The protocol of the review was prepared following the PRISMA-Protocol guidelines (PRISMA-P) [18] and it can be provided by the corresponding author on request.

For the present review, parental imprisonment was conceptualized as any kind of custodial confinement of a parent by the criminal justice system, including being held as a prisoner of war or for political reasons. Studies were included if they met the following criteria: (a) they assessed the presence of a PTSD diagnosis in offspring of prisoners, (b) they reported data necessary to calculate the effect sizes on the prevalence of a diagnosis of PTSD (total sample sizes and the number of participants with PTSD) or study authors were available to provide these data if they were not reported in the paper, (c) the diagnosis of PTSD was based on an official diagnostic classification system (i.e., any version of the DSM or ICD) and at least one structured/semi-structured diagnostic, clinician-administered interview [e.g., Clinician-Administered PTSD Scale (CAPS; [19]) or a self-report questionnaire, (d) the paper was published in English, French, German, Spanish or Italian language. Studies that were considered for inclusion could use any research design except reviews, single-case studies, case series, and case reports. Eligible designs included cross-sectional or case-control studies where the prevalence of PTSD was assessed in prisoners’ offspring. Longitudinal studies were included if they reported at baseline data regarding the prevalence of PTSD in this type of population or the authors were available to provide them if requested. Outcomes might have been measured any time after parental incarceration: while parents are in prison or after release, in childhood or in adulthood. Studies were included if they were conducted on child, adolescent or adult offspring.

Studies were identified by conducting an online systematic search of electronic databases and by using keywords related to PTSD (“Post-traumatic stress disorder” OR“Traumatization” OR “Trauma”) combined through the Boolean operator AND with keywords related to prisoners (“Prisoners” OR “Incarceration” OR “Pow” OR “Captivity”) and with keywords related to offspring (“Offspring” OR“ Sons” OR“Children”). The search procedure was conducted during the second week of November 2019 by using the databases Scopus and PubMed. No date restriction was applied.

In addition, all corresponding authors of the eligible studies were contacted to identify any further studies, irrespective of their publication status. Reference lists of the studies included in the meta-analysis were also examined.

Studies were screened against eligibility criteria by two authors (AP, FF) working independently in three stages. During the first and second stages, studies were examined with regards to the inclusion criteria after reading the title and the abstract, respectively. After each stage, the authors met to compare their selections. Studies were not excluded if there was disagreement between the authors on inclusion or exclusion. During the final stage, two authors examined independently the full text of the papers. At this stage, any disagreements about inclusion or exclusion of studies were discussed and resolved in a meeting with a third author (AC).

All the information was extracted from each included study by two authors (AP, FF) working independently and inserted into an Excel worksheet, which was piloted first on 2 included studies. The following information was extracted and coded from each study: (1) Title of the paper, (2) First author, (3) Publication date, (4) Country where the study was conducted, (5) Research design, (6) Recruitment setting (i.e., sites where participants were recruited and strategies used to recruit them), (7) Inclusion and exclusion criteria, (8) Total sample size, (9) Group size with PTSD, (10) Number of offspring without PTSD, (11) Mean age of the offspring total group, (12) Offspring generational cohort (coded as children/adolescents versus adults), (13) Total percentage of female offspring, (14) Instrument(s) used to make the PTSD diagnosis, (15) Total percentage of mothers, (16) Reasons for parental incarceration (war/political reasons versus crime reasons).

A third author (AC), not involved in the data extraction procedure, checked the correctness of the data entered in the worksheet. After entering the data, any discrepancies were discussed at a meeting between the authors who extracted the data and the third author.

A random-effect meta-analysis was conducted using the software Comprehensive Meta-Analysis, CMA version 2.00 [20]. The effect sizes were calculated as event rates, which were obtained as the ratio between the number of cases with PTSD and the total study sample size: higher effect sizes indicated higher prevalence rates of PTSD in offspring samples. The effect sizes were estimated by adopting a 95% confidence interval. Hedges’ correction for small sample bias was applied to all effect sizes [21].

Heterogeneity analysis of the effect sizes was conducted by calculating the I² statistic [22] and the Q index [21]. The I² index represents a measure of between-study heterogeneity in percentage, which is attributable to variability rather than chance [22]. A value approximating zero indicates homogeneity, whereas values of 25%-50%, 50%-75%, and 75%-100% represent low, medium, and large heterogeneity, respectively. The Q index is calculated 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 [21].

To assess the likelihood that the effect sizes were subjected to publication bias, two procedures were used: the visual inspection of the funnel plot and the Egger test [20]. The first represents a scatter plot in which the effect sizes computed from the included studies are plotted on the horizontal axis against an indicator of study precision, the standardized error, on the vertical axis [22]. In the absence of bias, the graph resembles a symmetrical inverted funnel, because the effect sizes derived from smaller studies scatter more widely at the bottom of the graph, with the spread narrowing with increasing precision among larger studies. If there is publication bias because smaller studies that report no significant effect sizes remain unpublished, the funnel plot appears asymmetrical [22]. The Egger test is an unweighted regression analysis based on the precision of each study as the independent variable and the effect size divided by its standard error as the dependent variable [23]. A non-statistically significant result of the t-test for the null hypothesis of an intercept equal to zero allows us to discard publication bias [23].

