Social Support and Symptom Severity Among Patients With Obsessive-Compulsive Disorder or Panic Disorder With Agoraphobia: A Systematic Review

Panic disorder with or without agoraphobia (PD/A) and obsessive-compulsive disorder (OCD) are characterized by major behavioral dysruptions that may affect patients’ social and marital functioning. The disorders’ impact on interpersonal relationships may also affect the quality of support patients receive from their social network. The main goal of this systematic review is to determine the association between social or marital support and symptom severity among adults with PD/A or OCD. A systematic search of databases was executed and provided 35 eligible articles. Results from OCD studies indicated a negative association between marital adjustment and symptom severity, and a positive association between accommodation from relatives and symptom severity. However, results were inconclusive for negative forms of social support (e.g. criticism, hostility). Results from PD/A studies indicated a negative association between perceived social support and symptom severity. Also, results from studies using an observational measure of marital adjustment indicated a negative association between quality of support from the spouse and PD/A severity. However, results were inconclusive for perceived marital adjustment and symptom severity. In conclusion, this systematic review generally suggests a major role of social and marital support in PD/A and OCD symptomatology. However, given diversity of results and methods used in studies, more are needed to clarify the links between support and symptom severity among patients with PD/A and OCD.

independent correctors after full text review was 97% (171/177). For the remaining six articles, a consensus was reached between the reviewers (three were included and three were excluded).

Results
The electronic database search provided 4011 articles, from which 2010 duplicates were removed. The other 2001 articles were screened by title, abstract and keywords, and this first selection led to the removal of 1826 articles. Two articles were impossible to retrieve (authors' contacts could not be found; Cohen, 1986;Kitch, 1983), which led to 173 articles that went through a full text review. Among these articles, 148 were written in English, six in Chinese, six in German, four in French, three in Italian, two in Portuguese, one in Japanese, one in Korean, one in Turkish, and one in Dutch. Finally, 30 of these articles were included in the present review.
Results from the independent search led to the inclusion of three other articles, for a total sample of 33 articles.
Reasons for exclusion of the remaining 143 articles are described in Table 1. The article selection is described in detail in Figure 1.

