Round 1

Reviewer 1 Report (Previous Reviewer 2)

Comments and Suggestions for Authors

This is a re-submitted manuscript which was rejected before. I have carefully checked all the revisions made by the authors to my comments in the original round. I believe the authors have already addressed all my concerns in this version of the manuscript.

Author Response

We thank the reviewer for checking our previous revisions.

Reviewer 2 Report (Previous Reviewer 1)

Comments and Suggestions for Authors

The changes and adjustments made to the manuscript improve the study. The authors are to be commended. In light of this, I recommend the manuscript for publication.

Author Response

We thank the reviewer for this positive assessment.

Reviewer 3 Report (New Reviewer)

Comments and Suggestions for Authors

The study combines a quantitative survey with a qualitative exploration of students' own explanatory models and this mixed design is one of its main strengths. The study shows how biomedical and cultural-spiritual interpretations coexist even among future health professionals, a finding with clear implications for medical education and mental health literacy. I particularly appreciate the authors' transparency regarding the limitations of their design. The comments below are intended to strengthen the methodological transparency of the instruments and analyses and to align the interpretation of the findings more closely with what the data can support.

 Observations:

1. Knowledge is measured with agreement-based Likert items, which capture beliefs rather than knowledge. Please provide all items (supplementary material) as this was not available to me as a reviewer. Furthermore, please consider to include the instrument as an appendix of the article, rather than as a supplementary material.
2. Considering the measurement of the outcome, justify why a validated instrument (e.g., SP-EPQ, or items from the Fearful Isolated Sleep Paralysis Interview) was not used.
3. If knowledge items describing SP preceded it, recognition may have been primed and prevalence inflated. Please state whether participants were shown a definition of SP before answering and discuss possible confusion with sleep inertia, hypnic jerks, nightmares or limb numbness. Also, please consistently refer to "self-reported lifetime SP experience" rather than implying a clinical condition, since a single episode does not meet criteria for recurrent isolated sleep paralysis.
4. The Polish study used a standardized instrument, the Sleep Paralysis Experience and Phenomenology Questionnaire (SP-EPQ), whereas the present study relies on a single ad-hoc item. The two prevalence figures are therefore not directly comparable. Please revise the Discussion accordingly: acknowledge this methodological difference explicitly, avoid describing the Guinean estimate as "substantially exceeding" the rates reported in Poland and Argentina and frame these comparisons with appropriate caution.
5. The statement that 61.9% "substantially exceeds" other rates should be tempered and anchored in meta-analytic estimates (e.g., Sharpless & Barber, 2011; Denis et al., 2018) and in available African data. Please clarify what is meant by "culturally specific contexts" for Turkey and Egypt.
6. Academic stress and irregular sleep were not measured in this study. Please present them explicitly as hypotheses for future research rather than as explanations of the findings.

Author Response

We thank the reviewer for these comments, which led to substantial changes in how knowledge and the outcome are measured and interpreted.

Comment 1. Knowledge is measured with agreement-based Likert items, which capture beliefs rather than knowledge. Please provide all items (supplementary material) as this was not available to me as a reviewer. Furthermore, please consider to include the instrument as an appendix of the article, rather than as a supplementary material.

We agree. The instrument is now included in the article as Appendix A. It gives the order of the questionnaire sections and the response format. Table A1 lists the eight knowledge, five attitude and five perception statements in French with an English translation, together with the distribution of answers and the proportion agreeing among students with and without SP.

We also changed how knowledge is scored. Re-examining the data, we found that agreement ratings were positively correlated across correct and incorrect statements alike. For instance, students with SP agreed more often than other students both with the correct statement that SP involves an inability to move or speak (92.4% vs. 64.8%) and with the incorrect statements that SP occurs only at sleep onset (79.8% vs. 48.4%) and that it is purely spiritual or supernatural (44.9% vs. 18.0%). This pattern supports the reviewer’s point: agreement measures endorsement, not knowledge. Each statement is now scored as correct when the student agreed with a correct statement or disagreed with an incorrect one, and the knowledge score is the number of correct answers (0–8; KR-20 0.59). Two statements scored as correct are less clear-cut (“SP is a common sleep disorder”; “effective medical treatments exist”); this is stated in the text and in Table A1, and a six-item score without them also differed between students with and without SP (p = 0.002). With this score, each additional correct answer was associated with higher odds of reporting SP (aOR 1.31, 95% CI 1.12–1.54). We also examined response style. Twenty-two students, none of whom had heard of SP, answered “neutral” to all 13 knowledge and attitude statements. The knowledge association persisted when they were excluded (aOR 1.26) and after adjustment for having heard of SP (aOR 1.24). It was weaker and not significant among the 226 students who had heard of SP (aOR 1.16, 0.92–1.46). The association with the attitude score did not persist when these 22 students were excluded. We report these analyses in Section 3.5.4 and interpret the scores in Section 4.3 as largely reflecting familiarity with SP. Changes: Sections 2.5, 3.3, 3.5, 4.2, 4.3, 4.4; Tables 4, 5a, 5b and A1; Figures 2 and 3.

Comment 2. Considering the measurement of the outcome, justify why a validated instrument (e.g., SP-EPQ, or items from the Fearful Isolated Sleep Paralysis Interview) was not used.

No validated French version of the SP-EPQ or of the Fearful Isolated Sleep Paralysis Interview was available, and we found none in the literature. The latter is also a clinician-administered interview, which could not be used in a self-administered survey completed in class. We now state this in Section 2.5 and in the limitations (Section 4.4). We also state that prevalence estimates obtained with these instruments are not directly comparable with ours. Section 4.5 recommends translating and validating one of them for future studies.

Comment 3. If knowledge items describing SP preceded it, recognition may have been primed and prevalence inflated. Please state whether participants were shown a definition of SP before answering and discuss possible confusion with sleep inertia, hypnic jerks, nightmares or limb numbness. Also, please consistently refer to “self-reported lifetime SP experience” rather than implying a clinical condition, since a single episode does not meet criteria for recurrent isolated sleep paralysis.

No definition was shown. The SP item (“Have you ever experienced an episode of sleep paralysis?”) was in the practices section. It came after the knowledge section, which included the statement “During sleep paralysis, a person is unable to move or speak”. This statement may have acted as an implicit definition and primed recognition, as the reviewer suggests. We now state this in Sections 2.5 and 2.6 and discuss it in Section 4.1. One finding is consistent with some misclassification: 31 of the 94 students who had never heard of SP reported having experienced it (Section 3.3). Possible confusion with hypnic jerks, sleep inertia, nightmares and transient limb numbness is discussed in Sections 2.5 and 4.1. The questionnaire did not screen for narcolepsy either, and this is acknowledged.

