A systematic review was conducted following the PRISMA guidelines [21]. The protocol of the review was regisStered on PROSPERO (2018 CRD42018106194). Eligibility criteria involved (a) Characteristics of participants, (b) ChaSracteristics of outcomes, (c) Characteristics of comparators, (d) Characteristics of design:

(a) Studies were included if they used a clinical group with a current primary OCD diagnosis. Diagnosis had to be established by a semi-structured clinical interview based on the criteria of a standardized diagnostic system such as the Structured Clinical Interview according to DSM-IV [SCID-I; 22] or by an unstructured clinical interview conducted by a mental health professional based on the criteria of a standardized diagnostic system such as the DSM-5 [1] or the ICD-10 [23]. Studies including patients with a lifetime diagnosis of OCD or including participants with subthreshold OCD were not included. Studies were included only if they used adolescent/adult participants, as the clinical characteristics of the disorder in children are significantly different than those of adolescent/adult OCD [24]. Studies using participants with primary hoarding were excluded since it is conceptualized as a distinct diagnosis in the DSM-5. Studies were included if they had recruited patients from primary, secondary or tertiary care settings. Concurrent psychological or pharmacological treatment was not a reason for exclusion. Studies were not considered to be excluded if the patients had concurrent general medical diseases. Studies conducted on OCD in the elderly where patients included individuals aged over 70 years old, were excluded.

(b) Studies were included if they evaluated perceived Physical Health Status by validated, internationally recognized self-report instruments, such as the Medical Outcomes Survey 36-Item Short-Form Health Survey [36; 25] or the World Health Organization Quality of Life-Brief Form [WHOQOL-BREF; 26].

(c) Studies were included if using control groups of screened participants not meeting the criteria for any of the mental disorders included in a standardized classification system and this condition was checked by a clinical interview by a mental health practitioner. In addition, studies were also included if they used control groups of unscreened participants drawn from the general population.

(d) Studies were included if they used a case-control design, where a group of patients with a current primary OCD diagnosis was compared with a healthy or a general population control group on perceived physical health. Any other type of research was allowed if the study provided the data necessary to compute an effect size estimate (for the data requested to compute the effect sizes, see paragraph “Meta-analysis and summary measures”). No publication data or language restriction was applied.

Studies were identified by carrying out an online systematic search of electronic databases and by using each of the keywords “Obsessive Compulsive Disorder”, “Obsessions”, “Compulsions” combined through the Boolean operator AND with the keywords “Physical Health”, “Physical Health Status”, “Physical Quality of Life”. The search procedure was conducted during the last week of November 2018 by using the electronic databases Scopus, PubMed, PsycINFO, EMBASE, Cochrane Library.

Subsequently, the corresponding authors of the included studies were contacted to request further data/to discover if they had further data. An inspection of all the references of the studies included in the review was also performed. An hand-search of conference proceedings was carried out in order to locate potential abstracts, papers, or posters relevant to OCD research presented at the following scientific associations: American Psychiatry Association, American Psychological Association, European Psychiatry Association, European Association of Psychology, British Psychological Society, Royal College of Psychiatrists. Theses and doctoral dissertations were hand-searched to identify additional unpublished eligible data.

Studies were assessed on eligibility criteria by two reviewers (AP, FF) independently during three different stages. During the first and the second stages, studies were assessed with regard to eligibility criteria after the reading of the title and of the abstract, respectively. During these stages, studies were retained when there was no agreement on inclusion between the reviewers. Finally, the studies remaining were assessed on eligibility criteria after the reading of the full text. In this selection stage, the reviewers discussed reasons for inclusion and any disagreements in judgement were addressed during meetings with another independent reviewer (AC) to obtain consensus on which studies to include in the pool.