Support and OCD Severity The Association Between OCD Severity and Marital Adjustment
Two studies looked at the association between marital adjustment and OCD. In 2006, Abbey conducted a study in order to examine romantic relationship functioning in individuals with OCD. The results indicated that relationship satisfaction negatively correlated with the obsessing (r = -.26, p < .05) and positively correlated with the neutralizing (r = .26, p < .05) subscales of the Obsessive-Compulsive Inventory, Revised (OCI-R). Selfdisclosure, which is an index of positive social support, negatively correlated with the obsessive subscale of the OCI-R (r = -.30, p < .05). Moreover, the obsessive subscale also negatively correlated with the emotional subscale ("my partner listens to me when I need someone to talk to") of the Personal Assessment of Intimacy in Relationship (PAIR; Schaefer & Olson, 1981;r = -.27, p < .05). No other significant associations between measures of marital adjustment and OCD severity were found. Riggs, Hiss, and Foa (1992) also looked at the link between marital distress and OCD symptom severity. The results indicated a significant negative correlation between marital adjustment and avoidance of the situation related to the main obsession (r (50) = -.28, p < .05), as rated by an independent assessor, but not between marital adjustment scores and ratings of main obsessions (r (52) = -.04, p > .70) or rituals (r (52) = .06, p > .65). For a summary of these results, see Table 2. Many studies found a significant positive correlation between accommodation and total scores on the measure of obsessive-compulsive disorder severity (Boeding et al., 2013;Calvocoressi et al., 1999;Cherian, Pandian, Bada Math, Kandavel, & Reddy, 2013;Cherian, Pandian, Bada Math, Kandavel, & Janardhan Reddy, 2014;Ferrão & Florão, 2010;Gomes et al., 2014;Ramos-Cerqueira, Torres, Torresan, Negreiros, & Vitorino, 2008;Stewart et al., 2008;Vikas et al., 2011;Wu, Pinto, et al., 2016), meaning that the more accommodation provided by the relatives, the more severe the OCD symptoms.
Some authors also found significant association between specific areas of accommodation and symptoms of OCD. Indeed, Vikas and colleagues (2011) found that participation in rituals was correlated with the level of obsessions (r = .52, p < .01) and compulsions (r = .54, p < .01). Some results demonstrated that the level of accommodation was associated with contamination/cleaning compulsions (r = .18, p = .03, Albert et al., 2010; r = .26, p = .007, Stewart et al., 2008). Also, Albert and colleagues (Albert, Brunatto, Maina, & Bogetto, (2009;Albert et al., 2010) found that accommodation total scores were significantly correlated with obsessions (r = .21, p = .013) but not compulsions (r and p are not available in the original article). However, results from Beoding's study (2013) indicated a significant positive correlation between accommodation and severity of compulsions (r = .39, p < .05) but not with severity of obsessions (r = .26, p > .05), which contradicts results obtained by Albert and colleagues (2010).
Other authors also found that accommodation by the family could predict the severity of OCD symptomatology.
Indeed, Van Noppen and Steketee ( , 2009 reported that the level of accommodation, as rated by patients, predicted OCD severity in regression analyses (b = 0.47, p < .01, partial correlation = .42, R 2 change = .16, p > .01). Accommodation was also predictive of OCD severity when rated by relatives (b = 0.50, p < .01, partial correlation = .46, R 2 change = .20, p > .01). Accommodation alone as rated by patients and relatives explained 16 and 20%, respectively, of the symptom severity (F (1,44) = 9.19, p < .01; F (1.44) = 11.7, p = .001). They also found that among many factors (e.g. relatives' attributions, emotional over-involvement, and criticism), accommodation was the strongest predictor of OCD severity, explaining 42% of the variance (direct causal effects .42, p < .05). On the other hand, other authors found that OCD symptomatology could predict accommodation behaviors among the relatives. For example, Stewart and colleagues (2008) reported significant correlations between OCD severity and cleaning/contamination symptoms (r = .26, p = .007). When entered in a stepwise regression analysis, these factors remained significant. Similarly, Albert and colleagues (2010) also entered the significant factors in a regression analysis and reported that a higher FAS total score was predicted by the contamination/cleaning symptom dimension score (β = 0.22, t = 2.87, p = .005).
The only study that found negative results is the one by Drury and colleagues (2014). They conducted a study on hoarders and their relatives in order to assess the impact of hoarding on functioning as well as burden for the relatives. They used a different measure, the Family Impact Scale for Hoarding disorder (FISH), in order to assess both the level of family accommodation displayed by the relatives and the associated burden. The results indicate that hoarding severity did not predict FISH scores (b = 0.20, t = 1.33, p = .190).
Finally, two recent meta-analyses investigated the association between family accommodation and OCD severity. In Strauss, Hale, and Stobie (2015), results from 14 studies (seven with adults and seven on pediatric OCD) showed a statistically significant medium effect size (r = .35, 95% CI [.23, .47]), so that family accommodation accounts for approximately 12% of the variance in OCD symptom severity. In , 41 studies on accommodation and OCD severity were included. Among those studies, 15 were on Palardy, El-Baalbaki, Fredette et al. 261 adults with OCD. Results showed a medium positive effect (r = .42, 95% CI [.36, .47], z = 13.00, p < .001), which indicates that higher OCD severity is associated with increased family accommodation. Also, there was no significant difference by categorical age groups (Q(1) btwn = 1.36, p = .24) and no significant effect when examining participant mean age (β = -0.002, SE = .003, z = -0.82, p = .41). This suggests that the association between family accommodation and OCD severity is similar whether participants are adults or children. For a summary of these results, see Table 3.

The Association Between OCD Severity and Expressed Emotion
Three studies were interested in the association between OCD severity and Expressed Emotion, with one of them demonstrating a significant association. Indeed, Cherian and colleagues' (2014) results indicate that perceived criticism was associated with OCD severity as measured by the Y-BOCS (r = .24, p < .01) and the Clinical Global Impression (CGI; r = .27, p < .01). Finally, Van Noppen and Steketee ( , 2009 found no significant association between Expressed Emotions variables (criticism, hostility, emotional over-involvement) and OCD severity, when rated by patients or relatives. When accommodation of the family was entered in the model, criticism lost its significance (Van Noppen & Steketee, 2003, 2009). See Table 4 for a summary of the results.