To show how much the estimate depends on the case definition, we added three stricter definitions. The proportion was 56.6% when inability to move was required, 39.7% when recurrent episodes with inability to move were required, and 26.9% when intense fear was also required; the last definition is the closest approximation to recurrent isolated sleep paralysis that our data allow (Section 3.3). The associations with knowledge and attitude were similar under the first two definitions, and the knowledge association was weaker under the strictest one (Section 3.5.4). The term “self-reported lifetime SP experience” is now used throughout; the shorthand “students with SP” is defined once in Section 2.5, and Section 4.3 is now titled “Factors Associated with Self-Reported SP”. Section 2.5 states that a single episode does not meet the ICSD-3 criteria for recurrent isolated sleep paralysis.

Comment 4. The Polish study used a standardized instrument, the Sleep Paralysis Experience and Phenomenology Questionnaire (SP-EPQ), whereas the present study relies on a single ad-hoc item. The two prevalence figures are therefore not directly comparable. Please revise the Discussion accordingly: acknowledge this methodological difference explicitly, avoid describing the Guinean estimate as “substantially exceeding” the rates reported in Poland and Argentina and frame these comparisons with appropriate caution.

We revised Section 4.1 accordingly and removed the phrase “substantially exceeds”. The Discussion now explains that the SP-EPQ describes the experience in its screening question and checks positive answers against the respondent’s own account, whereas we used a single undefined item. We also note that the instrument effect can work in both directions, since questionnaires designed for SP have given higher estimates than non-specific instruments in a recent meta-analysis. Our main figure is therefore presented as the proportion of students who identified with the label of SP, not as a clinical prevalence.

Comment 5. The statement that 61.9% “substantially exceeds” other rates should be tempered and anchored in meta-analytic estimates (e.g., Sharpless & Barber, 2011; Denis et al., 2018) and in available African data. Please clarify what is meant by “culturally specific contexts” for Turkey and Egypt.

The Introduction and Section 4.1 now cite the meta-analytic estimates. In Sharpless and Barber (2011), lifetime prevalence was 7.6% in the general population, 28.3% in students and 31.9% in psychiatric patients. In the 2024 meta-analysis by Hefnawy et al., the pooled prevalence was 30%, and 34% among students. Denis et al. (2018) review the correlates of SP but do not pool prevalence, so we cite them for the correlates (sleep quality, stress, trauma, anxiety). We added the African data we could find: in Nigeria, 26.1% of medical students, 44% of nursing students and 22.5% of young adults; in Sudan, 29.9% of college students; in Egypt, 43% of college students and 44% of the general population. We also report that no published prevalence data were found for other African countries.

On “culturally specific contexts”: re-reading the Turkish study, we found that it does not report a prevalence. It included only 59 students who had already experienced SP, so the figure of 38.4% in our previous version was wrong. The phrase has been removed. Instead, the Introduction states what these studies show about culture: in Turkey, 88% of affected students called SP Karabasan and 17% attributed their episodes to this creature; in Egypt, about half of general-population respondents attributed SP to the jinn.

Comment 6. Academic stress and irregular sleep were not measured in this study. Please present them explicitly as hypotheses for future research rather than as explanations of the findings.

Done. Section 4.1 now states that these factors were not measured and that whether they contribute to the high rate is “a hypothesis for future research, not an explanation of our findings”. The same framing is used in Sections 4.4 and 4.5 and in the Highlights. We also removed from the Abstract the sentence presenting sleep deprivation and academic stress as risk factors in this population.

Reviewer 4 Report (New Reviewer)

Comments and Suggestions for Authors

This manuscript investigates the prevalence of sleep paralysis (SP) and its associated factors among 320 medical science students from three universities in Conakry, Guinea. The study addresses an interesting and relatively understudied public health topic, particularly in the context of sub-Saharan Africa. The incorporation of knowledge, attitudes, practices, and cultural perceptions provides a broader perspective on SP beyond its clinical manifestations. Overall, the study has merit, and the authors have appropriately acknowledged several limitations associated with its cross-sectional design, convenience sampling, and self-reported measurements. However, there are several methodological and statistical concerns that should be addressed to strengthen the validity and interpretation of the findings.
1. The authors used backward stepwise selection with statistical significance thresholds of p < 0.20 for entry and p < 0.05 for retention. While this approach is commonly used in exploratory studies, relying entirely on data-driven variable selection can produce unstable models, particularly with a relatively modest sample size. It may also lead to biased coefficient estimates and overly optimistic confidence intervals because uncertainty associated with the selection process is not accounted for.
2. Participants were recruited from three universities, but the statistical analysis does not indicate whether the potential clustering of observations within institutions was considered. Students from the same university may share similar academic environments, stress levels, and educational experiences, potentially resulting in correlated observations. Ignoring such dependence could lead to inaccurate standard errors and confidence intervals. The authors should clarify whether university affiliation was considered in the regression analyses and discuss the potential impact of institutional clustering.
3. The manuscript states that data completeness was at least 98% and that no imputation was performed. However, it remains unclear whether the regression analyses were conducted using complete cases and how many observations were included in the final multivariable model. The authors should report the number of observations excluded because of missing data and the final analytic sample size.
4. Regarding model evaluation, the authors report using the Hosmer–Lemeshow goodness-of-fit test and AUC-ROC. If the AUC was calculated using the same sample employed for variable selection and model fitting, the reported discrimination may be optimistic. The authors should clarify whether any internal validation, such as bootstrap resampling, was performed.

There are also several minor comments:
1. Several sentences in Sections 2.3 and 2.4 appear to be duplicated. These should be removed to improve readability.
2. The authors should report the distribution of participants across the three universities and the number of SP cases within each institution to facilitate assessment of the sample composition.

 

Author Response

We thank the reviewer for these statistical comments, which we have addressed by changing the modelling strategy.

Comment 1. The authors used backward stepwise selection with statistical significance thresholds of p < 0.20 for entry and p < 0.05 for retention. […] relying entirely on data-driven variable selection can produce unstable models, particularly with a relatively modest sample size. It may also lead to biased coefficient estimates and overly optimistic confidence intervals because uncertainty associated with the selection process is not accounted for.

We agree. We also found that our previous description was inaccurate: the selection had been based on the Akaike information criterion (stepwise, both directions), not on p-value thresholds. The primary analysis is now a multivariable model specified in advance. It includes all candidate variables (sex, age, nationality, university, year of study, knowledge score, attitude category) without selection: 13 parameters, 9.4 students in the smaller outcome group per parameter (Section 2.7, Table 5b, Model 1). In this model, the knowledge score (aOR 1.31, 1.12–1.54) and a lower attitude score (aOR 0.35, 0.13–0.91) were associated with SP, whereas sex was not (aOR 0.67, 0.40–1.14).

