Remission in CBT for adult anxiety disorders: A meta-analysis

Clinical Psychology Review 61 (2018) 1–8

Contents lists available at ScienceDirect

Clinical Psychology Review journal homepage: www.elsevier.com/locate/clinpsychrev

Review

Remission in CBT for adult anxiety disorders: A meta-analysis a

a

Kristen S. Springer , Hannah C. Levy , David F. Tolin a b

a,b,⁎

T

Anxiety Disorders Center, Institute of Living, Hartford Hospital, 200 Retreat Avenue, Hartford, CT 06106, United States Department of Psychiatry, Yale University School of Medicine, 300 George St. New Haven, CT 06511, United States

H I GH L IG H T S

is no universally accepted definition of remission in anxiety disorders. • There conducted a meta-analysis of remission in cognitive-behavior therapy (CBT). • We overall remission rate for the intent-to-treat samples was 48% at post-treatment and 54% at follow up while the remission rate for completer samples was • The 53% at post-treatment and 56% at follow up. rates differed by definition, diagnosis, and various patient factors. • Remission • Although CBT is effective for anxiety, there is significant room for improvement.

A R T I C LE I N FO

A B S T R A C T

Keywords: Anxiety CBT Remission Meta-analysis PTSD OCD

Currently there is no universally accepted definition of remission in anxiety disorders. This may be causing significantly different estimates of treatment efficacy across anxiety disorders. The aim of this paper was to determine not only the overall remission rate in cognitive-behavioral therapy (CBT) for anxiety disorders, but also to examine whether the different definitions of remission lead to significantly different remission rates. From the initial 228 abstracts reviewed by the authors, 100 articles were retained. The overall mean remission rate was 51.0%. Remission rates were highest when remission was defined as good end state functioning or no longer meeting criteria for the primary diagnosis. Studies of posttraumatic stress disorder had the highest remission rates, while those of obsessive-compulsive disorder and social anxiety disorder had the lowest remission rates. Rates of remission differed by certain demographic (e.g., older age) and clinical (e.g., medication use) characteristics. Although CBT is an empirically supported treatment for anxiety disorders, it is clear that there is room for improvement, as many patients do not achieve remission status.

1. Introduction The efficacy of cognitive behavioral therapy (CBT) for anxiety disorders is well established with multiple controlled trials demonstrating that CBT outperforms not only waitlist and placebo controls but also other psychological treatments (Butler, Chapman, Forman, & Beck, 2006; Tolin, 2010). Meta-analytic reviews of CBT for anxiety disorders have commonly found medium to large effect sizes from pre- to posttreatment, depending on the anxiety disorder (Butler et al., 2006; Olatunji, Cisler, & Deacon, 2010). A recent meta-analysis of response rates in CBT for anxiety suggest a less optimistic view of CBT efficacy; there was a 49.5% response rate at post-treatment and a 53.6% response rate at follow-up (Loerinc et al., 2015). One complication with interpreting the Loerinc et al. (2015) metaanalysis is that the constructs of response and remission were

⁎

combined. Therefore, we examined how different definitions of remission lead to different remission rates among the anxiety disorders. Response refers to overall change over the course of treatment (e.g., a patient exhibited a 25% or greater reduction in symptoms). However, a patient could exhibit clinical response but still be quite impaired. For example, a patient with obsessive-compulsive disorder (OCD) might show a decrease in Yale-Brown Obsessive-Compulsive Scale (Y-BOCS; Goodman et al., 1989) scores from 32 to 20; although that is more than a 25% decrease (and thus the patient may be considered a responder), the post-treatment score is still in the moderate range, suggesting clinically elevated symptoms. Remission, on the other hand, refers to end status after treatment; in other words, a patient in remission is considered “well.” This is a critical distinction: in order to determine how well CBT impacts the substantial public health burden of anxiety disorders (Kessler, Alonso, Chatterji, & He, 2014), it is important to

Corresponding author at: Anxiety Disorders Center, Institute of Living, Hartford Hospital, 200 Retreat Avenue, Hartford, CT 06106, United States. E-mail address: [email protected] (D.F. Tolin).

https://doi.org/10.1016/j.cpr.2018.03.002 Received 30 June 2017; Received in revised form 14 March 2018; Accepted 14 March 2018 Available online 16 March 2018 0272-7358/ © 2018 Elsevier Ltd. All rights reserved.

Clinical Psychology Review 61 (2018) 1–8

K.S. Springer et al.

methodological quality factors impacted reported remission rates.