Support and PD/A Severity The Association Between PD/A Severity and Social Support
Three studies looked at the link between PD/A and social support, with two of them demonstrating that the level of support is associated with the severity of PD/A symptoms. Huang and colleagues (2010)  appraisal did not predict agoraphobia in a regression analysis, the authors did find a significant negative correlation between the two variables (r = -.27, p < .05). However, they did not find a link between social support appraisal and severity of panic symptoms. Renneberg, Chambless, Fydrich, and Goldstein (2002) conducted a study in order to investigate affect balance in dyads of patients and their relatives and its association with outcome following cognitive behavioral therapy.
Social Support and OCD/PDA Severity: A Systematic Review 264 Given that the relatives were nine parents and 26 spouses, this study was included under both social and marital support (see paragraph below). In order to assess the level of affectivity in their sample, the authors used an observational measure, which assesses both verbal and non-verbal behaviors in an interaction between partners. Based on these observations, they separated the group between affect-balanced and affectunbalanced dyads. The authors found that the two groups did not differ on pre-treatment scores of measures of agoraphobia and panic (t-tests p > .17 for all measures), which means that the quality of the interaction during a problem-solving task between partners was not associated with symptoms severity. Refer to Table 5 for a summary of these results.

The Association Between PD/A Severity and Marital Adjustment
Eleven studies evaluated the link between marital adjustment and the severity of either panic disorder and\or agoraphobia. Authors reported mixed results. Since authors did not all use the same measure of marital adjustment, results will be presented by the type of measure utilized in order to compare similar articles. It is thus possible that one study gets described in several places due to its use of multiple questionnaires.
the MMMQ, and the severity of agoraphobia. However, no statistics were presented for this result. Monteiro, Marks, and Ramm (1985) also used the MMMQ and found that at pre-treatment, there was no significant difference in agoraphobic symptoms, as assessed by the FQ, between participants qualified as being in a "good" versus a "less good" marriage (p > .05). Although not explicitly reported by the authors of the original study, the authors of the present review based their results on graphs presented in the original study. Chambless (1985) used the Marital Dissatisfaction Questionnaire (MDQ), which is a five-item questionnaire that assesses the discrepancy between the respondent's perception of his/her actual and ideal spouse. The author did not find any significant correlations between the level of marital dissatisfaction and severity of agoraphobia (r = .10, p > .05) or frequency of panic attacks (r = -.10, p > .05).
In their study, Lange and Van Dyck (1992) utilized the Interactional Problem Solving Inventory (IPSI), which is a self-report questionnaire that measures the extent to which partners are satisfied with their problem-solving abilities. They did not find any significant correlation between relationship quality and agoraphobic severity before treatment, except for the avoidance of busy streets subscale of the FAS-IR (r = -.29, p < .10). However, the significance level was set at .10 and there is no mention of whether or not the test was one-tailed or twotailed.
Finally, Tukel (1995) divided his 45 participants with PDA into three subgroups, those of housewives, working women, and working men. Participants were assessed on severity of PDA (FQ) and quality of marital relationship (MMQ). Results indicated a significant positive correlation between severity of PDA and quality of marital relationship for housewives (r = .61, p = .04). No significant correlations were found for the other subgroups (r = .15, p > .05 for working women; r = .10, p > .05 for working men).
Observational measures -Two studies included an observational measure of the interaction between patients and their relatives. Chambless and colleagues (2002) were interested in the marital interaction between couples in which one partner has PDA and a control group. They used the Kategoriensystem für Parnerschaftliche Interaktion (KPI), which is a system used to code a problem-solving interaction between two partners. During analysis of the interaction, each meaningful unit of speech is assigned a verbal and non-verbal code (e.g. positive, negative, or neutral). The authors found that panic frequency was not significantly related to any self-reported marital variables. However, they demonstrated that husbands whose wives were more Social Support and OCD/PDA Severity: A Systematic Review 266 avoidant engaged in a higher rate of negative verbal behavior (r = .44, p < .006) and were more critical (r = .35, p < .031) during the problem-solving interaction. The Renneberg and colleagues' study (2002), described in a section above, also used the KPI. They did not find a significant difference between marital adjustment and severity of panic or agoraphobic symptoms.