Stepwise selection is reported only for comparison (Model 2), and we assessed its stability in 1000 bootstrap samples. Knowledge was retained in 96.6% of samples and attitude in 76.6%, but sex in only 52.3%, and the selected model was chosen again in only 5.3% of samples (Table 5b, Section 3.5.2). Firth’s penalized likelihood gave similar estimates (Section 3.5.4). We therefore base our conclusions on the pre-specified model. The additional sensitivity analyses showed that the attitude association is not robust, as explained in our response to Reviewer 3, Comment 1. Male sex is no longer described as an independent factor in the Abstract, Highlights and Discussion.

Comment 2. Participants were recruited from three universities, but the statistical analysis does not indicate whether the potential clustering of observations within institutions was considered. […] The authors should clarify whether university affiliation was considered in the regression analyses and discuss the potential impact of institutional clustering.

University is now included as a fixed effect in the primary model (UKAG aOR 2.18 and La Source aOR 2.38 versus UGANC; likelihood-ratio test p = 0.08). With only three clusters, random-effects variance estimates and cluster-robust standard errors are unreliable, and we explain this in Section 2.7. As a check, we fitted a random-intercept model for university: the between-university variance was estimated at zero, the intraclass correlation in the model without covariates was 0.003, and the other estimates were unchanged (Section 3.5.4). Clustering within institutions therefore did not affect the conclusions. Section 4.3 notes that the difference between universities cannot be generalized with three institutions.

Comment 3. […] it remains unclear whether the regression analyses were conducted using complete cases and how many observations were included in the final multivariable model. The authors should report the number of observations excluded because of missing data and the final analytic sample size.

There were no missing values for any variable used in the models or for any of the Likert items. All models were therefore fitted on the full sample of 320 students, 198 of whom reported SP, with no exclusion. Missing values were confined to optional free-text fields and to the country of origin (6 of 212 foreign students). This is now stated in Section 2.7 and in the titles of Tables 5a and 5b. The previous statement “data completeness was ≥98%” has been replaced.

Comment 4. […] If the AUC was calculated using the same sample employed for variable selection and model fitting, the reported discrimination may be optimistic. The authors should clarify whether any internal validation, such as bootstrap resampling, was performed.

No internal validation had been performed. We now use Harrell’s bootstrap method (1000 resamples, with the selection step repeated in each resample for the stepwise model). The primary model had an apparent AUC of 0.71 (95% CI 0.65–0.77), an optimism-corrected AUC of 0.66 and a calibration slope of 0.75. For the stepwise procedure, the AUC fell from 0.68 to 0.64. The Hosmer–Lemeshow test did not indicate poor calibration (p = 0.13). These results are reported in Section 3.5.3, Table 5b and Figure 5. The text now states that discrimination is modest and that the model describes associations rather than predicting SP.

Minor comment 1. Several sentences in Sections 2.3 and 2.4 appear to be duplicated.

The duplicated sentences in Sections 2.3 and 2.4 have been removed, as has a duplicated sentence in Section 4.2.

Minor comment 2. The authors should report the distribution of participants across the three universities and the number of SP cases within each institution.

The sample comprised 52 students from UGANC, 144 from UKAG and 124 from La Source University. Of these, 26 (50.0%), 96 (66.7%) and 76 (61.3%), respectively, reported SP (p = 0.10). These figures are given in Section 3.1 and Tables 1 and 4.

We hope that the revised manuscript meets the reviewers’ expectations, and we remain available for any further changes.

Yours sincerely,

Houssainatou Bah and Aboubacar Sidiki Magassouba, on behalf of all authors

Round 2

Reviewer 3 Report (New Reviewer)

Comments and Suggestions for Authors

no suggestions

Reviewer 4 Report (New Reviewer)

Comments and Suggestions for Authors

I would like to thank the authors for their thorough and constructive response to the previous round of review. The revised manuscript has satisfactorily addressed my major statistical concerns and demonstrates substantial methodological improvement. Most notably, the authors have replaced the original stepwise variable-selection approach with a prespecified multivariable logistic regression model, while providing bootstrap stability assessments and sensitivity analyses to evaluate the robustness of their findings. They have also appropriately accounted for university affiliation through fixed effects and further examined potential institutional clustering using a random-intercept model. Additionally, the authors have clarified that the primary analyses included all 320 participants without missing data and have strengthened model evaluation through bootstrap internal validation, including optimism-corrected discrimination and calibration assessments. Importantly, the revised interpretation appropriately acknowledges the limited robustness of certain associations and the modest discriminatory performance of the model. These methodological improvements substantially enhance the transparency, statistical rigor, and reliability of the findings. 

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Editor and Authors,

This manuscript addresses a relevant and underexplored topic in African contexts, especially regarding sleep paralysis among health science students. The study has merit for integrating quantitative and qualitative analysis, as well as exploring cultural and spiritual dimensions associated with the disease. However, despite its thematic relevance, the manuscript presents important methodological, analytical, and conceptual limitations that reduce its scientific robustness.

Some items that could be addressed in the review: The fragility of the psychometric assessments and the obtaining of knowledge and attitude scores; some analytical inconsistencies were verified in the logistic regression and, consequently, an inadequate causal interpretation; which may be established by a lack of control for possible confounders. The qualitative analysis needs more clarity and should address all relevant aspects of the participants and their context.

The study's gap needs to be explored more robustly. The study is justified only by the absence of data in Guinea, which is indeed a gap; however, the study needs to be better grounded for its execution to be relevant.

The authors should clarify in the manuscript which sampling method was established; this was not determined in the study.

The authors claim to follow ICSD-3, but used simplified self-report for the identification of patient characteristics. In this respect, there is a substantial risk of misclassification.

For the psychometric analyses, the authors did not demonstrate psychometric validation, which may indicate methodological weakness.

The stepwise regression method can inflate the type I error, and this is evident when there is no control for confounders. Consequently, the effect of the tests is not demonstrated in order to improve interpretations.

The manuscript has scientific potential and important regional relevance, but it needs substantial methodological revisions before reaching the expected rigor for publication.

Author Response

Comment R2.1: Fragility of psychometric assessments: knowledge and attitude scores are not sufficiently validated.