All the information was extracted from each of the included studies by two reviewers (AP, FF) independently and inserted into an excel worksheet which was firstly piloted on 2 included studies. The following information was extracted and coded from each of the studies: (a) Title of the paper, (b) First author, (c) Publication date, (d) Country where the study was conducted, (e) Inclusion and exclusion criteria, (f) Total sample size, (g) Number of patients with OCD, (h) Number of controls, (i) Types of controls (screened participants without psychiatric disorders or unscreened participants from general population), (j) Mean and standard deviation of the OCD group on the perceived Physical Health Status outcome, (k) Mean and standard deviation of the control group on the perceived Physical Health Status outcome, (l) Mean and standard deviation of the OCD group on the measure of role limitations due to physical problems, (m) Mean and standard deviation of the control group on the measure of role limitations due to physical problems, (n) Mean and standard deviation of the OCD group on the measure of bodily pain, (o) Mean and standard deviation of the control group on the measure of bodily pain, (p) Total mean age and age range, (q) Total percentage of females, (r) Measure(s) used to assess perceived Physical Health Status, (s) Measure(s) adopted to evaluate role limitations due to physical problems, (t) Measure(s) to assess bodily pain, (u) OCD symptom severity (measured in terms of the Y-BOCS scores), (v) Research design, (w) Instrument(s) used to establish the OCD diagnosis, (x) Percentage of patients on concurrent medication, (y) Patients’ recruitment strategies, (z) Strategies used to recruit the controls, (aa) Comorbid personality disorders, (ab) Comorbid depression symptoms, (ac) Percentage of patients with concurrent general medical diseases.

The third independent reviewer (AC) not involved in the extraction procedure checked the correctness of the data inserted in the worksheet by the other two reviewers. After the insertion of the data was completed, discrepancies in the data extracted by the two reviewers were discussed and resolved in a final meeting between the reviewers who conducted the data extraction and the third independent reviewer.

When inconsistency analyses showed significant and high heterogeneity between the effect sizes, the role of moderators was investigated. Two independent reviewers (FF and AP) coded the moderators independently. Subsequently, during meetings between the two reviewers, data insertion in the worksheet was checked for accuracy and potential discrepancies were resolved with a third reviewer (AC). The following variables were coded as moderators: (a) mean sample age; (b) sample gender (coded as the percentage of females); (c) OCD symptom severity, coded as a continuous variable based on Yale-Brown Obsessive Compulsive Scale scores [Y-BOCS; 27], which is the gold standard symptom measure; (d) study publication date; (e) study methodological quality according to the Newcastle-Ottawa Scale scores [NOS; 28] (see Quality Assessment paragraph below).

The NOS was used to judge the methodological quality of the studies. This tool has recently been recommended by systematic review practice guidelines as the most reliable instrument for conducting a quality assessment of cross-sectional/cohort studies [29]. The NOS includes eight items grouped into three domains: (a) Selection, (b) Comparability, (c) Outcome (cohort studies) or exposure (case-control studies) according to the study design. For each item, a series of response options are provided. A star system is adopted to allow a semi-quantitative quality assessment. The highest quality studies are assigned a maximum of one star for each item, excepting the item related to comparability where two stars are allowed. The scores on the NOS range from zero to nine stars. Two reviewers (AP, FF) performed the quality assessment independently. Discrepancies in the assignment of the scores were resolved in a consensus meeting with an independent third reviewer (AC).

The meta-analysis was calculated using random-effect models, which assume that the included studies are drawn from populations of studies that systematically differ from each other [30]. Effect sizes were calculated as standardized mean differences (SMD) by computing the following formula reported in Cohen [31]: (M_OCD-M_CONTROL)/SD_POOLED, where M_OCD is the mean of the OCD groups on the perceived Physical Health Status instruments (or the measure of role limitations due to physical problems or the measure of bodily pain), M_CONTROL indicates the mean of the controls on that measure and SD_POOLED is the pooled standard deviation. The effect sizes were estimated with a 95% confidence interval and interpreted according to the criteria described by Cohen [31]: values equal to 0.80 or higher were judged as large, values up to 0.50 as medium, and values up to 0.20 as small. When a study reported the data on more than one instrument to assess Physical Health Status, such as on both the SF-36 Physical Health Status scale and the WHOQOL-BREF Physical Health Status scale, a mean effect size was calculated by combining the effect sizes related to the scores on all the instruments. A standardized mean difference was calculated separately also for the data obtained from the SF-36 Role Limitations due to Physical Problems scale to summarize the evidence about the perceived negative interference of physical health-related problems. Finally, a standardized mean difference was calculated separately also for the data obtained from the SF-36 Bodily Pain scale to summarize the data related to perceived physical pain in OCD.

To verify publication bias, three different procedures were adopted including the Duval and Tweedie's trim and fill procedure [32], the visual inspection of the funnel plot and the Egger test.