El-Baalbaki and colleagues (2011) used a different observational measure, the Global Couple Interaction
Coding System (GCIS). It also evaluates partners during a problem-solving situation but it evaluates each partner on five components of their verbal and non-verbal marital interaction. The five components are divided into three negative dimensions: (a) avoidance of and withdrawal from the discussion, (b) dominance, asymmetry in the control of the conversation, and (c) hostility, criticism, and conflict; and two positive dimensions: (a) support and validation, which reflect active listening and warmth, and (b) problem-solving skills.
Behaviors are rated according to four levels of severity (absent, mild, moderate, excessive). The authors found many significant correlations between aspects of the interaction and symptoms of panic and agoraphobia.
Lastly, quality of solutions by the spouse was negatively associated with fear of bodily sensations (r = -.25, p < .05). For further information, refer to Table 6. Table 6 Description

The Association Between PD/A Severity and Expressed Emotion
Two studies assessed the level of Expressed Emotion in relation to the severity of agoraphobia. Peter and colleagues (1993) reported significant associations between the severity of agoraphobia and the critics and emotional warmth subscales of the CFI (critics: r = .55, p < .01; emotional warmth: r = -.56, p < .01). In Rodde and Florin's (2002) study, 46 participants with PD/A and their partner were included. However, only results for 32 couples were reported (14 couples dropped out). Expressed Emotion status was assessed with the Five-Minute Speech Sample (FMSS), a five-minute monologue during which the partner is asked to talk about the patient and their relationship. There were no significant associations between Expressed Emotion status and fear of bodily sensations (BSQ). For a summary, see Table 7.

OCD Studies
Suffering from OCD can create major changes in the dynamics of an intimate relationship and the family.
Results presented above generally demonstrate that the level of support influence the severity of OCD. Indeed, all but one study assessing family accommodation found significant results, indicating that the more accommodation behaviors performed by the relatives, the more severe the OCD symptoms. Both metaanalyses also found a positive association between family accommodation and OCD severity. However, half of the sample of studies in the Strauss et al. (2015) meta-analysis were studies on pediatric OCD. Given that the authors did not examine the effect of age, it cannot be concluded that results would have been the same for adult patients only.
Although family members wish to alleviate the burden on the patients by modifying their habits and participating in the rituals, their behaviors seem to maintain and contribute to the severity of the disorder by favouring avoidance by the patients. It is also interesting to note that the authors of this study consider accommodation as a specific measure of social support because it is considered as support that is directly linked to the symptoms of OCD. A systematic review by Fredette and colleagues (2016) also found that PTSD specific measures of social support tended to be more associated with the outcome of cognitive behavioral therapy than global measures of support. Results relating to Expressed Emotion and OCD severity are mixed. Cherian and colleagues (2014) found that the level of criticism influenced OCD severity so that victims with relatives who are more critical of them seem to experience more severe symptoms. Van Noppen and Steketee (2009)  considered. This latter result supports the other studies, which found a robust link between accommodation and OCD severity. Finally, marital adjustment also seems to be associated with OCD severity, so that a better quality of relationship between partners is correlated with less severe symptoms. These results are based on two studies.