know not only how many patients show a change in symptoms, but also how many can be considered “well” after treatment. Remission has been defined in several different ways. One common definition is that the patient no longer meets diagnostic criteria for the disorder being treated. An important limitation of this definition is that a patient may no longer meet criteria for a diagnosis simply because he/ she is one symptom short of the criteria, yet continue to have severe and impairing symptoms. Another definition of remission is scoring below a clinical cutoff on a continuous measure. However, some researchers have used cutoff scores that could still reflect the presence of mild to moderate symptoms (e.g., a Y-BOCS score of 15 or less). A third and related definition is clinically significant change (CSC; Jacobson & Truax, 1991), which can reflect post-treatment scores that are outside the pathological range, within the normative range, or past the midpoint between the pathological and normative ranges (McGlinchey, Atkins, & Jacobson, 2002). It is noted that some authors have confused CSC with reliable change (i.e., a degree of change that is unlikely to be due to measurement error); however, these constructs should be considered separately, as reliable change refers to movement over time (analogous to the construct of response) whereas CSC refers to end status (analogous to the construct of remission; Jacobson, Roberts, Berns, & McGlinchey, 1999). A fourth definition of remission is good (or high) end state functioning, a less well-defined construct that refers to overall psychological health across multiple domains of functioning, such as anxiety, depression, and general stress. Practically speaking, good end state functioning is often defined in treatment studies as scoring below cutoffs on measures not only of the disorder being treated, but also on measures of commonly co-occurring problems such as depression. As can be seen from the wide range of remission definitions, as well as the fact that some authors have defined remission as a combination of two or more of these factors, it is clear that there is no universally accepted definition of remission in anxiety disorders, which may lead to significantly different estimates of treatment efficacy. It is important, therefore, to determine not only the overall remission rate in CBT for anxiety, but also to understand whether different definitions of remission lead to significantly different estimates. Of course, universally accepted and applied definitions are needed in order to clarify true remission rates in CBT. To this end, the present meta-analysis examined rates of remission (defined as end state after treatment) in CBT for DSM-IV (American Psychiatric Association, 1994) anxiety disorders. Our primary aim was to compare remission rates across various definitions, including no longer meeting criteria for the primary diagnosis, CSC, scoring below a clinical cutoff, good end state functioning, and commonly used combinations of these definitions. A secondary aim was to compare remission rates across anxiety disorders. We also examined whether certain demographic variables (e.g., age, gender) and psychological factors (e.g., medication status, comorbidity) impacted remission rates. In addition to establishing standardized definitions for remission, it is important to consider the methodological quality of the research. One might expect that key indices of methodological rigor (e.g., Foa & Meadows, 1997; Jadad et al., 1996), would affect remission estimates, with more rigorous trials yielding more conservative estimates. Accordingly, randomized controlled trials (RCTs) would be expected to yield lower remission rates than would open trials. Intent-to-treat (ITT) analyses would be expected to yield lower rates than would completer analyses. Studies that used independent evaluators, blind to treatment condition, would be expected to yield lower rates than would studies that used unblinded evaluators or self-report measures. Studies that used manualized treatments and assessed treatment fidelity would be expected to yield lower remission rates than would those that did not. Finally, studies that employed reliable and valid measures for the disorder being treated would be expected to yield lower remission rates than would those that used non-validated measures (e.g., daily diaries). In the present meta-analysis, we investigated whether these

2. Method 2.1. Data sources Journal articles were identified using searches of the PsycINFO and Medline electronic databases from January 2000 through February 2018 in order to keep the literature current. Other meta-analyses have used similar time frames (e.g., Loerinc et al., 2015). The following search terms were used: (Cognitive Behavior Therapy or CBT or Cognitive Therapy or Behavior Therapy or Exposure Therapy) and (Remission or Recovery or Clinically Significant Change or Good End State or High End State) and (Anxiety or Anxiety Disorders or Obsessive-Compulsive Disorder or OCD or Posttraumatic Stress Disorder or PTSD or Phobia or Phobic Disorders). Searches were limited to English-language articles and adult populations. Studies were also identified through the reference lists of originally obtained articles, review papers, and meta-analyses. 2.2. Study selection and data extraction We included studies that met the following inclusion criteria: participants were adults (age 18 years and older) who met DSM-IV criteria for an anxiety disorder; that it was a treatment study (randomized controlled trial or open trial) for a specific DSM-IV anxiety disorder, in which at least one treatment condition was CBT monotherapy; and a number or percentage of participants remitted was reported. We excluded samples that combined CBT with other treatments (e.g., medications), web-based and self-help interventions, concurrent treatment (e.g., concurrent CBT for co-occurring anxiety and substance use), previously failed treatment (e.g., participants received CBT after failing to respond to pharmacotherapy), and studies that did not require medication stabilization prior to treatment. For articles that did not specify medication stabilization, we attempted to contact the authors for clarification; if they did not respond, we retained the study for analyses. Articles reporting secondary analyses of a data set already included in the meta-analysis were excluded. The first and second author independently reviewed all abstracts from the initial search and coded whether or not they met initial inclusion criteria. Any disagreements were resolved by mutual discussion until 100% agreement was reached. This process resulted in 228 abstracts that met initial inclusion criteria (see Fig. 1). Full-text articles were obtained for these abstracts and divided between the first and second author, who independently coded their assigned articles. From the initial 228 abstracts, 100 full-text articles (see Appendix for full list) were retained for analyses and consisted of more completer samples than ITT samples, representing a total of 3278 ITT patients and 4129 completer patients.

Fig. 1. Selection of articles for meta-analysis. k = number of studies.

2

Clinical Psychology Review 61 (2018) 1–8

K.S. Springer et al.

2.3. Coding variables

2.5. Statistical analyses

2.3.1. Remission criteria and measurement properties We included the number of remitted and non-remitted patients at post-treatment and follow-up. The average length of time at follow-up was 6.59 months. The following remission criteria were included: 1) no longer meeting criteria for the primary diagnosis; 2) good end state functioning, defined as scoring below threshold on multiple dimensions of mental health including but not restricted to the primary diagnosis; 3) scoring below a clinical cut-off; 4) CSC, defined as demonstrating reliable change and scoring below a clinical cutoff; 5) panic-free (i.e., no panic attacks for a specified amount of time); 6) combination of a certain decrease and scoring below a clinical cutoff on a measure; and 7) combination of scoring below a clinical cutoff and no longer meeting criteria for the primary diagnosis. If a cutoff was used, we coded the cutoff type (very mild, mild, or moderate). If there was a follow-up period, we included its length (in months). We also included the number of measures and different measurement modalities used to determine remission.

We used random-effects models with Comprehensive Meta-Analysis software (Borenstein et al., 2009). Study was the unit of analysis, resulting in pooled event rates, except for measure-specific questions, in which the measure was the unit of analysis. Because some remission criteria categories yielded small numbers of studies (e.g., combination of scoring below a clinical cut-off and no longer meeting criteria for the primary diagnosis, k = 2), we analyzed the measure-specific questions two ways: first, using the original remission criteria categories, and second, using new categories that combined related remission criteria together (e.g., all studies that included scoring below a clinical cutoff were classified as such, even if other criteria were also included; this led to some studies being examined in more than one comparison). Remission rates were weighted by inverse variance (therefore, larger samples were weighted more strongly than were smaller samples). Publication bias was examined using the Trim and Fill procedure (Duval & Tweedie, 2000), which estimates missing studies to the left side of the mean event rate (indicating lower remission rate). Withinand between-group heterogeneity was examined using the Q statistic and its p value. Within-group heterogeneity was further examined using the I2 statistic, which reflects the percentage of variation due to true heterogeneity rather than chance and is interpreted as follows: 25% = little heterogeneity, 50% = moderate heterogeneity, and 75% = high heterogeneity (Higgins, Thompson, Deeks, & Altman, 2003).