Sensitivity analyses were performed by computing the effect sizes only in the studies (a) using the SF-36 Physical Health Status scale, (b) using adults, (c) using healthy screened controls, (d) using OCD patients without general medical disorders.

To assess between-studies heterogeneity, two indices were used, the I² [33] and the Q statistic [34], respectively. The I² is expressed as a percentage attributable to variability rather than chance [33]. A value close to zero indicates homogeneity, whereas values of 25%-50%, 50%-75%, and 75%-100% indicate low, moderate, and high heterogeneity, respectively. The Q statistic is computed by summing the squared deviations of each study’s effect estimate from the overall effect estimate while weighting the contribution of each study by its inverse variance [35]. In the hypothesis of homogeneity among the effect sizes, the Q statistic follows a chi-square distribution with k − 1 degrees of freedom, k being the number of studies. The moderators' analysis was conducted by performing weighted least squares meta-regressions. The meta-analysis was carried out using the Comprehensive Meta-analysis software version 2.00.

The electronic search and the search through additional sources produced 2512 records after removing duplicates. Of these, 2467 were excluded by the title or abstract as being on irrelevant constructs. Thus, 45 studies were full-text screened for inclusion. Sixteen studies were excluded for not using Physical Health Status measures or not assessing Physical Health Status/Role Limitations due to Physical Problems/Pain. Nine studies were excluded as they did not include a control group. Six studies were excluded since they were conducted on child samples. After this selection, fourteen studies were included by the consensus of the three independent assessors in the systematic review and meta-analysis (n= 20,223, 15 effect sizes). The Flow Chart is shown in Fig. (1).

All included studies were in English and published in peer-reviewed journals. Publication date ranged from 1996 to 2018. Three studies were conducted in Europe [36-38], three in North America [39-41], three in South America [5, 42, 43], three in Asia [44-46], two in the Middle East [47, 48]. All studies used adults, except for one [44] using adolescents/ adults (age range = 16 -70 years) and one using adolescents [42]. Six studies compared OCD patients with screened healthy controls [40, 42, 44, 46-48], eight compared OCD patients with controls recruited from the general population [5, 36-39, 41, 43, 45]. Eight studies used the SF-36 Physical Health Status scale [5, 36, 37, 39-41, 43, 48]; seven studies used the WHOQOL-BREF Physical Health Status scale [38, 39, 42, 44-47]. The included studies used different measures of depression: four studies [36, 37, 39, 48] used the HAM-D, three [5, 42, 44] used the BDI or the BDI-II, one [45] used the DASS-21, whereas the other studies did not assess depression.

Three studies [36, 38, 46] reported the proportion of the patients with personality disorders (51%, 0%, and 0%, respectively). Three studies [41, 44, 48] reported the proportion of the patients on concurrent pharmacotherapy (100%, 54%, and 100%, respectively). Five studies excluded patients with general medical disorders [5, 45-48], seven did not report information on comorbidity with general medical disorders [36, 38-40, 42-44], one reported that 74.60% of the group had comorbid medical disorders [37] and one reported 61.66% [41]. The supplementary file presents an overview of study characteristics.

Fig. (1). PRISMA flowchart.

Six studies received 5 points on the NOS [5, 38, 39, 41, 45], three received 6 points [36, 37, 47], four 7 points [40, 42, 44, 48] and one study [46] 9 points.

The definition of cases was judged as adequate for all the studies. All the studies were considered to have reported some independent validation. Representativeness of cases was judged as adequate for all the studies. Selection and definition of controls were judged as adequate for all the studies. Comparability of the subjects across the included studies was judged as adequate only in one study [46]. Ascertainment of exposure was judged as adequate for all the studies. Six studies did not use the same method of ascertainment of exposure for cases and controls and this item of the NOS was not judged as adequate for this subgroup of studies [5, 36-39, 41]. For six studies the provision of the information about the non-response rate was not judged as adequate [5, 38, 39, 47, 41, 45]. The quality assessment according to the criteria of the NOS is in Table 1.

