Survivorship in Hodgkin lymphoma : childbearing and treatment-related disease

Abstract

Cancer is often considered a disease with, historically, poor survival that affects middle-aged and elderly individuals. Hodgkin lymphoma (HL) is a lymphatic malignancy that affects both young and old individuals, with the age-specific incidence curve having its first peak at ages 20-30 years. As survival has improved substantially over the last decades, there is an increasing number of survivors – some still at a young age. The primary purpose of this thesis was to address issues related to childbearing and treatment-associated disease among HL survivors. As a means to investigate these issues, novel statistical methods were developed and applied.

Childbearing in relation to HL: Some HL survivors will be at the beginning, or in the midst, of their childbearing years. Both pregnancy and HL are associated with changes to the immune system, making it plausible that pregnancy could affect the progression of the disease. Study I in this thesis aimed to answer if pregnancy affects the risk of relapse among female patients in remission from HL. Ever since the introduction of radio- and chemotherapy with the possibility to cure HL, the negative effects of therapy on fertility have been a concern. Contemporary treatments are believed to be less gonadotoxic than those previously used, but few studies have compared childbearing potential between the main treatment regimens administered today (ABVD and BEACOPP) in a real-world setting. In Study II, temporal trends in childbearing among female HL patients (of childbearing ages) were investigated, within groups of treatments, and in comparison with the general population. Both studies utilized a cohort of women diagnosed between 1992 and 2009, at ages 18-40 years, for whom detailed information on relapse as well as patient and disease characteristics was available. For the purpose of comparing childbirth rates with the general population, HL patients were individually matched to HL-free comparators. Childbirth rates were studied separately within two time windows during follow-up: 0-3 years and 3-7 years, and cumulative probability of childbirth was calculated in the presence of the competing risks of death and relapse.

No evidence to support the hypothesis of pregnancy-associated relapse was found. However, since the absolute risk of relapse is at its highest levels during the first 2-3 years after diagnosis, female HL patients could, if possible, be advised to delay childbearing to avoid co-occurence. Childbearing potential improved over calendar time, reflecting reduced toxicity and changes in counseling. Women treated during recent years had childbirth potential in line with that of matched comparators three years after diagnosis. Even women treated with BEACOPP, the most gonadotoxic chemotherapy, had an increasing cumulative probability of having children after HL. Importantly, no women had children after a relapse within the first seven years after diagnosis, which motivates a need for fertility advice and counseling at time of HL diagnosis.

Treatment-related disease: Late effects from cancer therapy are becoming increasingly important to quantify as the number of cancer survivors grows. Chemo- and radiotherapy used to treat HL increases the risk of cardio- and cerebrovascular diseases, and secondary malignancies (SMs). In Study III, excess incidence of diseases of the circulatory system (DCS) among HL patients was investigated to describe temporal trends in DCS morbidity attributable to HL and its treatment. Data on patients diagnosed with HL between 1985 and 2013 at ages 18-80 years, for whom information on inpatient DCS records was available, was used. Relative survival methods were applied to estimate excess incidence rates indirectly from the observed and expected rates of DCS. Cumulative excess incidence of DCS was calculated in the presence of competing risks. The treatment-related incidence of DCS declined between the mid-1980s and mid-1990s, after which no substantial improvements were observed. The risk of a treatment-related DCS persists for up to 10 years among patients who completed their treatment in the new millennium.

When studying late effects it is important to attempt to capture the additional disease incidence associated with cancer treatment. Additionally, to gain insight in real-life risks, it is of interest to study not only time to first event, but continue to follow patients as they experience different types of late effects before reaching an absorbing state. Doing both of these simultaneously requires estimating excess transition rates to transient states in a multi-state model framework, for which no methods have existed. Study IV suggested a way to achieve this, using a recently developed simulation strategy to predict transition probabilities. As an illustrative example, data from Study III on HL patients and DCS incidence was used. Combining methods from relative survival with the multi-state framework enables investigation of complex patient pathways and can be useful for several applications related to survivorship among cancer patients.