2.3.2. Quality criteria We coded the following variables, in addition to ITT vs. completer analyses, to evaluate study quality: 1) independent rater blinded to treatment condition was used to determine remission (yes or no); 2) type of treatment study (randomized controlled trial or open trial); 3) reliable and valid measure (was the measure used to determine remission reliable and valid for the disorder being assessed, yes or no); 4) treatment fidelity (was fidelity assessed, yes or no), and 5) treatment manual (was a manual used, yes or no).

3. Results Across all analyses, the mean remission rate was 51.0% (95% CI = 47.8%–54.2%) at post-treatment and 55.1% (95% CI = 51.0%–59.1%) at follow-up. ITT and completer samples were not associated with significantly different overall mean remission rate at post-treatment (47.9% vs. 53.0%, Q = 2.51, p = 0.11) or at follow-up (53.6% vs. 56.1%, Q = 0.39, p = 0.53) (see Table 1). The Trim and Fill procedure suggested that no ITT studies were missing at post-treatment; however, at follow-up, seven missing studies imputed by the procedure decreased the estimated mean remission rate to 46.1% (see Fig. 2). At post-treatment there was significant between-group heterogeneity in the ITT sample (Q = 9.85, p = 0.05) and in the completer sample (Q = 15.70, p = 0.008) across diagnoses. In both the ITT and completer samples, PTSD and GAD studies were associated with higher remission rates, while OCD and SAD studies were associated with lower remission rates (see Table 2). Heterogeneity across diagnoses was also significant at follow-up in the completer sample (Q = 11.12, p = 0.02), though not in the ITT sample (Q = 5.93, p = 0.20). Again, PTSD and GAD studies were associated with higher remission rates, while OCD and SAD studies were associated with lower remission rates. When using the original remission criteria categories, betweengroup heterogeneity was significant for the ITT sample (Q = 19.40, p = 0.007) and for the completer sample (Q = 19.78, p = 0.006) according to the remission criteria used. As shown in Table 3, in the ITT sample reported remission rates were highest in studies defining remission as no longer meeting criteria for the primary diagnosis or good

2.3.3. Sample characteristics The following sample characteristics were coded: 1) principal anxiety disorder treated, including social anxiety disorder (SAD), posttraumatic stress disorder (PTSD), obsessive-compulsive disorder (OCD), specific phobia (SP), generalized anxiety disorder (GAD), and panic disorder with or without agoraphobia (PD/A); 2) age, in years; 3) percent female; 4) percent of ethnic/racial minority participants; 5) percent taking psychiatric medications; 6) percent taking antidepressant medication; 7) percent taking benzodiazepine medication; 8) percent with comorbid depressive disorder; and 9) percent with comorbid substance use disorder.

2.4. Reliability Eleven articles (11%) were randomly selected from the pool of included studies for inter-rater reliability analyses. The first and second author independently coded each of these articles, and then intraclass correlation coefficients and kappa coefficients were calculated depending on whether the variable was continuous or categorical, respectively. All values were 0.80 or higher, indicating strong inter-rater reliability across variables.

Table 1 Mean remission rates for intent-to-treat and completer samples. Sample

Intent-to-treat Completer

Post-treatment

Follow-up 2

k

Mean

95% CI

Q

I

k

Mean

95% CI

Q

I2

52 95

47.9% 53.0%

43.4%–52.4% 48.6%–57.4%

285.60⁎ 608.15⁎

82.14 84.54

31 66

53.5% 56.1%

47.6%–59.3% 50.4%–61.6%

182.21⁎ 399.15⁎

83.54 83.72

Note. k = number of samples. CI = confidence interval. ⁎ p < 0.05.

3

Clinical Psychology Review 61 (2018) 1–8

K.S. Springer et al.

remission as panic-free status and were lowest in studies defining remission as scoring below a cutoff score. At follow-up, between-group heterogeneity was also significant for the ITT sample (Q = 21.34, p = 0.002) and for the completer sample (Q = 18.20, p = 0.011) according to the remission criteria used. As shown in Table 3, in both the ITT and completer samples, the number of studies using various definitions of remission was small, making interpretation difficult. We therefore combined the definitions into broader categories for clarity: (1) panic-free status (for studies of PD/ A), (2) scoring below a cutoff (with or without additional requirements), (3) no longer meeting criteria for the primary diagnosis, and (4) good end state functioning. When using these simplified remission definitions, at post-treatment, between-group heterogeneity was significant for the ITT sample (Q = 10.39, p = 0.034), though not in the completer sample (Q = 7.55, p = 0.11) according to the remission criterion used. As shown in Table 3, reported remission rates were highest when remission was defined as good end state functioning or as no longer meeting criteria for the primary diagnosis, and lowest when remission was defined as a score below a cutoff. Remission rates were also lower for those PD/A studies defining remission as panic-free status. At follow-up, betweengroup heterogeneity was not significant for the ITT sample (Q = 1.25, p = 0.742) or for the completer sample (Q = 2.94, p = 0.401). Meta-regression was used to examine the relationship between pretreatment patient variables and remission rates, examining each variable independently. It should be noted that pre-treatment characteristics differed in terms of the number of available samples. This is because certain articles only provided demographic characteristics for the full sample (e.g., CBT and control group combined), not the CBT group specifically, and had to be excluded from the analyses. As shown in Table 4, in the ITT sample, mean sample age did not significantly predict remission rates at post-treatment; however, at follow-up the effect was significant, indicating lower remission rates in older samples. The percentage of patients taking psychiatric medications was a significant predictor at both post-treatment and follow-up, with lower remission rates in samples with higher medication use. At post-treatment, this effect appears largely attributable to the percentage of patients taking benzodiazepines (antidepressant medications, conversely, were associated with significantly higher remission rates). Samples with higher rates of comorbid depressive disorders were associated with lower remission rates at post-treatment and follow-up. There was a nonsignificant trend toward a lower remission rate for samples with higher rates of comorbid substance use disorders at post-treatment; at followup this effect was significant. Neither the percentage of women nor the percentage of minority participants was significantly associated with remission rates, though a non-significant trend is noted for lower posttreatment remission rates among samples with greater minority representation. In the completer sample, at both post-treatment and at follow-up, higher remission rates were associated with younger age, lower rates of minority participation, and lower rates of both depressive and substance use disorders. Greater medication use was associated with lower rates of remission at post-treatment, though curiously, for those samples reporting either antidepressant or benzodiazepine use, both of these medications were associated with higher remission rates. Table 5 shows remission rates according to the presence or absence of methodological quality variables. In the ITT sample, remission rates did not differ according to randomized controlled trials vs. open trials (Q = 1.21, p = 0.271), use of reliable and valid measures (Q = 3.21, p = 0.200), use of a treatment manual (Q = 0.01, p = 0.920), or acceptable treatment fidelity (Q = 1.13, p = 0.287). Measures administered by a blind evaluator had significantly (Q = 7.20, p = 0.007) lower remission rates than did those not administered by a blind evaluator (either patient self-report or a non-blinded evaluator). In the completer sample, remission rates did not differ significantly according to randomized vs. open trials (Q = 2.45, p = 0.117), use of reliable and valid measures (Q = 3.07, p = 0.215), blind evaluators (Q = 1.30, p = 0.255), or

