The treatment of primary breast cancer usually consists of surgery often followed by adjuvant therapy (radiotherapy, chemotherapy, hormonal treatment, etc.) to reduce the risk of recurrence. The cancer diagnosis and the treatments may have significant impact on the patients' quality of life. This thesis deals with scientific aspects and clinical results of a study aimed at assessing the impact of breast cancer (and its treatment) on the patients' quality of life. Studies such as this assessing the problems and symptoms experienced by the patients are often referred to as health-related quality of life (HRQL) research. HRQL research deals with subjective experiences and raises challenging, scientific questions. Therefore, much attention was directed towards methodological issues in this clinically motivated project. The study was a prospective, longitudinal, questionnaire-based investigation of women with newly diagnosed breast cancer registered in the Danish Breast Cancer Co-operative Group's DBCG 89 Program. The patients were sub-divided into low-risk and high-risk patients. High-risk patients were offered randomisation in one of three randomised adjuvant therapy trials involving chemotherapy, ovarian ablation, and endocrine therapy. After a literature study and interviews with breast cancer patients, a questionnaire was composed that included two widely used standard questionnaires (EORTC QLQ-C30 and Hospital Anxiety and Depression (HAD) Scale) and a DBCG 89 Questionnaire developed for this study. A total of 1,898 eligible patients were invited by post to participate in the study involving six assessments over a 2-year period, and 1,713 patients (90%) completed the first questionnaire. Furthermore, a questionnaire was sent to 872 women selected at random from the general population; 608 (70%) responded. The multi-item scales of the two standard questionnaires were analysed for so-called differential item functioning (DIF) in order to investigate whether the (summary) scale scores were adequate representations of the information obtained by the individual items. The DIF analyses identified a number of cases of DIF, which, among other things, contributed to detection of possible problems in the HAD Scale. It was concluded that DIF analyses are relevant when important analyses based on multi-item scales are made. A new way to evaluate the validity of questionnaires was developed. The results from questionnaires completed by patients were compared against results from open ended interviews with the same patients rated by observers. The idea was that if results were similar, the patients had then probably understood and completed the questionnaire items as intended. On the other hand, if results from self-assessment and interviews deviated, misunderstandings or other errors might have taken place, and the study would give insight into possible problems. Of 57 breast cancer patients, 46 (81%) were successfully interviewed. In general, the agreement between patient-completed questionnaires and interviews was excellent, indicating very good validity. The median weighted kappa for the EORTC QLQ-C30 was 0.85 (range 0.49-1.00); it was 0.79 (range 0.65-0.95) for the HAD Scale, and 0.92 (range 0.51-1.00) for the DBCG 89 Questionnaire. However, the study identified a mechanism called selective reporting, which may affect results from most HRQL questionnaires: in order to provide correct and useful answers some patients do not report symptoms they believe are irrelevant to the study, e.g., symptoms unrelated to cancer. This mechanism may lead to bias if results from patients are compared to results from populations reporting their symptoms more completely, e.g., general population samples. In contrast, this mechanism has little importance when results from different sub-groups of cancer patients are compared. In this study multiple variables were assessed at multiple points in time and we did not have a priori hypotheses for all these potential comparisons. Therefore, a staff survey involving experienced doctors and nurses was conducted in order to generate hypotheses that could be tested in the data from patients. We contacted 46 health care professionals and 36 (78%) responded. Overall, the staff survey did not prove very useful for the intended purpose. The main reason for this was probably that the health care professionals had limited insight into the patients' HRQL. A different approach to the problem of multiple hypothesis testing proved more useful. Hypotheses generated from the initial literature review were tested in the comparison of patients in chemotherapy against patients not in chemotherapy. The study of women selected at random from the general population showed that these women experienced a considerable degree of "morbidity" according to all three questionnaires. This shows that symptoms and problems reported by cancer patients may have causes other than cancer, and thus constitutes a good justification for the use of data from general population studies when interpreting data from cancer patients. The levels of anxiety and depression of low-risk breast cancer patients were found to be lower than those from the general population sample. After careful consideration we concluded that this finding was probably incorrect. The most important explanations were thought to be the wording of some HAD Scale items as well as two mechanisms that are not specific to the HAD Scale, the "selective reporting mechanism" found in the validation study, and the response-shift problem. These findings indicate - in contrast to the conclusion above - that the comparability of HRQL data from cancer patients and general population data must be questioned. However, as this is the first study to raise the problem, this issue needs further investigation. Based on the initial literature review and interviews we hypothesised that 30 different HRQL issues would be impaired in patients undergoing CMF chemotherapy compared to patients not in chemotherapy; 23 of these hypotheses were confirmed. In addition, our study and other research suggest that other HRQL aspects may also be affected by chemotherapy. Thus, there is considerable evidence that patients in chemotherapy may experience effects on a wide spectrum of HRQL issues. Most other studies have assessed surprisingly few of the HRQL issues shown in our study to be impaired in patients receiving chemotherapy. Similarly, current review articles on HRQL effects of adjuvant chemotherapy mention only relatively few of these topics. Concerning HRQL after the treatment period, our main finding was that many symptoms and problems had declined or disappeared, but some persisted: anticipatory nausea, weight gain, endocrine effects (e.g., hot flushes/sweats, irregular bleedings/amenorrhea, vaginal dryness), disturbed sleep, and sexual dysfunction. These findings are in agreement with the literature. The staff study showed that experienced physicians and nurses did not expect many of the "scientifically well documented" consequences of chemotherapy. Taken together, our findings suggest that information to patients about chemotherapy should be more comprehensive than that which has been practised in most places. When compared against ovarian ablation, chemotherapy was associated with more impact on HRQL during the treatment period; only hot flushes/sweats were more pronounced in the ovarian ablation group. Thus, from an overall "HRQL perspective" ovarian ablation or suppression may be preferable. However, younger women may preserve their premenopausal status (including fertility) by having chemotherapy, and this may be an argument for chemotherapy or for temporary ovarian ablation via goserelin, rather than permanent ovarian ablation. Furthermore, while ovarian ablation/suppression may be preferable because of less impairment of HRQL, contemporary chemotherapeutic regimens may be more effective. These results indicate that for some patients, the HRQL data and results on treatment efficiency may be in conflict. There is no simple, universally correct solution to this dilemma. More research into patients' views and expectations to the health-care system in cases where medical decision-making involves complex trade-offs between treatment efficiency and HRQL issues is needed. Contrary to expectations, the analyses showed that fatigue and emotional function predicted the risk of recurrence and death independently of biological and clinical prognostic variables. In multivariate Cox regression analyses patients who were more fatigued or had poorer emotional function had a worse prognosis. These results are consistent with one small study, but are inconsistent with five similar studies in patients with primary breast cancer, which found no such associations. The reasons for these important differences are currently unknown. In conclusion, this study consisted of methodological and clinical investigations of HRQL in primary breast cancer patients. The initial questionnaire development resulted in a combination of questionnaires that was more comprehensive than in other similar studies. The results of the methodological studies generally supported the validity of the questionnaires but also gave important insights into potential scientific problems that are probably not restricted to the present study. These insights helped to prevent misinterpretations of the clinical data. The study provided the most detailed description of HRQL during and after breast cancer adjuvant chemotherapy to date, and compared results of chemotherapy against ovarian ablation. It also provided controversial results concerning the prognostic value of HRQL data. The combination of a large empirical study and several methodological sub-studies thus proved useful and gave new results.