Three sensitivity analyses were performed calculating the effect sizes in (a) the studies including only adults or in those including only children/adolescents; (b) the studies conducted on offspring of parents incarcerated for political/war reasons or those incarcerated for crime; (c) the studies conducted in Western (Anglophone and European Union countries) or Non-Western countries (African, Middle-East, Central and Southern America, Asian countries).

The search through the databases and the additional sources produced a total of 181 records, assessed independently in the first phase by two authors. After duplicates were removed, 135 records were assessed via title/abstracts. This led to the exclusion of 76 records. Reviewing the full text of the remaining 49 records resulted in the exclusion of 43 articles. After this selection process, 6 studies (2512 participants overall) were included in the systematic review and meta-analysis. The PRISMA flowchart of the study selection process is provided in (Fig. 1).

The sample sizes in the included studies ranged from 79 to 1869 participants. Three studies were conducted in the United States, one in Israel, one in the State of Palestine, one in Lithuania. Three studies were conducted on adults, 3 studies on children. Two studies were conducted on the offspring of parents incarcerated for crime reasons and four studies conducted on the offspring of parents incarcerated for political/war reasons. All the studies were based upon a cross-sectional research design, except a longitudinal one [17]. All the papers were written in English and published in peer-review journals. The descriptive characteristics of the included studies are presented in Table 1.

Fig. (1). PRISMA flowchart of the study selection.

The mean effect size was statistically significant and showed that about 15% of the offspring of prisoners had PTSD (Event Rate = 0.146, 95% CI: 0.081 - 0249, Z = -5.228, p< 0.001, k = 6). The forest plot with study and mean effect sizes is provided in Fig. (2). There was no evidence of publication bias, as suggested by the funnel plot (Fig. 3) and by the Egger test of intercept which was not statistically significant[β = -2.853,SE = 1.909, t₍₄₎ = 1.494, p = 0.209]. For this analysis, a significant heterogeneity was found [I² = 92.637, Q₍₅₎ = 67.906, p < .001].

Three sensitivity analyses were carried out by calculating the mean effect size separately in (a) adult samples and child/adolescent samples, (b) in the studies conducted on offspring of parents incarcerated for political/war reasons and for crime reasons, and (c) in studies conducted in Western and non-Western countries. There was no significant difference in the mean effect sizes between the studies on adults and those on children/adolescents [Q₍₁₎ = 0.00, p= .999] (Fig. 4). No significant differences in the mean effect sizes were found between the studies conducted on the offsprings of parents incarcerated for political/war reasons and those for the crime [Q₍₁₎ = 0.00, p = .979] (Fig. 5) and between the studies conducted in Western and non-Western countries [Q₍₁₎ = 0.854, p = .355] (Fig. 6).

Subsequently, offspring’s gender and parents’ gender, coded as the percentage of female offspring and as the percentage of mothers, respectively, were assessed by running simple meta-regressions. While offspring’s gender was not related to the effect sizes [β = -0.01, 95% CI: -0.02 – 0.02, p = .452, k = 5], parents’ gender was significantly and positively associated with the effect sizes, suggesting that for the studies with higher percentages of incarcerated mothers, the prevalence of PTSD among the offspring was higher [β = 0.01, 95% CI: 0.0 – 0.01, p = .019, k = 5].

The relatively small number of studies highlights the importance of future studies on this topic and suggests that it is still neglected by researchers. The cross-sectional design of the included studies did not allow us to draw firm conclusions about the effect of parental incarceration as a risk factor for the development of PTSD. Other variables may explain the prevalence rates and may amplify the effects of incarceration. The role of further moderators should be assessed such as the quality of parent– child relationships before the incarceration, pre-existing life difficulties (e.g., early childhood trauma or prenatal exposure to drugs and alcohol), time since trauma occurrence (duration of untreated illness), child’s age at the time of incarceration, nature and length of the sentence, alternative care arrangements, number/quality of the contact with the incarcerated parent, parents’ psychiatric disorders and types of detention systems, or protective factors such as how other family members cope with the event and the wider social context [35-37]. The lack of control groups prevented us from comparing the prevalence rates of PTSD in prisoners’ offspring with those reported in children recruited from the general population. In addition, it may be interesting to compare the prevalence rates with those of children who have experienced the separation from parents for other reasons than incarceration. Another research gap regards the under-representation of European countries. Future research should investigate which PTSD symptom clusters are more severe among prisoners’ children. Finally, new investigations should focus on concurrent psychiatric conditions associated with PTSD in youth such as severe social withdrawal.

In conclusion, the present paper describes the first meta-analysis on the prevalence of PTSD among prisoners’ offspring. This psychiatric condition can be a serious mental health problem in this population, particularly for the offspring of incarcerated mothers. Our findings point out the importance of thorough assessment and timely intervention/prevention strategies implemented by professionals in mental health settings and detention systems.