Funnel Plot of Precision by Logit event rate 10

8

Precision (1/Std Err)

6

4

2

0

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

Logit event rate

Funnel Plot of Precision by Logit event rate 8

7

6

Precision (1/Std Err)

5

4

3

2

1

0

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

Logit event rate

Fig. 2. Funnel plots for remission rates in intent-to-treat samples at post-treatment (top) and follow-up (bottom).

end state functioning, and lowest in studies defining remission as scoring below a clinical cutoff, with or without other requirements such as longer meeting criteria for the primary diagnosis. In the completer sample, reported remission rates were highest in studies defining 4

Clinical Psychology Review 61 (2018) 1–8

K.S. Springer et al.

Table 2 Mean remission rates for intent-to-treat samples (top) and completer samples (bottom) according to primary diagnosis. Diagnosis

Post-treatment

Follow-up

k

Mean

95% CI

Q

I2

k

Mean

95% CI

Q

I2

OCD SAD PD/A GAD PTSD

6 10 13 3 20

37.9% 40.1% 48.0% 51.4% 53.3%

31.2%–45.1% 30.4%–50.7% 39.8%–56.4% 35.5%–66.9% 45.3%–61.1%

2.71 40.99⁎ 53.26⁎ 7.43⁎ 154.46⁎

0.00 78.04 77.47 73.08 87.70

4 4 8 2 13

52.3% 43.2% 55.6% 65.0% 54.8%

42.0%–62.4% 37.8%–48.7% 44.4%–66.2% 43.6%–81.7% 44.7%–64.4%

2.23 6.34 38.27⁎ 3.97⁎ 97.96⁎

0.00 52.66 81.71 74.79 87.75

OCD SAD PD/A GAD PTSD

27 11 18 11 23

45.2% 40.4% 55.5% 56.3% 62.8%

38.2%–52.3% 30.2%–51.5% 45.2%–38.2% 50.2%–62.2% 52.1%–72.3%

130.38⁎ 52.23⁎ 87.33⁎ 13.21 249.67⁎

80.06 80.85 80.53 24.30 91.19

18 3 20 8 17

43.9% 45.4% 57.5% 65.2% 63.5%

35.4%–52.7% 22.3%–70.8% 50.4%–64.3% 53.1%–75.6% 48.7%–76.1%

57.89⁎ 12.14⁎ 46.12⁎ 24.18⁎ 219.22⁎

70.63 83.53 58.81 71.05 92.70

Note. OCD = obsessive-compulsive disorder. SAD = social anxiety disorder. PD/A = panic disorder and/or agoraphobia. GAD = generalized anxiety disorder. PTSD = posttraumatic stress disorder. k = number of samples. CI = confidence interval. ⁎ p < .05.

acceptable treatment fidelity (Q = 3.13, p = 0.077). Significantly lower remission rates were found in studies that used a treatment manual, compared to those that did not (Q = 5.63, p = 0.018).

2015). We argue that it is important to consider each of these variables separately with response being change over the course of therapy, and remission being a healthy end-state. In our study, there were no significant differences in remission rates in anxiety disorders between the ITT and completer samples, with an overall remission rate of 51.0%. The overall remission rate across anxiety disorders for ITT samples was 47.9% at post-treatment and 53.0% at follow up, while the remission rate for completer samples was 53.6% at post-treatment and 56.1% at

4. Discussion One complication with interpretations of previous research is that the constructs of response and remission were combined (Loerinc et al.,

Table 3 Mean remission rates for intent-to-treat samples (top) and completer samples (bottom) according to criterion used. Expanded criteriaa

Post-treatment

Follow-up k

Mean

95% CI

Q

I2

77.61 75.26 86.14 70.88 86.85 71.90 0.00 0.00

18 6 10 4 1 3 1 0

56.2% 54.3% 51.4% 58.4% 75.0% 38.0% 46.0% –

48.3%–63.7% 43.9%–64.3% 37.3%–65.3% 48.3%–67.8% 65.6%–82.5% 24.6%–53.4% 31.7%–60.9% –