An overview of all the analyses is in Table 2. The comparison of perceived Physical Health Status showed a large significant-negative-effect size favouring controls over patients with OCD (SMD = -0.97, SE = 0.25, 95% CI: -1.46 - 0.45, p < 0.001). Controls reported significantly higher levels of perceived Physical Health Status than patients with OCD (Fig. 2). Evidence of publication bias was not observed as the values of the Egger test were not significant [B = -6.12, SE = 4.00, 95% CI: -14.86 - 2.60, t₍₁₂₎ = 1.52, p = 0.15] and the funnel plot did not appear asymmetrical (Fig. 3). Absence of publication bias was also supported by the trim and fill procedure showing that the mean effect size did not change when it was adjusted for publication bias (SMD = -0.95, 95% CI: -1.46 - 0.45; number of trimmed studies =0]. A significant heterogeneity was observed since the result of the Q test was significant [Q₍₁₃₎ = 560.80, p <0.001], and the value of the I² was higher than 75 (I² = 97.68).

Fig. (2). Forest plot of perceived Physical Health Status between OCD patients and controls.

Fig. (3). Funnel plot.

In a sensitivity analysis, the mean effect size was calculated by including only the studies (k = 8) using the SF-36 Physical Health Status scale [5, 36, 37, 39-41, 43, 48]. The results showed a medium significant effect size favouring controls over patients with OCD (SMD = -0.63, SE = 0.26, 95% CI: -1.15 - 0.10, p < 0.05). Evidence of publication bias was not observed as the result of the Egger test was not significant [B = -8.88, SE = 4.75, 95% CI: -20.851 - 2.74, t₍₆₎ = 1.86, p = 0.11]. Absence of publication bias was also supported by the trim and fill procedure which showed that the mean effect size did not change when it was corrected for publication bias (SMD = -0.63, 95% CI: -1.15 - 0.10; number of trimmed studies =0].

Another sensitivity analysis included only the studies using adults (k = 12) [5, 36-41, 43, 45-48]. A medium significant effect size emerged favouring controls over patients with OCD (SMD = -0.60, SE = 0.22, 95% CI: -1.05 - 0.16, p < 0.01). Evidence of publication bias was not observed as the Egger test was not significant [B = -2.52, SE = 4.06, 95% CI: -11.58 - 6.53, t₍₁₀₎ = 0.10, p = 0.54]. Absence of publication bias was also supported by the trim and fill procedure showing that the mean effect size did not change when corrected for publication bias (SMD = -0.60, 95% CI: -1.05 - 0.15; number of trimmed studies =0).

The mean effect size was calculated also by including only the studies using screened controls (k = 6) [40, 42, 44, 46-48]. The findings showed a large significant effect size favouring controls over patients with OCD (SMD = -1.53, SE = 0.83, 95% CI: -3.16 - 0.09, p < 0.001). Evidence of publication bias was not found as the Egger test was not significant [B = -14.65, SE = 12.23, 95% CI: -48.62 - 19.31, t₍₄₎ = 1.19, p = 0.29]. Absence of publication bias was supported by the trim and fill procedure which showed that the mean effect size did not change when it was corrected for publication bias (SMD = -1.53, 95% CI: -3.16 - 0.09; number of trimmed studies = 0).

A further sensitivity analysis included only the studies using patients without comorbid medical conditions (k = 6) [5, 39, 45-48]. The results showed a large yet non-significant effect size favouring controls over patients with OCD (SMD = -0.91, SE = 0.51, 95% CI: -1.90 - 0.08, p = 0.08). Evidence of publication bias was not found by the Egger test, which was non-significant [B = -1.45, SE = 6.35, 95% CI: -48.62 - 19.31, t₍₄₎ = 0.22, p = 0.82] but it was supported by the trim and fill procedure showing that the mean effect size changed when corrected for publication bias (SMD = -1.57, SE = 0.51, 95% CI: -1.90 - 0.08; number of trimmed studies = 2).

The comparison on bodily pain showed a significant, yet small, positive effect size favouring patients with OCD over controls (SMD = 0.22, SE = 0.05, 95% CI: 0.11 - 0.33, p < 0.001, k = 6): patients reported significantly higher levels of bodily pain than controls. No evidence of heterogeneity emerged [Q₍₅₎ = 6.50, p = 0.26, I^{2 =} 23.09]. Absence of publication bias was supported by the trim and fill procedure showing that the mean effect size did not change when corrected for publication bias (SMD = -0.22, 95% CI: 0.11 - 0.33; number of trimmed studies =0) and confirmed also by the Egger test, which was not significant [B = 2.01, SE = 0.97, 95% CI: -0.70 - 4.73, t₍₄₎ = 2.06, p = 0.10].