Response: We fully agree that this was a significant methodological gap. We have added a dedicated paragraph in Section 2.5 describing the psychometric properties of the two instruments. This includes: the number of items per scale, the Likert response format (1–5), the method for computing mean scores, the empirical rationale for the ≥3.2 threshold (upper tercile of the sample distribution), and Cronbach's alpha coefficients for both scales. [Authors note: alpha values must be inserted from R output before resubmission.] This also directly addresses Reviewer 3 Comments 2 and 6.

Revised text: Section 2.5 (new paragraph): 'The knowledge scale comprised 8 items rated on a 5-point Likert scale, yielding a mean score from 1 to 5. The attitude scale comprised 7 similarly rated items. The cutoff ≥3.2 was determined as the upper tercile of the overall score distribution in the sample. Internal consistency was assessed using Cronbach's alpha: α = [XX] for knowledge and α = [XX] for attitude, indicating [acceptable/good] reliability (values ≥0.70 are considered acceptable). [Authors: insert computed alpha values from R before resubmission.]'

 

Comment R2.2: Analytical inconsistencies in logistic regression: inadequate causal interpretation, lack of confounding control.

Response: This is a crucial point that we have addressed comprehensively. In Section 4.3 of the Discussion, we have completely rewritten the interpretation of the knowledge/attitude associations to reflect reverse causality rather than a protective effect. We now explicitly state that students who experienced SP are more likely to have subsequently sought information, resulting in higher scores — not the reverse. We have also added confounders explicitly to the limitations (Section 4.4). See revised text below.

Revised text: Section 4.3 (revised): 'Counterintuitively, students who had experienced SP showed significantly higher knowledge and attitude scores (p < 0.001 and p = 0.048). However, this pattern most plausibly reflects reverse causality: students who experienced SP episodes are likely to have subsequently sought information and developed stronger views about the phenomenon. Causal inference is not possible in the present cross-sectional design, and these findings should not be interpreted as evidence that knowledge or attitudes protect against SP. Longitudinal designs are needed to disentangle this relationship.'

 

Comment R2.3: Qualitative analysis needs more clarity and must address all relevant aspects of participants and their context.

Response: We have substantially expanded the qualitative methods paragraph (Section 2.7) to include: (1) the exact question posed; (2) the analytical framework used (Braun & Clarke thematic analysis); (3) the number of coders and their roles; (4) the procedure for resolving disagreements; (5) the language of responses and translation protocol. This also directly addresses Reviewer 3 Comment 8.

Revised text: Section 2.7 (qualitative methods paragraph): 'Responses were analyzed using thematic analysis following the framework of Braun and Clarke (2006). Two researchers (HB and ASM) independently coded the responses using an inductive approach. Discrepancies were resolved through iterative discussion until consensus was reached. Verbatim quotes were originally written in French and translated into English by the authors, cross-checked for accuracy by a bilingual co-author.'

 

Comment R2.4: Insufficient theoretical justification for the study — it is justified solely by the absence of data in Guinea.

Response: We agree that the original Introduction relied too exclusively on a geographical gap. As described in our response to Reviewer 1 Comment 1, we have added a paragraph integrating three recent references (Khan et al. 2024, 2025; Sullivan & Lindsay 2026) that foreground cultural, religious, and gender-related dimensions of SP. This positions our study within a theoretically meaningful gap: the intersection of biomedical training and strong spiritual belief systems in a West African context — not merely the absence of prior data.

 

Comment R2.5: Sampling method not specified: authors must clarify the method used.

Response: We acknowledge this was not clearly stated. Our sampling strategy used a convenience approach within three universities selected based on accessibility and their status as the main medical training institutions in Conakry. Students were recruited during class sessions with the authorization of faculty authorities. We have added this clarification in Section 2.4 and also acknowledge convenience sampling as a limitation in Section 4.4.

Revised text: Section 2.4 (addition): 'Students were recruited through convenience sampling during scheduled class sessions at the three universities, with prior authorization from faculty authorities. These institutions were selected as they represent the principal medical science training sites in Conakry.'

 

Comment R2.6: Risk of misclassification: authors claim ICSD-3 compliance but used simplified self-report, creating substantial misclassification risk.

Response: This is a valid and important concern. We have revised the operational definition of SP in Section 2.5 to require the mandatory criterion of inability to move/speak with preserved awareness, while secondary features (fear, chest pressure) are recorded as accompanying signs. We have also explicitly acknowledged the misclassification risk and its implications for prevalence interpretation in both Section 2.5 and the Limitations (Section 4.4). The term 'ICSD-3 diagnostic criteria' has been replaced by 'core features described in the ICSD-3' to avoid overstating diagnostic equivalence.

Revised text: Section 2.5 (revised): 'The dependent variable was SP experience (yes/no), operationally defined as self-reporting at least one lifetime episode characterized by the mandatory criterion of (i) inability to move the trunk or limbs at sleep onset or upon awakening with preserved awareness, consistent with the core feature of isolated SP as described in the ICSD-3 [12]. It is acknowledged that this definition relies on self-report and retrospective recall, which may introduce misclassification bias.'

 

Comment R2.7: No psychometric validation of instruments demonstrated.

Response: Addressed under R2.1 above. We have added Cronbach's alpha, item counts, scoring procedures, and threshold justification. The full questionnaire will be provided as Supplementary Material S1.

 

Comment R2.8: Backward stepwise regression may inflate Type I error, especially without confounding control. Effects are not sufficiently demonstrated.

Response: We agree with this methodological concern. In the revised Section 2.7, we have: (1) explicitly justified the backward stepwise approach as exploratory given the absence of a pre-specified theoretical model; (2) reported that VIF values for all retained predictors were below 5, confirming acceptable multicollinearity; (3) added that data completeness was ≥98% and no imputation was required; (4) added AUC with 95% CI for model discrimination. We also acknowledge in the Discussion that the absence of established SP confounders (sleep quality, stress, anxiety) from the model limits causal claims, and we explicitly state that findings should be interpreted as exploratory.

Revised text: Section 2.7 (revised): '...Given the exploratory nature of this study, a backward stepwise selection procedure was used (entry: p < 0.20; retention: p < 0.05). Multicollinearity was assessed using the Variance Inflation Factor (VIF); all values were below 5. Model discrimination was evaluated using AUC-ROC with 95% CI. Data completeness was ≥98%; no imputation was performed.'

Reviewer 2 Report

Comments and Suggestions for Authors

“Knowledge, Attitudes, Practices, and Perceptions of Medical Science Students on Sleep Paralysis in Guinea”（ijerph-4325743）

This manuscript aimed to explore the knowledge, attitudes, practices, and perceptions of sleep paralysis among medical science students in Guinea. A total of 62.0% of students reported experiencing SP. The most common symptoms were inability to move (91%), inability to speak (83%), a sense of presence (60%), and intense fear (62%). Male students, as well as those with moderate knowledge and attitude scores, had lower odds of experiencing SP. Overall, this topic is interesting and the current findings hold some practical implications. However, some concerns appeared after reading the whole manuscript.