79.68⁎ 14.85⁎ 101.28⁎ 2.98 – 3.90 – –

78.66 66.33 91.11 0.00 – 48.69 – –

86.85 82.18 77.61 75.26

1 17 18 6

75.0% 52.2% 56.2% 54.3%

65.6%–82.5% 42.1%–62.1% 48.3%–63.7% 43.9%–64.3%

– 112.15⁎ 79.68⁎ 14.85⁎

– 85.73 78.66 66.33

k

Mean

95% CI

Q

I

Not meeting diagnostic criteria Good end state functioning Below cutoff score Clinically significant change Panic free Decrease + below cutoff No diagnosis + below cutoff Panic free + below cutoff

22 9 17 7 5 5 2 2

55.8% 53.9% 42.5% 45.3% 49.6% 47.9% 34.1% 31.7%

48.7%–62.6% 41.7%–65.7% 32.9%–52.6% 31.0%–60.3% 32.6%–66.7% 37.4%–58.6% 22.4%–48.2% 22.0%–43.2%

93.80⁎ 32.33⁎ 115.419⁎ 20.60⁎ 30.42⁎ 14.24⁎ 0.34 0.66

Simplified criteriaa Panic free Below cutoff score No diagnosis Good end state functioning

5 32 22 9

49.6% 42.9% 55.8% 53.9%

32.6%–66.7% 36.5%–49.5% 48.7%–62.6% 41.7%–65.7%

30.42⁎ 173.98⁎ 93.80⁎ 32.33⁎

Expanded Criteriaa

Post-treatment

2

Follow-up 2

k

Mean

95% CI

Q

I

k

Mean

95% CI

Q

I2

Not meeting diagnostic criteria Good end state functioning Below cutoff score Clinically significant change Panic free Decrease + below cutoff No diagnosis + below cutoff Panic free + below cutoff

32 8 15 36 4 9 3 5

61.5% 62.3% 37.1% 51.0% 70.8% 54.2% 65.8% 41.4%

53.0%–69.3% 50.1%–73.0% 27.2%–48.1% 44.1%–57.9% 46.3%–87.1% 41.8%–66.0% 30.1%–89.6% 28.8%–55.3%

204.19⁎ 16.42⁎ 76.75⁎ 196.33⁎ 10.37⁎ 44.92⁎ 20.68⁎ 7.98

84.82 57.38 81.76 82.17 71.07 82.19 90.33 49.87

27 4 6 22 4 7 2 6

62.9% 60.3% 30.0% 54.9% 67.6% 45.4% 85.4% 56.0%

52.5%–72.3% 47.4%–71.9% 17.9%–45.8% 46.7%–62.8% 35.6%–88.8% 36.7%–54.3% 30.1%–98.8% 46.0%–65.4%

202.74⁎ 5.91 33.23⁎ 61.81⁎ 9.55⁎ 9.49 11.43⁎ 6.50

87.18 49.23 84.96 66.02 68.59 36.78 91.25 23.07

Simplified criteriaa Panic free Below cutoff score No diagnosis Good end state functioning

4 68 32 8

70.8% 48.2% 61.5% 62.3%

46.3%–87.1% 43.1%–53.4% 53.0%–69.3% 50.1%–73.0%

10.37⁎ 407.32⁎ 204.19⁎ 16.42⁎

71.07 83.55 84.82 57.38

4 43 27 4

67.6% 51.1% 62.9% 60.3%

35.6%–88.8% 44.5%–57.6% 52.5%–72.3% 47.4%–71.9%

9.55⁎ 191.75⁎ 202.74⁎ 5.91

68.59 78.10 87.18 49.23

Note. Expanded criteria refers to the original categories used, whereas simplified criteria refers to new categories used in order to reduce the number of categories. k = number of samples. CI = confidence interval. ⁎ p < .05. a Measure was the unit of analysis.

5

Clinical Psychology Review 61 (2018) 1–8

K.S. Springer et al.

researchers and clinicians do not rely solely on diagnostic criteria to determine remission status. Consistent with the response rates identified by Loerinc et al. (2015), patients with GAD and PTSD achieved the highest remission rates, while those with OCD and SAD had the lowest remission rates. Although GAD follows a chronic course (Wittchen, 2002) with lower rates of spontaneous remission (Bruce et al., 2005), the present analysis suggests a relatively favorable prognosis with CBT, though the majority of GAD patients did not achieve remission and the findings for GAD are based on only three ITT studies, rendering any conclusions tentative. Unlike GAD, even in the absence of treatment, PTSD prevalence follows a declining course, with the steepest declines seen over the first 12 months after trauma exposure (Kessler, Sonnega, Bromet, Hughes, & Nelson, 1995). Among the PTSD studies included in the present metaanalysis, several required PTSD of as little as three to four months' duration (Bryant, Moulds, Guthrie, Dang, & Nixon, 2003; Cottraux et al., 2008; Ehlers et al., 2013; Ehlers et al., 2014; Ehlers, Clark, Hackmann, McManus, & Fennell, 2005; Resick et al., 2008; Resick, Nishith, Weaver, Astin, & Feuer, 2002), and it is possible that the higher remission rates partially reflect the course of the disorder. In a recent meta-analysis of CBT for OCD, Öst, Havnen, Hansen, and Kvale (2015) reported large effect sizes of 1.31 and 1.33 for pre- to post-treatment symptom change when CBT was compared to waitlist and placebo conditions, respectively. Given that both SAD and OCD had the lowest remission rates in the present study, it could be that although symptoms reduce substantially with CBT, this symptom reduction frequently does not translate to remission. Remission rates differed by several patient variables (e.g., medication use, age, comorbid disorders). Samples with high rates of antidepressant use had better outcomes with CBT, whereas samples with high rates of benzodiazepine use had significantly worse outcomes. This is consistent with research showing that benzodiazepines, at least when used on an as-needed basis (PRN), are associated with poorer outcomes in CBT (Westra, Stewart, & Conrad, 2002). However, the current metaanalysis lacks the details of the benzodiazepine usage to determine whether the medications were used PRN or on a regular basis. More research is needed to clarify the impact of medication usage on CBT outcomes, both cross-sectionally and longitudinally. Although older samples were as likely to achieve remission status at post-treatment, they were less likely to do so at follow-up, suggesting a possible need for age-related treatment modifications (e.g., Mohlman et al., 2003). Finally, samples with high rates of comorbid depression and/or substance use disorder had lower rates of remission. Whereas Olatunji, Cisler, and Tolin (2010) did not find a significant relationship across studies between depressive and substance use disorders and anxiety symptom reduction, the present findings suggest that the impact of these disorders is more evident in rates of remission, perhaps because