The comparison of SF-36 Role Limitations due to Physical Problems showed a medium significant positive effect size favouring patients with OCD over controls (SMD = 0.55, SE = 0.14, 95% CI: 0.83 - 0.26, p < 0.001, k = 6): patients reported significantly higher levels of role limitations due to physical problems than controls. Significant heterogeneity emerged [Q₍₅₎ = 41.59, p < 0.001, I²⁼ 87.98]. Absence of publication bias was supported by the trim and fill procedure showing that the mean effect size did not change when corrected for publication bias (SMD = 0.55, 95% CI: 0.83 - 0.26; number of trimmed studies =0). Absence of publication bias was confirmed also by the Egger test which was not significant [B = -1.60, SE = 3.45, 95% CI: -11.19 - 7.98, t₍₄₎ = 0.46, p = 0.65].

Age was negatively associated with the effect sizes: higher age was associated with larger standardized mean differences in effect sizes on perceived Physical Health Status between patients with OCD and controls (B = -0.01, SE = 0.01, 95% CI: -0.02 - 0.01, p < 0.001, k = 13). Female gender was negatively associated with the effect sizes: higher percentage of females was associated with larger standardized mean differences in effect sizes on perceived Physical Health Status between patients with OCD and controls (B = -0.03, SE = 0.01, 95% CI: -0.04 - 0.02, p < 0.001, k = 14). OCD severity was positively associated with the effect sizes: higher OCD severity was associated with narrower standardized mean differences in effect sizes on perceived Physical Health Status between patients with OCD and controls (B = 0.21, SE = 0.01, 95% CI: 0.18 - 0.23, p < 0.001, k = 11).

Publication date was negatively associated with the effect sizes: more recent publication dates were associated with larger standardized mean differences in effect sizes on perceived Physical Health Status between patients with OCD and controls (B = -0.05, SE = 0.01, 95% CI: -0.06 - 0.03, p < 0.001, k = 14).

Methodological quality, coded through the scores on the NOS, was positively associated with effect sizes. Studies with higher methodological quality were associated with lower standardized mean differences in effect sizes on perceived Physical Health Status between patients with OCD and controls (B = 0.10, SE = 0.03, 95% CI: 0.02-0.16, p = 0.008, k = 14).

Some shortcomings should be pointed out. Firstly, the cross-sectional design of the studies does not allow conclusions to be drawn about the causal effect of OCD symptoms on perceived physical health but can only suggest an association. It might be argued that a poorer perception of Physical Health Status can induce obsessive fears or exacerbate to some extent pre-existing fears or compulsive behaviours. Thus, the investigation of perceived physical health needs for prospective designs. Secondly, seven of the studies did not report information about comorbid medical conditions. This lack of information prevented the adjustment of the analyses, and the sensitivity analysis was conducted on a small subgroup.

Only three out of the studies provided information about the number of patients on concurrent psychopharmacological treatment: 100% of the samples in two studies [41, 44] and 54% in one [48]. The lack of this information does not allow the effects of psychopharmacotherapy on perceived physical health to be accounted for. It may be argued that medication routinely prescribed for OCD at higher dosages than for anxiety or depressive disorders, such as Selective Serotonin Reuptake Inhibitors (SSRIs), may be associated with side-effects that impact physical health such as nausea, dizziness sedation, insomnia, and sexual dysfunction [54]. In addition, since OCD is typically a resistant disorder, it is often necessary to associate different classes of drugs in addition to the SSRIs such as Atypical Antipsychotics, as evidenced also by a recent review [55], with the risk of greater side effects and less perception of one's health. Another variable potentially moderating the association between OCD and physical health might be depression [56, 57]. However, in the present studies there was a large heterogeneity in the instruments used to measure it and this did not allow depression to be investigated as a moderator. As previously mentioned, future research should compare perceived physical health in OCD with other severe mental disorders such as psychosis or potentially with other conditions belonging to the so-called OCD spectrum, such as body dysmorphic disorder or skin picking disorder which typically involve a negative body experience [58, 59].

In conclusion, this is the first systematic review on Perceived Physical health in OCD: this quality of life domain should be considered more carefully by researchers in future investigations and by clinicians as a target of treatment, particularly with older, female and less severe patients. New interventions for Physical Health status in OCD may be evaluated.