PD/A Studies
Studies analyzing the association between social or marital support and the severity of PD/A present mixed results. Researches using measures of social support seem to indicate that people with good support, either in terms of their perception of the quality of their support or the size of their network, present less severe symptoms. Concerning marital adjustment, results are mixed and the methods used were diverse. Nine studies used self-reported measures, from which two found that better marital adjustment was negatively correlated with symptoms of panic and agoraphobia. Interestingly, these two studies used the same questionnaire, the DAS, and found strikingly similar results. Another study found a difference between satisfied and unsatisfied dyads on measures of symptom severity. However, significance was not assessed. Surprisingly, one study (Tukel, 1995) also found that marital adjustment between housewives with PDA and their spouses was positively correlated with severity of disorder. There were no significant correlations between marital adjustment and PDA severity among working men and women. These results may be understood using the assortative mating hypothesis, which suggests that partners choose each other on the basis of perceived attributes (Hafner, 1977). For example, a woman with agoraphobia who has dependent traits may choose a partner with more dominant traits. Both partners would thus benefit from a dynamic where the husband endorses more responsibilities and takes care of his agoraphobic wife. Given that improvement in agoraphobic symptoms would lead to more autonomy from the wife and break this dynamic, it may also lead to a decline in marital satisfaction for both partners. Thus, it is possible that housewives and their husbands are more likely to have these attributes that allow them to benefit from an agoraphobic dynamic, in comparison to working men and women.
However, when patients are not distinguished according to their working status, most results seem to indicate that level of marital adjustment, as assessed by self-reported questionnaires, is not associated with the severity of panic and agoraphobic symptoms. Given that marital adjustment is a concept that is larger than marital support, it would be interesting to create more specific measures of marital support in order to eliminate factors that are not directly in link with support (e.g. sexuality).
Three studies also assessed marital adjustment with observational measures, two of which found significant results. These results indicated that negative social support (e.g. criticism and dominance) is associated with more severe symptoms while positive support (e.g. proposing positive solutions) is associated with less severe symptoms. However, one other study (Renneberg et al., 2002) found no significant differences between balanced-affected and unbalanced-affected dyads on measures of panic and agoraphobia severity. In order to understand the latter result, we performed statistical analyses to determine the effect sizes and statistical power, using descriptive data from Finally, two studies analyzed the link between Expressed Emotion and severity of PD/A. In the first study (Peter et al., 1993), it was found that criticism was negatively associated with the symptomatology of PD/A. Also, positive aspects of support were assessed with the CFI, and it was found that emotional warmth was associated with less severe symptoms. Given that Expressed Emotion assesses relatives' attitudes and behaviors toward the patient's illness, it is considered a specific measure of social support. As reported for the results concerning OCD, the specific way people deal with their partner's symptoms seems to be correlated with the severity of the symptomatology. In the second study (Rodde & Florin, 2002), no significant association was found between Expressed Emotion and the severity of PD/A. However, no statistics were presented, which makes it difficult to draw conclusions.

Conclusion
Results presented in this systematic review generally indicate that social and marital support is associated with the severity of OCD and PD/A, which supports our hypotheses. Indeed, negative social support seems to be associated with more severe symptoms whereas positive social support might be beneficial for people suffering from OCD or PD/A. These results stress the importance of relatives in helping people recover from their illness.
Living with someone suffering from a mental illness can be difficult for the relatives, as they might not know how to support or react to the patient's behaviors. Thus, solutions such as integrating the relatives in the therapy as well as providing them with tools (e.g. psychoeducation, personalized therapy for the relatives) on how to deal with the symptoms of the disorder might be beneficial to both the patient and the relatives. However, more studies would be needed to assess the level of support, using both observational and self-report measures of social support, as well as more specific measures as they seem to be more strongly correlated with the severity of OCD and PD/A. Moreover, studies that assess and compare both the negative and positive forms of social support would be interesting, since negative social support has been found to be more strongly correlated with symptom severity in a study of post-traumatic stress disorder (Zoellner, Foa, & Brigidi, 1999).
To conclude, this systematic review has some limitations. Indeed, the wide spectrum of questionnaires used by different researchers rendered it difficult to compare studies adequately. Moreover, the authors of the present study did not always have full access to the description of the questionnaires, which at times made it necessary to infer their content (e.g. MDQ). Moreover, the authors decided to cover a broad spectrum of concepts relating to social support (e.g. accommodation and marital adjustment) in order to render this systematic review as exhaustive as possible. Readers need to keep this in mind when interpreting the results. In order to compensate for this, the authors tried to only present results pertaining to social support when it was possible.
For example, only FAS total scores and results of the participation in the rituals subscale were used, since other subscales concerning the impact of accommodation on relatives were not manifestations of social support as it was defined in this review. Also, results concerning emotional over-involvement were left out, since it is not included in our definition of social support. Finally, some studies had limited statistics, which makes the interpretation of their results difficult.