Table 4 Meta-regression of pre-treatment demographic variables and remission rates for intent-totreat samples (Top) and completer samples (Bottom). Demographic variable

Post-treatment

Follow-up

k

Z

p

k

Z

p

Age % Female % Minority % on Medications % on Antidepressants % on Benzodiazepines % with Comorbid Depressive Disorder % with Comorbid Substance Use Disorder

37 41 17 22 13 15 23

−1.11 0.05 −1.79 −3.31 2.18 −2.25 −4.01

0.27 0.95 0.07 < 0.001 0.03 0.02 < 0.001

20 24 8 10 7 8 12

−5.10 0.87 −0.93 −4.13 −0.22 0.17 −6.22

< 0.001 0.38 0.35 < 0.001 0.82 0.87 < 0.001

30

−1.93

0.05

11

−5.41

< 0.001

Age % Female % Minority % on Medications % on Antidepressants % on Benzodiazepines % with Comorbid Depressive Disorder % with Comorbid Substance Use Disorder

61 57 23 40 18 15 29

−2.47 0.50 −3.22 −4.45 4.39 2.17 −5.03

0.013 0.62 0.001 < 0.001 < 0.001 0.03 < 0.001

44 41 13 26 7 3 18

−2.94 0.00 −6.09 −5.66 0.93 1.18 −6.12

0.003 0.69 < 0.001 < 0.001 0.35 0.24 < 0.001

43

−6.00

< 0.001

25

−10.34

< 0.001

Note. k = number of samples.

follow up. Although CBT is an empirically supported treatment for anxiety disorders and often outperforms other treatments (Butler et al., 2006; Tolin, 2010), the present results suggest that there is significant room for improvement as only half of patients achieve remission status. Remission rates at post-treatment varied according to the way in which remission was defined. The lack of one universally accepted definition of remission is problematic in that it is difficult to obtain a clear understanding of true remission rates. In the present meta-analysis, the criteria of no diagnosis and good end state functioning yielded the highest rates of remission, whereas scoring below a clinical cutoff yielded lower remission rates. No longer meeting diagnostic criteria for an anxiety disorder is likely too weak of a definition since a patient can lose one DSM criterion for a particular disorder but still be significantly impaired. It is promising that good end state functioning, which examines how well the patient is doing across multiple domains, yielded relatively high remission rates, suggesting that CBT is able to treat problems across broad domains beyond the primary diagnosis. However, of the eight samples using this remission criterion, it is noted that six of them were studies of PTSD, potentially inflating results as discussed below. Therefore, to be conservative, we suggest that remission be defined as scoring below a clinical cutoff. We recommend that

Table 5 Mean post-treatment remission rates for ITT samples (Top) and completer samples (Bottom) according to the presence or absence of methodological quality criteria. Quality criterion

Present

Absent

k

Mean

95% CI

Q

I2

k

Mean

95% CI

Q

I2

Blind evaluatora Randomized trial Reliable and valid measuresa Treatment fidelity Treatment manual

49 47 72 40 46

47.1% 47.2% 52.4% 49.3% 48.0%

41.8%–52.4% 42.3%–52.0% 47.7%–57.0% 44.2%–54.4% 43.3%–52.8%

237.94⁎ 235.28⁎ 436.77⁎ 221.53⁎ 257.40⁎

79.83 80.45 83.74 82.40 82.52

35 5 11 12 6

58.2% 54.3% 47.5% 42.9% 47.2%

52.0%–64.2% 42.6%–65.5% 36.8%–58.5% 32.9%–53.5% 31.9%–63.0%

205.89⁎ 25.661⁎ 59.61⁎ 63.94⁎ 257.40⁎

83.49 84.41 83.22 82.80 82.52

Blind evaluatora Randomized trial Reliable and valid measuresa Treatment fidelity Treatment manual

78 70 133 54 84

56.0% 55.0% 53.2% 56.1% 51.3%

50.5%–61.3% 49.5%–60.3% 49.3%–57.1% 49.9%–62.0% 46.5%–56.0%

463.76⁎ 410.48⁎ 807.69⁎ 369.24⁎ 535.34⁎

83.40 83.19 83.66 85.65 84.50

69 25 13 41 11

51.7% 47.5% 62.1% 48.3% 65.5%

46.8%–56.6% 39.9%–55.1% 53.0%–70.4% 42.4%–54.3% 54.8%–74.9%

369.12⁎ 189.76⁎ 24.66⁎ 206.65⁎ 50.34⁎

81.58 87.35 51.35 80.64 80.13

Note. CI = confidence interval. k = number of samples. ⁎ p < .05. a Measure was the unit of analysis.

6

Clinical Psychology Review 61 (2018) 1–8

K.S. Springer et al.