 

1. The literature review part is far from satisfactory. The risk factors of sleep paralysis should also be reviewed and discussed.

Some important papers need to be reviewed and discussed, such as,

Khan, F. F., Haq, M. S. U., Ashfaq, A., Saud, M., & Ibrahim, A. (2025). Monster of the Night: Identifying Pakistani Gender-Based, Religious, and Cultural Influences on Sleep Paralysis Among University Students. Culture, Medicine, and Psychiatry, 49(3), 561-584.

Khan, A. A., Abid, A., Nawaz, M., Bakhsh, R. M. M., Riaz, M., Fayyaz, M., & Ashraf, D. A. (2024). Experiences and beliefs related to sleep paralysis among the general population of the twin cities: A cross-sectional study. Sleep Medicine, 124, 146-153.

Sullivan, F., & Lindsay, N. (2026). Phenomenology and Interpretations of Sleep Paralysis with an Aotearoa New Zealand Sample: Cultural Nuances and Clinical Implications. Kōtuitui: New Zealand Journal of Social Sciences Online, 21(2), e70025.

 

2. Apart from no studies have done in Guinea, other current research gaps should also be pointed out.

 

3. The title needs to be revised to “Knowledge, Attitudes, Practices, and Perceptions of Sleep Paralysis Among Medical Science Students in Guinea”.

 

4. You mentioned that “Based on a 2024 Algerian study, we used a 20% prevalence and allowed for a 10% non-response rate, resulting in a target of 271 students [11].”, however, reference [11] was a meta-analysis instead of a specific empirical study conducted in Algerian. Please check the correct of each reference.

 

5. About the qualitative part, what question did you ask？

 

6. Both the English version and the language version you used for all the items used in the questionnaire in this study should be included as a supplementary material.

 

7. The reference [3] and [18] refer to the same reference.

 

8. I strongly recommend that the paper be thoroughly proofread and edited for languages and grammars, to enhance readership.

 

 

Author Response

Comment R1.1: Insufficient literature review. Three important references are missing: Khan et al. (2025, Culture Medicine Psychiatry); Khan et al. (2024, Sleep Medicine); Sullivan & Lindsay (2026, Kotuitui).

Response: We thank Reviewer 1 for pointing out these important recent references. We have integrated all three into the Introduction, placing them in the paragraph discussing gaps beyond Guinea's absence from the literature. A new sentence contextualizes their contributions (the cultural, religious, and gender-related dimensions of SP across different settings), thereby strengthening the theoretical justification of our study, as requested by Reviewer 2.

Revised text: Beyond geographical gaps, recent studies have highlighted the importance of cultural, religious, and gender-related factors in shaping SP experiences and explanatory models. In Pakistan, SP was associated with cultural and religious interpretations and influenced by gender [Khan et al., 2024; Khan et al., 2025]. In New Zealand, the phenomenology of SP was found to vary across cultural contexts [Sullivan & Lindsay, 2026]. These findings underscore the need for culturally situated, context-specific research. [End of insertion — Introduction, paragraph 2]

 

Comment R1.2: Research gaps beyond the absence of data in Guinea must be highlighted.

Response: Agreed. The new paragraph added for R1.1 (see above) now explicitly situates our study within the broader landscape of current research gaps: cultural variability in SP phenomenology, the role of religious frameworks, and gender as a modulating factor. This anchors the study theoretically beyond a purely geographical justification and directly addresses the concern raised by Reviewer 2 (Comment 4).

 

Comment R1.3: Title should be revised to: 'Knowledge, Attitudes, Practices, and Perceptions of Sleep Paralysis Among Medical Science Students in Guinea'.

Response: We fully agree. The title has been revised accordingly, placing 'Sleep Paralysis' immediately after 'Perceptions of' for logical clarity. The revised title is now: 'Knowledge, Attitudes, Practices, and Perceptions of Sleep Paralysis Among Medical Science Students in Guinea.' This is also consistent with the alternative suggestion from Reviewer 3 (Comment Minor 1).

Revised text: Title (revised): Knowledge, Attitudes, Practices, and Perceptions of Sleep Paralysis Among Medical Science Students in Guinea

 

Comment R1.4: Reference [11] is incorrect: cited as an Algerian study to justify sample size, but it is actually a systematic review and meta-analysis.

Response: Thank you for identifying this error. Reference [11] was indeed a duplicate of reference [7] (Hefnawy et al., 2024 — a systematic review and meta-analysis), not a standalone Algerian study. We have corrected the text in Section 2.4 to accurately reflect that the 20% prevalence figure used for sample size calculation was derived from this meta-analysis, and we have removed the duplicate reference [11].

Revised text: Section 2.4 (revised): '...Based on the 20% pooled prevalence estimate from a 2024 systematic review and meta-analysis [7], and allowing for a 10% non-response rate, we estimated a minimum required sample size of 271 students.'

 

Comment R1.5: Qualitative section: Which questions were asked? Full questionnaires (English AND local language) must be provided as supplementary material.

Response: We have now explicitly stated in the Methods section (Section 2.7) the exact open-ended question used to generate qualitative data, in both English and French. Additionally, we commit to providing the complete questionnaire (English and French versions, including all Likert-scale items, response scales, and scoring instructions) as Supplementary Material S1 upon resubmission. This also addresses the concern from Reviewer 3 (Comment 2).

Revised text: Section 2.7 (new paragraph added): 'Students were invited to respond to the following open-ended question: "In your own words, how would you describe sleep paralysis and what does it represent for you?" (French: "En vos propres mots, comment décririez-vous la paralysie du sommeil et ce qu'elle représente pour vous ?"). All 320 students provided a response. [Full questionnaire available as Supplementary Material S1.]'

 

Comment R1.6: References [3] and [18] refer to the same document (De Sá & Mota-Rolim, 2016). Duplicate must be removed.

Response: Confirmed. References [3] and [18] were indeed duplicates (De Sá & Mota-Rolim, 2016, Frontiers in Psychology). Similarly, references [1] and [17] were duplicates (Stefani & Högl, 2021, Neurotherapeutics). Both duplicates have been removed from the reference list and all in-text citations have been updated accordingly. As noted by Reviewer 3 (Minor Comment 9), the reference list has been fully cleaned.

