The Cutting Edge
RECURRENCE RISK IN MAJOR DEPRESSION
Article first published online: 5 DEC 2012
© 2012 Wiley Periodicals, Inc.
Depression and Anxiety
Volume 30, Issue 1, pages 1–4, January 2013
How to Cite
Patten, S. B. (2013), RECURRENCE RISK IN MAJOR DEPRESSION. Depress. Anxiety, 30: 1–4. doi: 10.1002/da.22030
- Issue published online: 8 JAN 2013
- Article first published online: 5 DEC 2012
Major Depressive Disorder (MDD) is typically regarded as an episodic condition characterized by recurrent (albeit sometimes chronic) major depressive episodes (MDEs). Understanding patterns of recurrence is extremely important for clinical practice and public health, but this topic has been relatively neglected in the research literature. A classic paper defining the concept of recurrence noted, in 1991, that “few published studies” were available to validate the definitions provided. A review published nearly 20 years later referred to such studies as “scarce.” Due to this paucity of research, decisions about ongoing or long-term treatment cannot be firmly supported by evidence. Clinical decisions that are related to prognostic information (e.g. weighing the risks and benefits of long-term treatment) are difficult to make in the absence of clear information. In contrast, in the cancer field, tools for calculating prognosis are readily available (http://www.lifemath.net/cancer/) and patients can consult such tools while considering their options. Such algorithms are not yet available for MDD.
An influential attempt to clarify terminology surrounding the longitudinal course of MDD was produced by a task force of the MacArthur Foundation Research Network on the Psychobiology of Depression. Here, recurrence was defined as re-emergence of the syndrome of MDE during a period of recovery, with recovery being defined as a sustained asymptomatic state following a period of remission. In this classification, emergence of MDE during the remission period, but prior to the recovery period, would be considered a relapse rather than recurrence. With reference to Research Diagnostic Criteria, the MacArthur Foundation task force asserted that ≥8 weeks of symptom-free status following ≥2 to <8 (asymptomatic) weeks of remission constituted recovery. Since then, a point in time 4 months after achievement of remission was favored by another task force as a definition for the start of the recovery interval, corresponding to the upper limit of the MacArthur Foundation definition.
The definitions put forward by these task forces were driven by a belief that the distinction between relapse and recurrence reflects a distinction between symptomatic expression of an ongoing episode (that had perhaps been temporarily suppressed by treatment or other factors) and the emergence of a new episode. The MacArthur Foundation task force hypothesized that the distinction could be supported empirically by documenting a “point of rarity”  presumed to occur after the usual duration of a prior episode but before the usual onset of another one. This reflects a conception of MDE duration that can also be found in DSM-IV, where it is stated that MDE “typically” lasts about 4 months. This kind of statement implies that episodes lasting much less than 4 months (e.g. a few weeks) or more than 4 months (e.g. a few years) are less common or less typical than episodes lasting about 4 months. If one imagines a bell-shaped curve representing a distribution of episode durations with a central tendency around 4 months, then one can predict a “point of rarity” as posited by the task force. However, MDE duration does not appear to have a central tendency of this type. Most episodes are brief and a few are very long[5, 6], the mean duration represents an average that is strongly influenced by the very long episodes. A “point of rarity” does not appear to exist, or at least has not been confirmed to exist. In this sense, the distinction between relapse and recurrence, as originally conceived, increasingly appears to be an arbitrary distinction. Nevertheless, the basic clinical question: “what is the risk that another episode will occur?” remains just as salient as ever. How can the epidemiologic literature, at its current stage of development, help to inform the answer to such questions?
The majority of psychiatric epidemiologic studies have used a cross-sectional design and have focused on estimating prevalence. Prevalence, however, merely provides a snapshot of an underlying dynamic in a population. Prevalent cases are members of a prevalence “pool” within the population, a pool that has an inflow (incidence, relapse, and recurrence) and an outflow (remission and death). Clinical interventions that diminish the prevalence of disease in populations do so by increasing the outflow from the prevalence pool (e.g. increasing the rate of remission and recovery among cases) and by reducing the rate of relapse and recurrence among remitted and recovered cases. Epidemiologic and economic analyses indicate that longer term strategies aimed at reducing recurrence are likely to be the most effective way to reduce the population burden of MDD .
DSM-IV divides cases of MDD into two categories: single episode and recurrent. Most population-based cross-sectional studies have reported that a majority, but not necessarily an overwhelming majority, of lifetime MDD cases are recurrent, about 75% according to one data synthesis. However, such statistics can be deceptive. Lifetime prevalence is the proportion of a population that has met diagnostic criteria for MDD during their lifetime up to the date of assessment. This is a difficult parameter to interpret. Lifetime prevalence has been described as a “cumulative incidence proportion among survivors” or CIPAS, a term that emphasizes the complex nature of this parameter. Even though a sizable proportion of people with lifetime MDD report only a single episode, many of these will experience subsequent recurrences during their lifetime. There is also growing evidence that many episodes are not recalled later. Through these mechanisms, it follows that recurrent cases can be expected often to be misclassified as single-episode cases by cross-sectional studies. It is safe to assume that a very large majority of people with MDD will experience a recurrent course.