Acknowledgements

comorbid symptoms interfere with the ability to learn therapy skills and/or to adequately engage in therapy. Surprisingly, there were no differences when reliable and valid assessment measures were used in the studies. Those studies that used a more methodologically rigorous study design (e.g., RCT vs. open trial) did not have difference in reported remission rates. As expected, lower remission rates were reported in the completer samples when a treatment manual was used. In the ITT sample, there were lower remission rates when blind evaluators were performing the assessments. Therefore, it is recommended that future studies continue to use independent evaluators to avoid inflated remission rates. Several limitations of the present analyses are worth noting. First, given that there is no universally accepted definition of remission, we used only the most common definitions that we encountered, and this was not an exhaustive list. Second, certain anxiety disorders (e.g., ITT GAD sample, k = 3) and certain definitions of remission (e.g., panic free status, k = 5) had small sample sizes, rendering those findings very tentative. Due to these small sample sizes we were also unable to address more specific and nuanced questions about remission, such as whether or not certain definitions yield more accurate remission rates for certain anxiety diagnoses compared to others. Finally, we did not include control conditions in our meta-analysis; it is therefore unclear how remission rates in CBT compare to spontaneous remission or remission in other treatments. With these limitations in mind, the present results suggest a number of directions for future research. First, the field would benefit from a universally accepted and applied definition for remission, which would yield more accurate estimates of the efficacy of CBT on its own and in comparison to control treatments. Such consensus would facilitate modifications and improvements in CBT. Second, it will be critical to address more specific questions about remission for specific patients and disorders, such as whether certain definitions, measures, or evaluation criteria yield more accurate estimates of remission rates for certain individuals and problems. Ultimately, a more nuanced understanding of remission and how it may differ between individuals and presenting problems may lead to more effective and tailored treatments. Future research should also address whether variants of CBT (e.g., concurrent treatment for anxiety and substance use) and varying methods of service delivery (e.g., telehealth) result in different reported remission rates. To conclude, we encourage treatment researchers to place more of an emphasis on remission when designing and evaluating their treatments. Now that CBT is an established, evidence-based treatment for anxiety disorders, the field should shift its focus from efficacy to remission as the ultimate goal of treatment.

The authors wish to thank Bailey D'Antonio and Pamela Scalise for their assistance with collecting and organizing articles. Appendix A. Supplementary data Supplementary data to this article can be found online at https:// doi.org/10.1016/j.cpr.2018.03.002. References American Psychiatric Association (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author. Borenstein, M., Hedges, L. V., Higgins, J. P., & Rothstein, H. R. (2009). Introduction to meta-analysis. Chichester, England: Wiley. Bruce, S. E., Yonkers, K. A., Otto, M. W., Eisen, J. L., Weisberg, R. B., Pagano, M., & Keller, M. B. (2005). Influence of psychiatric comorbidity on recovery and recurrence in generalized anxiety disorder, social phobia, and panic disorder: A 12-year prospective study. The American Journal of Psychiatry, 162(6), 1179–1187. http://dx.doi. org/10.1176/appi.ajp.162.6.1179. Bryant, R. A., Moulds, M. L., Guthrie, R. M., Dang, S. T., & Nixon, R. D. (2003). Imaginal exposure alone and imaginal exposure with cognitive restructuring in treatment of posttraumatic stress disorder. Journal of Consulting and Clinical Psychology, 71(4), 706–712. http://dx.doi.org/10.1037/0022-006X.71.4.706. Butler, A. C., Chapman, J. E., Forman, E. M., & Beck, A. T. (2006). The empirical status of cognitive-behavioral therapy: A review of meta-analyses. Clinical Psychology Review, 26(1), 17–31. http://dx.doi.org/10.1016/j.cpr.2005.07.003. Cottraux, J., Note, I., Yao, S. N., de Mey-Guillard, C., Bonasse, F., Djamoussian, D., ... Chen, Y. (2008). Randomized controlled comparison of cognitive behavior therapy with Rogerian supportive therapy in chronic post-traumatic stress disorder: A 2-year follow-up. Psychotherapy and Psychosomatics, 77(2), 101–110. http://dx.doi.org/10. 1159/000112887. Duval, S., & Tweedie, R. (2000). Trim and fill: A simple funnel-plot–based method of testing and adjusting for publication bias in meta-analysis. Biometrics, 56(2), 455–463. http://dx.doi.org/10.1111/j.0006-341X.2000.00455.x. Ehlers, A., Clark, D. M., Hackmann, A., McManus, F., & Fennell, M. (2005). Cognitive therapy for post-traumatic stress disorder: Development and evaluation. Behaviour Research and Therapy, 43(4), 413–431. http://dx.doi.org/10.1016/j.brat.2004.03. 006. Ehlers, A., Grey, N., Wild, J., Stott, R., Liness, S., Deale, A., & Clark, D. M. (2013). Implementation of cognitive therapy for PTSD in routine clinical care: Effectiveness and moderators of outcome in a consecutive sample. Behaviour Research and Therapy, 51(11), 742–752. http://dx.doi.org/10.1016/j.brat.2013.08.006. Ehlers, A., Hackmann, A., Grey, N., Wild, J., Liness, S., Albert, I., & Clark, D. M. (2014). A randomized controlled trial of 7-day intensive and standard weekly cognitive therapy for PTSD and emotion-focused supportive therapy. The American Journal of Psychiatry, 171(3), 294–304. http://dx.doi.org/10.1176/appi.ajp.2013.13040552. Foa, E. B., & Meadows, E. A. (1997). Psychosocial treatments for posttraumatic stress disorder: A critical review. Annual Review of Psychology, 48, 449–480. http://dx.doi. org/10.1146/annurev.psych.48.1.449. Goodman, W. K., Price, L. H., Rasmussen, S. A., Mazure, C., Fleischmann, R. L., Hill, C. L., & Charney, D. S. (1989). The Yale-Brown obsessive compulsive scale. I. Development, use, and reliability. Archives of General Psychiatry, 46(11), 1006–1011. http://dx.doi. org/10.1001/archpsyc.1989.01810110048007. Higgins, J. P., Thompson, S. G., Deeks, J. J., & Altman, D. G. (2003). Measuring inconsistency in meta-analyses. BMJ, 327(7414), 557–560. http://dx.doi.org/10.1136/ bmj.327.7414.557. Jacobson, N. S., Roberts, L. J., Berns, S. B., & McGlinchey, J. B. (1999). Methods for defining and determining the clinical significance of treatment effects: Description, application, and alternatives. Journal of Consulting and Clinical Psychology, 67(3), 300–307. http://dx.doi.org/10.1037/0022-006X.67.3.300. Jacobson, N. S., & Truax, P. (1991). Clinical significance: A statistical approach to defining meaningful change in psychotherapy research. Journal of Consulting and Clinical Psychology, 59(1), 12–19. http://dx.doi.org/10.1037/0022-006X.59.1.12. Jadad, A. R., Moore, R. A., Carroll, D., Jenkinson, C., Reynolds, D. J., Gavaghan, D. J., & McQuay, H. J. (1996). Assessing the quality of reports of randomized clinical trials: Is blinding necessary? Controlled Clinical Trials, 17(1), 1–12. Kessler, R. C., Alonso, J., Chatterji, S., & He, Y. (2014). Disability and costs. In P. E. T. Ehring (Vol. Ed.), The Wiley handbook of anxiety disorders. Vol. I. The Wiley handbook of anxiety disorders (pp. 47–57). Chichester, UK: John Wiley & Sons. Kessler, R. C., Sonnega, A., Bromet, E., Hughes, M., & Nelson, C. B. (1995). Posttraumatic stress disorder in the National Comorbidity Survey. Archives of General Psychiatry, 52(12), 1048–1060. http://dx.doi.org/10.1001/archpsyc.1995.03950240066012. Loerinc, A. G., Meuret, A. E., Twohig, M. P., Rosenfield, D., Bluett, E. J., & Craske, M. G. (2015). Response rates for CBT for anxiety disorders: Need for standardized criteria. Clinical Psychology Review, 42, 72–82. http://dx.doi.org/10.1016/j.cpr.2015.08.004. McGlinchey, J. B., Atkins, D. C., & Jacobson, N. S. (2002). Clinical significance methods: Which one to use and how useful are they? Behavior Therapy, 33, 529–550. http://dx. doi.org/10.1016/S0005-7894(02)80015-6. Mohlman, J., Gorenstein, E. E., Kleber, M. S., de Jesus, M., Gorman, J. M., & Papp, L. A. (2003). Standard and enhanced cognitive-behavior therapy for late-life generalized