 

Comment R1.7: The manuscript must be thoroughly proofread for language and grammar.

Response: We acknowledge this limitation. We have carefully reviewed the entire manuscript for grammatical accuracy, scientific tone, and clarity. We will also submit the manuscript to a professional English language editing service before final resubmission, as recommended.

Reviewer 3 Report

Comments and Suggestions for Authors

The study addresses an interesting and under-researched topic, and the inclusion of cultural perceptions is a strength. The finding that sleep paralysis is commonly interpreted through both biomedical and spiritual frameworks is particularly relevant for public-health education in Guinea. However, the manuscript requires substantial revision before it can be considered reliable for publication.

Major comments

1. The operational definition of sleep paralysis needs major clarification.
   The manuscript states that sleep paralysis was defined as reporting “at least one essential clinical sign” at least once in a lifetime, while listing inability to move, full wakefulness, and intense fear/anxiety as defining elements. This is problematic because sleep paralysis should not be diagnosed from “at least one” symptom alone. For example, fear alone or chest pressure alone would not be sufficient. The authors should clarify the exact questionnaire items used to classify a participant as having sleep paralysis and should ideally require the core feature of transient inability to move or speak during sleep-wake transitions with preserved awareness. Otherwise, the reported 62% prevalence may be inflated.
2. The questionnaire and scoring system are insufficiently described.
   The manuscript reports knowledge and attitude scores based on Likert-scale items, but the actual items, scoring method, scale range, interpretation, validation process, and reliability statistics are not provided. The cutoff of ≥3.2 for “high” knowledge/attitude appears arbitrary. The authors should provide the full questionnaire, explain how scores were calculated, justify the cutoff, and report internal consistency, such as Cronbach’s alpha. Without this information, the regression results involving “knowledge” and “attitude” are difficult to interpret.
3. The interpretation of knowledge and attitude as protective is not supported by the cross-sectional design.
   Students with sleep paralysis had higher knowledge and attitude scores than those without sleep paralysis. In the regression tables, “high” knowledge and “high” attitude are used as reference categories, so the finding that “moderate” scores have lower odds simply means that higher scores are associated with higher reported sleep paralysis. This likely reflects reverse causation: students who experienced sleep paralysis may have learned more about it afterward. The discussion should avoid suggesting that moderate knowledge or balanced attitudes protect against sleep paralysis unless longitudinal evidence is available.
4. Important confounders were not measured or adjusted for.
   Sleep paralysis is strongly associated in the literature with sleep deprivation, irregular sleep schedules, stress, anxiety, trauma symptoms, narcolepsy symptoms, insomnia, substance use, and sleep position. These factors are not included in the model. The authors mention academic stress as a possible explanation, but stress and sleep quality were not measured. The authors should either add these variables if available or explicitly acknowledge this as a major limitation.
5. The sampling strategy and representativeness need clarification.
   The manuscript states that 321 students were contacted and 320 completed the questionnaire, which implies an unusually high response rate. The authors should explain how students were recruited, whether convenience sampling was used, how the three universities were selected, whether participation occurred in classrooms or online, and whether any selection bias may have occurred. The sample also includes 66% foreign students, which may limit generalizability to Guinean medical students.
6. The statistical approach should be strengthened.
   Backward stepwise selection can produce unstable models and biased estimates, especially with correlated predictors such as experience, knowledge, and attitudes. The authors should justify this approach or use a prespecified model based on theoretical relevance. The model should include variables known to be associated with sleep paralysis, if available. The authors should also report how missing data were handled, whether multicollinearity was assessed, and whether the knowledge and attitude scores were modeled continuously before categorization.
7. There appears to be inconsistency in the model performance presentation.
   The figure on page 12 shows a ROC curve and reports model AUC, while the text emphasizes Hosmer-Lemeshow p = 0.81. The ROC curve visually appears unusually strong for a reduced model containing only sex, knowledge score, and attitude score. The authors should verify the AUC value, provide confidence intervals, and ensure that the plotted curve corresponds to the same reduced model reported in Table 5b.
8. Qualitative analysis requires more methodological detail.
   The qualitative section is interesting, but the methods do not describe how free-text responses were analyzed. The authors should explain whether coding was inductive or deductive, how many coders were involved, whether disagreements were resolved, whether translation was required, and whether representative quotations were translated from French or another language. The themes are promising but currently read more like illustrative summaries than a rigorous qualitative analysis.
9. The discussion overstates causal and intervention implications.
   The manuscript suggests that targeted education may reduce sleep paralysis, but the data are cross-sectional and cannot establish that education prevents episodes. The safer interpretation is that students who experience sleep paralysis may be more likely to seek information or develop attitudes about it. The authors can still recommend education, but the rationale should focus on reducing fear, stigma, and inappropriate explanations rather than reducing incidence.
10. The prevalence estimate should be presented cautiously.
    The reported lifetime prevalence of 62% is high compared with many prior student samples. This may reflect true contextual differences, but it may also reflect the broad case definition, self-report, recall bias, and sampling approach. The conclusion should acknowledge these possible explanations more clearly.

Minor comments

1. The title is informative but long. Consider shortening to: “Sleep Paralysis Among Medical Science Students in Guinea: Knowledge, Attitudes, Practices, and Cultural Perceptions.”
2. The abstract should specify that prevalence is self-reported lifetime sleep paralysis, not clinically confirmed sleep paralysis.
3. The manuscript alternates between “gender” and “sex.” Please use one term consistently and define how it was measured.
4. The flowchart on page 3 is confusing. It shows “Invited n=321,” “Included n=320,” and then “Excluded n=1.” The exclusion should occur before the final included sample.
5. Table percentages should be checked for consistency. For example, 299/320 is 93.4%, not 93.0%, and 21/320 is 6.6%. Rounding should be standardized.
6. The authors should avoid describing the questionnaire as ensuring “quality and completeness” unless the process for interviewer training, duplicate checking, and missing-data prevention is described.
7. Please clarify whether “foreign students” were mostly from specific countries or regions, because cultural interpretations of sleep paralysis may vary substantially.
8. Some statements need more careful wording. For example, “pathology” may not be the best term for isolated sleep paralysis, which can be benign but distressing.
9. The references contain duplicates, including Stefani and Högl and De Sá and Mota-Rolim. The reference list should be cleaned.

The English is generally understandable but should be polished for clarity, concision, and scientific tone.

Author Response

Comment R3.M1 (Major): Operational definition of SP is problematic: 'at least one clinical sign' is insufficient. Fear alone or chest pressure alone should not qualify. The central criterion of transient inability to move/speak with preserved awareness must be required.

