: Saunders. Haynes, R. B. (1979). Determinants of compliance: The disease and the mechanics of treatment. In R. B. Haynes, D. W. Taylor, & D. L. Sackett (Eds.), Compliance in health care (pp. 49-109). Baltimore: Johns Hopkins University Press. Haynes, R. B., Taylor, D. W., & Sackett, D. L. (1979). Compliance in health care. Baltimore: Johns Hopkins University Press. Hulka, B. S. (1979). Patient-clinician interaction and compliance. In R. B. Haynes, D. W. Taylor, & D. L. Sackett (Eds.), Compliance in health care (pp. 63-77). Baltimore: Johns Hopkins University Press. Inui, T. S., Carter, W. B., Pecoraro, R. E., Pearlman, R. A., &Dohan, J. J. (1980). Variation in patient compliance with common long term drugs. Medical Care, 18, 986-993. Johnson, C. A., & Beardsley, R. S. (1978). Life style as a mediator of post-discharge adherence to a hypertension treatment regimen following an in-hospital behavioral modification program. Preventive Medicine, 7, 95. Kasl, S. V. (1975). Issues in patient adherence to health care regimens. Journal ofHuman Stress, 1, 5-17. Korsch, B. M., Gozzi, E. & Francis, V. (1968). Gaps in doctor-patient communication: Doctor-patient interaction and patient satisfaction. Pediatrics, 42, 855-871. Korsch, B. M., & Negrete, V. F. (1972). Doctor-patient communication. Scientific American, 227, 66-74. Leventhal, H., Zimmerman, R., &Gutmann, M. (1984). Compliance: A self regulation perspective. In D. Gentry (Ed.), Handbook of behavioral medicine (pp. 369-436). New York: Guilford. Ley, P. (1979). Improving clinical communications: Effects of altering doctor behaviour. In D. J. Oborne, M. M. Gruneberg, & J. R. Eiser (Eds.), Research in psychology andmedicine (Vol. 2, pp. 221-229). London: Academic. Ley, P., Bradshaw, P. W., Kincey, J. A., & Atherton, S. T. (1976). Increasing patients satisfaction with communications. British Journal of Social and Clinical Psychology, 15, 403-413. This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. PATH OF COMPLIANCE WITH CANCER THERAPY 207 Marston, M. V. (1970). Compliance with medical regimens: A review of the literature. Nursing Research, 19, 312-323. McKelvy, E. M., Gottlieb, J. A., Wilson, H. E., Haut, A., Talley, R. W., & Stephens, R. (1976). Hydroxyldaunomycin (adriamycin) combination chemotherapy in malignant lymphoma. Cancer, 38, 1484-1493. Mishel, M. H. (1980). Perceived ambiguity ofevents associated with the experience of illness and hospitalization. Development and testing of a measurement tool. Unpublished doctoral dissertation, Clarement Graduate College, Claremont, CA. Mishel, M. H. (1981). The measurement of uncertainty in illness. Nursing Research, 53, 258-263. Morisky, D. E., Levine, D. M., Green, L. W., Shapiro, S., Russell, R. P., & Smith, C. R. (1983). Five year blood pressure control and mortality. American Journal of Public Health, 73, 153-162. Patterson, P. P. (1981). Cancer patients and informed consent: A study of relationships among information, perceived ambiguity and stress during radiation therapy. Unpublished master's thesis, California State University, Los Angeles. Plumb, M. M., & Holland, J. (1977). Comparative studies of psychological function in patients with advanced cancer— 1. Self-reported depressive symptoms. Psychosomatic Medicine, 39, 264-276. Sackett, D. L. (1980). Hypertension: V. Compliance with antihypertensive therapy. Canadian Journal of Public Health, 71, 153-156. Sackett, D. L., & Snow, J. C. (1979). The magnitude of compliance and non-compliance. In R. B. Haynes, D. W. Taylor, & D. L. Sackett (Eds.), Compliance in health care (pp. 11-22). Baltimore: Johns Hopkins University Press. Santoro, A., Bonfante, V., & Bonadonna, G. (1982). Salvage chemotherapy with ABVD in MOPP-resistant Hodgkin's disease. Annals of Internal Medicine, 96, 139-143. Schein, P. S., DeVita, V. T., Hubbard, S., Chabner, B. A., Canellos, G. P., Berard C , & Young, R. C. (1976). Bleomycin, adriamycin, cyclophosphamide, vincristine, and prednisone (BACOP) combination chemotherapy in the treatment of diffuse histiocytic lymphoma. Annals of Internal Medicine, 85, All-422. Skarin, A. T., Canellos, G. P., Rosenthal, D. S., Case, D. C , Maclntyre, J. M., Pinkus, G. S., Moloney, W. C , & Frei, E. (1983). Improved prognosis of diffuse hisiocytic and undifferentiated lymphoma by use of high dose methotrexate alternating with standard agents (M-BACOD). Journal of Clinical Oncology, 1, 91-98. Smith, S. D., Rosen, D., Trueworth, R. C , & Lowman, J. T. (1979). A reliable method for evaluating drug compliance in children with cancer. Cancer, 43, 169-173. Taylor, S. E., Lichtman, R. R., & Wood, J. V. (1984). Compliance with chemotherapy among breast cancer patients. Health Psychology, 3, 553-562. Zimmerman, R. (1982, April). The dental practitioner-patient interaction: Dental health, expectations, and satisfaction. Paper presented at the meeting of the American Sociological Association, San Francisco. Zung, W. W. K. (1965). A self-rating depression scale. Archives o f General Psychiatry, 12, 63-70.