While conventional wisdom suggests that side effects are common, appear early in treatment, and are associated with treatment discontinuation, the current findings suggest otherwise. Although 41 of the most expected SSRI side effects (gastrointestinal, activating, sedating, cholinergic, and sexual symptoms) were reported as worsening by 4-27% of the sample that returned for the week 2 visit, none of the 55 symptoms queried by the SAFTEE were independently associated with attrition during the remainder of the 8-week acute treatment. Overall side effect frequency, severity, and burden as experienced by patients at week 2, based on either the SAFTEE burden item or the FIBSER, were also not related to attrition.
Although side effects have been linked to attrition,6, 10, 22, 28, 40
attrition in this study was independent of the prospectively measured presence, frequency, severity, or burden of side effects at week 2, the time typically associated with the appearance of side effects.24, 26, 27
This is despite 25% of participants reporting side effects of at least moderate intensity and 14% reporting at least moderate burden. There were few participants, however, who reported severe intensity or burden of side effects. Since bothersome or severe side effects have been associated with treatment discontinuation,10, 22
the low frequency of severe side effects in this sample could account for the failure to find an association between global severity or burden of side effects and attrition.
Since there was no difference in early symptom change between dropouts and completers, early lack of efficacy does not provide an alternate explanation for attrition in this sample.
There are a number of other possible explanations for the lack of association between side effects and attrition in this study, in contrast to findings from prior studies. First, side effects may be more prominent in decisions about continuing treatment for patients with limited concomitant disorders in randomized controlled trials since they may have less experience with treatment for chronic conditions and less tolerance for inconvenience. Second, studies often identify only one main reason for treatment discontinuation,22
such as side effects. It may be that patients consider multiple issues, weighing the benefits and psychological, physical, financial and other costs of treatment in making a decision about treatment continuation.41-44
Side effects may be one cost, but other perceived costs or benefits for individual patients are not typically measured in studies or discussed in clinical practice. Third, in this study the CRC inquired each week via a structured assessment5
about the likelihood of retention and discussion and resolution of patient specific barriers to retention likely occurred. There was also a high frequency of contact with a CRC including phone contact three times a week for 4 of the 8 weeks of the study, and use of an algorithm that specifically requires that side effects be addressed if indicated based on the FIBSER burden score each week. Increased contact, use of an individual such as a coordinator to support the patient, objective measurement and feedback to patients, and explicit focus on attrition are similar to interventions associated with increased adherence in collaborative care studies,45, 46
and these interventions may have mitigated the valence of adverse events for patients in this study and/or reduced attrition.
The trajectory of what happens to side effects after their emergence may also be more important to keeping patients in treatment than the appearance itself or the overall frequency, intensity, or burden of the side effects. A small prior study found that dropouts and completers had the same frequency of adverse events; however, the completers habituated to the side effects more rapidly than the dropouts.28
It may also not be the presence or burden of side effects that is related to discontinuation but the impact of the side effect on specific aspects of daily functioning.
Although adverse events are the rule rather than the exception,8
they may not be identified if clinicians do not ask about them directly. Adverse events often resolve in 2-4 weeks,27
although for many patients they do not, suggesting a strong need to identify and treat side effects aggressively when they do emerge or worsen early in treatment. It is therefore important to use a systematic approach to identifying them at each treatment visit, such as the use of a scale like the FIBSER so they can be treated when they do emerge or worsen early in treatment.
Weight loss was associated with attrition in analyses that did not control for confounding factors such as depression severity. It is possible that worsening of appetite (reported by 27%) and weight loss (reported by 16%) in this sample were related to worsening of depression rather than an adverse event in response to medication. Depression severity has been associated with attrition.4, 47
The few significant findings in the bivariate analyses may also have been due to chance.
There are several limitations of this study. We did not evaluate worsening or emergence of side effects or symptom change later than the week 2 visit. We also do not have side-effect ratings for participants who left the study between baseline and week 2, or for participants who missed their week 2 clinic visit. Only 12 participants dropped out between baseline and week 2. but they may have experienced severe intensity or burden of side effects. The distribution of side effects and/or their association with attrition may have been different with a different SSRI distribution, with antidepressants from different classes, or in a blinded clinical trial that includes a placebo. The frequency of side effects of moderate or higher intensity or burden is lower than that reported in other studies.28
We also do not have data in this study addressing patient perception of the costs and benefits of treatment or the relationship of adverse events to patient perception of these costs.
A further limitation of the study is having sufficient power to detect a moderate to small effect. For example, for a side effect that was relatively prominent, such as appetite decrease, which had an overall prevalence rate of 26.7% (in a sample of 206, 55 with appetite decrease and 151 without), there would be 80% power to detect a two-sided difference of 16%, 18% and 20%, assuming the rate of drop out in those without an appetite decrease was 10%, 15%, and 20%, respectively, with a type I error rate of .05. However, for a side effect that was relatively rare, such as drooling, which had an overall prevalence rate of 5.8% (in a sample of 206, 12 with drooling and 194 without), there would be 80% power to detect a two-sided difference of 31%, 34% and 36%, assuming that the rate of drop out in those without an appetite decrease was 10%, 15%, and 20%, respectively, with a type I error rate of .05.
In summary, this study does not support the commonly held belief that the presence, severity, or burden of side effects early in treatment with SSRIs is related to discontinuation of treatment, at least in patients who have frequent visits with their physician and are provided education and supportive medical management. By focusing on side effects, we may not be identifying the attitudes or issues of most importance to patients or how they are weighed in making a decision about treatment continuation. It may also be that what happens to side effects after their emergence is more related to dropout than the emergence/worsening, intensity, or burden of side effects early in treatment. A consistent approach to identifying side effects and aggressive management of side effects may be helpful in keeping patients in treatment long enough to recover. However, further assessment of how patients make decisions to remain in treatment is important.