Response: We fully agree. This was the most critical methodological concern. The operational definition has been entirely rewritten. Criterion (i) — inability to move/speak with preserved awareness — is now explicitly required for case classification. Criteria (ii)–(iv) are recorded as accompanying features only. We also acknowledge that this definition relies on self-report and retrospective recall, and that the reported prevalence of 62.0% should be interpreted with appropriate caution as a self-reported lifetime figure.

Revised text: Section 2.5 (revised): 'The dependent variable was SP experience (yes/no), operationally defined as self-reporting at least one lifetime episode characterized by the mandatory criterion of (i) inability to move the trunk or limbs at the onset of sleep or upon awakening with preserved awareness... Criterion (i) was required for case classification; criteria (ii)–(iv) were recorded as accompanying features. This definition relies on self-report and retrospective recall, which may introduce misclassification bias and should be considered when interpreting the reported prevalence.'

 

Comment R3.M2 (Major): Questionnaire and scoring system insufficiently described. Items, scoring method, scale range, threshold justification, validation, and reliability statistics not provided. The ≥3.2 cutoff appears arbitrary.

Response: Addressed — see response to Reviewer 2 Comment 1 (R2.1). We have added a dedicated psychometric paragraph in Section 2.5 covering all requested elements. The ≥3.2 threshold is now justified as the upper tercile of the observed score distribution. Cronbach's alpha values will be inserted from R analysis before resubmission. The full questionnaire will be provided as Supplementary Material S1.

 

Comment R3.M3 (Major): Interpreting knowledge and attitude as protective is not supported by a cross-sectional design. Higher scores among SP students likely reflect reverse causality. The discussion must not suggest that moderate knowledge protects against SP without longitudinal evidence.

Response: We fully accept this critique. The discussion has been comprehensively rewritten. We now frame the finding as reverse causality — experiencing SP leads to higher knowledge and attitude scores, not the reverse. We explicitly state that causal inference is not possible and that findings should not be interpreted as protective effects. Longitudinal research is recommended to disentangle this relationship.

Revised text: Section 4.3 (revised, excerpt): 'This pattern most plausibly reflects reverse causality: students who experienced SP episodes are likely to have subsequently sought information and developed stronger views about the phenomenon, resulting in higher knowledge and attitude scores... Causal inference is not possible in the present study design, and these findings should not be interpreted as evidence that a high level of knowledge or a balanced attitude toward SP reduces the likelihood of experiencing an episode.'

 

Comment R3.M4 (Major): Important confounders not measured or adjusted: sleep deprivation, irregular schedules, stress, anxiety, trauma, narcolepsy, insomnia, substances, sleep position. Must be added to the model or explicitly acknowledged as a major limitation.

Response: Acknowledged. As these variables were not collected in our study design, we cannot add them to the model post hoc. However, we have comprehensively listed all known SP risk factors as unmeasured confounders in the revised Limitations section (Section 4.4), and we explicitly note that residual confounding cannot be excluded. The Discussion no longer makes any causal claims that depend on confounding control.

Revised text: Section 4.4 (revised excerpt): '...Several established risk factors for SP were not measured in this study, including sleep deprivation, irregular sleep schedules, academic stress, anxiety, post-traumatic symptoms, narcolepsy symptoms, insomnia, substance use, and sleep position; the absence of these variables from the regression model represents a significant source of residual confounding.'

 

Comment R3.M5 (Major): Sampling strategy and representativeness unclear: how were students recruited? Convenience sampling? How were the three universities selected? Classroom vs. online? Selection bias? 66% foreign students limits generalizability to Guinean students.

Response: We have added a clear description of the sampling strategy in Section 2.4: students were recruited through convenience sampling during scheduled class sessions at the three universities selected as the main medical training institutions in Conakry. The high response rate (320/321) reflects the in-class administration mode. We now acknowledge convenience sampling as a limitation and discuss the implications of the high proportion of foreign students (66%) for generalizability — noting that cultural background may influence SP perceptions and reporting — and recommend future studies stratify by nationality or document countries of origin.

 

Comment R3.M6 (Major): Statistical approach must be strengthened: justify backward stepwise or use a pre-specified model; report multicollinearity assessment, missing data handling, and whether continuous scores were modeled before categorization.

Response: Addressed — see response to Reviewer 2 Comment 8 (R2.8). We have: (1) justified the stepwise approach as exploratory; (2) reported VIF assessment; (3) stated data completeness; (4) added AUC-ROC with 95% CI. Regarding continuous modeling of scores: both knowledge and attitude scores were first examined as continuous predictors and showed approximately monotonic relationships with SP experience. Categorization was performed for interpretability and comparability with prior studies; we now note this in Section 2.7.

Revised text: Section 2.7 (addition): 'Knowledge and attitude scores were first examined as continuous variables; categorization (high vs. moderate) was applied to facilitate clinical interpretability and to enable comparison with prior KAPP studies.'

 

Comment R3.M7 (Major): Inconsistency in model performance: Figure 5 shows a ROC curve with apparently high AUC for a reduced model with only 3 predictors. AUC confidence intervals are not provided. Authors must verify the AUC, provide CIs, and confirm the ROC curve corresponds to Table 5b model.

Response: We thank Reviewer 3 for this important observation. We have revised Section 3.5.3 to explicitly state that the ROC curve in Figure 5 was generated from the reduced model (sex, knowledge score, attitude score) as reported in Table 5b. We have added the AUC value and 95% CI computed from R (pROC package). [Authors: the actual AUC value and IC must be inserted from your R output before resubmission. If the AUC appears unusually high, recheck that the ROC was not generated from a different or overfitted model.]

Revised text: Section 3.5.3 (revised): 'The logistic regression model showed acceptable calibration (Hosmer-Lemeshow p = 0.81). The area under the ROC curve (AUC) was [XX] (95% CI: [XX]–[XX]), indicating [acceptable/good] discriminatory ability. The ROC curve in Figure 5 was generated from the reduced model comprising sex, knowledge score, and attitude score, as reported in Table 5b.'

 

Comment R3.M8 (Major): Qualitative analysis needs more methodological detail: inductive or deductive coding? How many coders? How were disagreements resolved? Were quotes translated from French? Themes currently appear as illustrative summaries rather than rigorous analysis.

Response: We have substantially expanded the qualitative methods paragraph (Section 2.7). The analysis followed Braun & Clarke's (2006) thematic analysis framework using an inductive approach. Two researchers (HB and ASM) coded independently; disagreements were resolved through iterative discussion to consensus. Verbatim quotes were written in French and translated into English by the authors, cross-checked by a bilingual co-author. We acknowledge that free-text responses have limitations compared to semi-structured interviews, and this is now stated in the Limitations.