A distinction should be drawn between a recurrence risk and a recurrence rate. A recurrence risk is a probability that is estimated as a frequency or proportion over a defined time interval. Mathematically, a more flexible way expressing the probability of recurrence is through the use of a rate. Similar to other familiar rates (e.g. velocity), and unlike an incidence proportion, incidence rates have units that are expressed in terms of person-time at risk, for example as 1/year or per year. A recurrence rate refers to an instantaneous probability of transition to an active episode from a recovered state in the same way that velocity refers to an instantaneous change in distance per time unit. This approach is flexible in the sense that a rate-based approach, given certain assumptions, can be used to predict the probability (risk) of recurrence over any desired time period. For example, a patient trying to decide whether to opt for long-term antidepressant treatment may want to know the risk of recurrence over 1 year, 2 years, 3 years (and so on). When a recurrence rate is known, it is possible to calculate this using the equation: cumulative risk of recurrence (over “t” years) = 1 − exp(−rt) where “r” is the recurrence rate defined with per years units and “t” is the number of years in question. The function “exp” refers to the inverse of the natural logarithm, as determined using a calculator or spreadsheet. For example, using a prediction algorithm the PredictD study identified scenarios associated with various annual risks of recurrence. In one recurrence scenario, this was 7% (see  for a description of the scenario). Taking this as an annual rate estimate (0.07/year), a patient's 5-year risk can be estimated as 1 − exp(−0.07 × 5) or 30%. Another approach is to consider each of the 5 years as a series of separate 1-year intervals, in which case the 5-year risk can be estimated as: cumulative incidence (over “n” successive intervals) = 1 − (1 − annual incidence)n, or, in this case as: 1−(1 − 0.07)5. This formula treats the 5-year cumulative incidence as the complement of the probability of remaining well (1 − 0.07 = 0.93) over five successive intervals and leads to approximately the same 5-year cumulative incidence proportion in this specific example. Such strategies may be helpful for engaging patients in clinical decisions. This kind of approach will be increasingly useful as the epidemiologic literature provides more specific and accurate estimates of recurrence risks and rates for MDD. Thus far, unfortunately, the literature has been inconsistent and there is a danger that such quantitative approaches can suggest a false degree of accuracy.
A recent systematic review of recurrence studies noted considerable heterogeneity in estimates both of the prevalence and determinants of recurrence. Nevertheless the review offered estimates of the risk of recurrence in several different populations. For example, 60 to 85% recurrence proportions were reported after follow-up intervals of 5–15 years among patients seen in specialized mental health centers. A 35% recurrence proportion was reported after 15 years in the single general population study included in the review (the Baltimore Epidemiologic Catchment Area Follow-Up). According to the formula presented above (rearranged to solve for the recurrence rate rather than the cumulative incidence), the general population cumulative recurrence probability of 35% over 15 years would roughly suggest a 3%/year recurrence rate. The most consistent determinants of recurrence were the number of past episodes and the presence of residual subthreshold symptoms. With each past episode, the risk recurrence may increase by as much as 15% to 18%.
The most well-known study of the long-term course of MDD is the 15-year follow-up of the NIMH Collaborative Program on the Psychobiology of Depression. An often-cited recurrence proportion of 85% over 15 years derives from this study. Assuming a constant incidence rate, this cumulative proportion would represent a relapse rate of 0.13/year, much higher than the general population estimate cited above. The recurrence rate was highest in the years immediately following an index episode and diminished with an increasing duration of recovery. As a result, the median time to recurrence was only 132 weeks, about 2.5 years.
An aspect of predicting recurrence risk that has been almost entirely ignored is the role played by episode characteristics. Major depression is known to be heterogeneous in terms of episode characteristics. One study recently examined patterns of recurrence according to various features of a baseline episode: subthreshold symptom levels (versus meeting the MDE threshold); episode duration longer than 4 weeks; and whether the episode was associated with high distress, restriction of activities, or suicidal ideation. Among those with no MDE and no subthreshold depression at a baseline, there was an 18% cumulative incidence of MDE over 14 years of follow-up. This corresponds roughly to an episode incidence rate of 0.014/year. When the episodes met diagnostic criteria for MDE, lasted at least 4 weeks and were associated with high distress and/or restriction of activities and/or suicidal ideation the recurrent proportion was 71% (a recurrence rate of approximately 0.09/year). A proviso to all of these rate calculations is that the tendency of recurrence rates to decline with time since a prior episode means that these rates are likely to underestimate the true rate in the early years after an episode and over estimate it in later years.
Although one may expect that increasing availability of longitudinal data will lead to better prediction and more informed clinical decisions around the issue of recurrence, it is also true that certain lines of evidence are beginning to cast some doubt on the idea that MDD is fundamentally a recurrent condition at all. It has been reported that more than 90% of MDEs in community populations are preceded by a loss or other apparent triggering event. Such observations suggest a diathesis-stress conceptualization of MDD. Early life experiences, through epigenetic mechanisms, may inactivate genes that contribute to downregulation of stress responses, providing a candidate mechanism for diathesis in addition to the long-recognized genetic contribution. Dannlowski et al. reported that amygdala responsiveness to threat-related facial expression was higher in subjects with a history of childhood trauma. They described this abnormality as a “limbic scar” because it did not correlate with current depressive symptom levels. These findings suggest a model whereby interactions between environmental stressors and exaggerated stress-responsiveness may account for periodic increases and decreases in depressive symptoms even in the absence of any inherently or fundamentally episodic underlying pathophysiologic process.
Whether MDD is best conceptualized as an intrinsically recurrent condition or as a persisting abnormality that produces variable manifestations, there is can be no doubt of the clinical value of quantifying recurrence risk in clinical practice. The science of epidemiology provides a set of tools for understanding and manipulating recurrence risk and rate data. However, the existing literature does not yet adequately support the application of such tools in clinical practice. Additional longitudinal data are urgently needed.
- 4American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders (DSM-IV-TR). Washington: American Psychiatric Association; 2000.