Role of funding sources There has been no significant financial support for this work that could have influenced its outcome. Contributors We confirm that the manuscript has been read and approved by all named authors and that there are no other persons who satisfied the criteria for authorship but are not listed. We further confirm that the order of authors listed in the manuscript has been approved by all of us. Authors Kristen Springer and Hannah Levy coded all abstracts and articles, generated the data set, and collaboratively wrote the first draft of the manuscript. Author David Tolin ran all of the analyses and provided feedback on all drafts of the manuscript. All authors contributed to and have approved the final manuscript. Conflict of interest We wish to confirm that there are no known conflicts of interest associated with this publication. 7

Clinical Psychology Review 61 (2018) 1–8

K.S. Springer et al.

Westra, H. A., Stewart, S. H., & Conrad, B. E. (2002). Naturalistic manner of benzodiazepine use and cognitive behavioral therapy outcome in panic disorder with agoraphobia. Journal of Anxiety Disorders, 16(3), 233–246. http://dx.doi.org/10.1023/ A:1005403025084. Wittchen, H. U. (2002). Generalized anxiety disorder: Prevalence, burden, and cost to society. Depression and Anxiety, 16(4), 162–171. http://dx.doi.org/10.1002/da. 10065.

anxiety disorder: Two pilot investigations. American Journal of Geriatric Psychiatry, 11, 24–32. http://dx.doi.org/10.1176/appi.ajgp.11.1.24. Olatunji, B. O., Cisler, J. M., & Deacon, B. J. (2010). Efficacy of cognitive behavioral therapy for anxiety disorders: A review of meta-analytic findings. The Psychiatric Clinics of North America, 33(3), 557–577. http://dx.doi.org/10.1016/j.psc.2010.04. 002. Olatunji, B. O., Cisler, J. M., & Tolin, D. F. (2010). A meta-analysis of the influence of comorbidity on treatment outcome in the anxiety disorders. Clinical Psychology Review, 30(6), 642–654. http://dx.doi.org/10.1016/j.cpr.2010.04.008. Öst, L.-G., Havnen, A., Hansen, B., & Kvale, G. (2015). Cognitive behavioral treatments of obsessive–compulsive disorder. A systematic review and meta-analysis of studies published 1993–2014. Clinical Psychology Review, 40, 156–169. http://dx.doi.org/10. 1016/j.cpr.2015.06.003. Resick, P. A., Galovski, T. E., O'Brien Uhlmansiek, M., Scher, C. D., Clum, G. A., & YoungXu, Y. (2008). A randomized clinical trial to dismantle components of cognitive processing therapy for posttraumatic stress disorder in female victims of interpersonal violence. Journal of Consulting and Clinical Psychology, 76(2), 243–258. http://dx.doi.org/10.1037/0022-006X.76.2.243. Resick, P. A., Nishith, P., Weaver, T. L., Astin, M. C., & Feuer, C. A. (2002). A comparison of cognitive-processing therapy with prolonged exposure and a waiting condition for the treatment of chronic posttraumatic stress disorder in female rape victims. Journal of Consulting and Clinical Psychology, 70(4), 867–879. http://dx.doi.org/10.1037/ 0022-006X.70.4.867. Tolin, D. F. (2010). Is cognitive-behavioral therapy more effective than other therapies? A meta-analytic review. Clinical Psychology Review, 30(6), 710–720. http://dx.doi.org/ 10.1016/j.cpr.2010.05.003.

Kristen S. Springer received her Ph.D. in Clinical and Health Psychology from the University of Florida in Gainesville, FL in 2015 with a specialty in cognitive behavioral treatments for chronic pain and anxiety disorders. She then completed her pre-doctoral internship at Eastern Virginia Medical School in Norfolk, VA. Dr. Springer subsequently completed a clinical post-doctoral fellowship at the Anxiety Disorders Center/Center for Cognitive Behavioral Therapy at the Institute of Living in Hartford, CT and remains on staff as a licensed psychologist. Dr. Springer has authored several scientific journal articles in the fields of both anxiety and chronic pain.

8