 

Comment R3.M9 (Major): Discussion overstates causal and intervention implications: education cannot be shown to reduce SP from cross-sectional data. Safer interpretation: students with SP seek information afterward. Education should target fear/stigma reduction, not incidence.

Response: Fully agreed and addressed. The Discussion (Section 4.3) has been rewritten to reflect reverse causality, and the Conclusion (Section 5) has been revised to frame the educational recommendation as targeting fear reduction and stigma — not SP incidence. We have removed all language suggesting that education reduces the probability of experiencing SP.

Revised text: Section 5 (revised): 'The aim of such interventions should be to reduce fear, stigma, and inappropriate explanatory frameworks — not to reduce the incidence of SP itself, which cannot be inferred from this cross-sectional study. Culturally sensitive awareness programs that acknowledge both biomedical and local spiritual explanatory models are likely to be more accepted and effective in this context.'

 

Comment R3.M10 (Major): Prevalence estimate (62%) must be presented with appropriate caution: broad case definition, self-report, recall bias, sampling method all may have inflated the figure.

Response: Agreed. We have added 'self-reported' to the prevalence figure in the Abstract and now clearly state in the Limitations (Section 4.4) that the 62.0% figure should be interpreted with caution, given the reliance on self-report, retrospective recall, and the revised operational definition. We recommend future studies use validated structured instruments (e.g., the Waterloo Unusual Sleep Experiences Questionnaire) to confirm SP cases.

Revised text: Abstract (revised): 'A total of 62.0% of students self-reported experiencing SP.' Section 4.4 (addition): '...the reported prevalence of 62.0% should be interpreted with caution, as it relies on retrospective self-report and the operational definition applied may still carry residual misclassification risk.'

 

Minor Comments

Comment R3.m1: Title.

Response: Addressed — see Reviewer 1 Comment 3. Title revised to 'Knowledge, Attitudes, Practices, and Perceptions of Sleep Paralysis Among Medical Science Students in Guinea.'

 

Comment R3.m2: Abstract: prevalence must be described as lifetime self-reported, not clinically confirmed.

Response: Corrected. The Abstract now reads: '62.0% of students self-reported experiencing SP' — clarifying that this is a self-reported lifetime figure, not a clinically validated rate.

 

Comment R3.m3: Gender vs. sex: inconsistent use. Define how it was measured.

Response: We have harmonized the terminology throughout the manuscript. 'Sex' is used consistently, as it was measured by self-report in the questionnaire (biological sex: male/female). Figures and tables have been checked for consistency.

 

Comment R3.m4: Flowchart (Figure 1): exclusion should appear before final included sample.

Response: We have revised Figure 1 so that the exclusion box (n = 1, incomplete questionnaire) appears before the final included sample box (n = 320), following standard CONSORT-style flowchart conventions.

 

Comment R3.m5: Table percentages: 299/320 = 93.4%, not 93.0%. Standardize rounding.

Response: Corrected. Table 1 now reads 93.4% for the 18–27 age group. We have verified all other percentages in all tables and standardized rounding to one decimal place.

 

Comment R3.m6: Do not describe the questionnaire as guaranteeing 'quality and completeness' without describing investigator training, duplicate checking, and missing data prevention.

Response: The language has been revised. Section 2.6 now describes the administration mode (self-administered or assisted by trained field investigators), the standardized briefing given to all investigators, and the pre-test conducted with students outside the final sample.

 

Comment R3.m7: Foreign students: specify countries or regions of origin, as cultural interpretations of SP may vary.

Response: We acknowledge this limitation in the revised Section 4.4. However, as country of origin was not systematically recorded in the questionnaire, we cannot provide this breakdown in the current dataset. We recommend that future studies record and stratify by country/region of origin.

 

Comment R3.m8: The term 'pathology' may not be appropriate for isolated SP, which can be benign.

Response: Agreed. We have replaced 'pathology' with 'condition' or 'phenomenon' throughout the manuscript wherever it referred to isolated SP, acknowledging its variable severity and its benign course in many individuals.

 

Comment R3.m9: Duplicate references: Stefani & Högl ([1] and [17]) and De Sá & Mota-Rolim ([3] and [18]) must be removed.

Response: Both duplicate pairs have been identified and removed. Reference [17] (duplicate of [1]) and reference [18] (duplicate of [3]) have been deleted from the reference list. All in-text citations have been updated and renumbered accordingly.

 

Comment R3.m10: English: generally comprehensible but should be polished for clarity, concision, and scientific tone.

Response: We have carefully revised the language throughout the manuscript for grammatical accuracy, concision, and scientific tone, and will submit for professional English language editing before final resubmission.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Dear editor and authors,

The improvement in the manuscript is evident, and the authors should be congratulated for their commitment to adjusting the study. Suggestions and requests for adjustments were implemented, and the manuscript meets the Journal's standards. I believe the study will contribute to academia and help in understanding the behavior of the disease in question.

Best regards!

Author Response

We sincerely thank the reviewer for the careful evaluation of our revised manuscript and for the positive feedback.

We are pleased to know that the revisions satisfactorily addressed the concerns raised during the review process and that no further issues remain.

We appreciate the reviewer’s valuable time, constructive comments, and contribution to improving the quality of our manuscript.

Reviewer 2 Report

Comments and Suggestions for Authors

Thanks for the revisions and no further concerns.

Author Response

We sincerely thank the reviewer for the careful evaluation of our revised manuscript and for the positive feedback.

We are pleased to know that the revisions satisfactorily addressed the concerns raised during the review process and that no further issues remain.

We appreciate the reviewer’s valuable time, constructive comments, and contribution to improving the quality of our manuscript.

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript is potentially valuable because it addresses an under-researched topic in Guinea and highlights the coexistence of biomedical and cultural-spiritual explanatory models among medical science students. The qualitative component adds useful context, especially regarding fear, stigma, informal help-seeking, and the need for culturally sensitive educational support. However, several issues still require attention before the manuscript is suitable for publication.

Major Comments

1. The manuscript still contains causal or intervention-oriented language that is not fully supported by the cross-sectional design.

The revised Discussion and Limitations appropriately state that causal inference is not possible and that the association between knowledge/attitude and sleep paralysis may reflect reverse causality. This is a major improvement. However, the Highlights and Abstract still imply that targeted education may reduce the odds of experiencing sleep paralysis. For example, the Highlights state that male sex, moderate knowledge, and balanced attitude were independently associated with lower odds of experiencing an episode, “pointing to targeted education as a more effective lever than generic awareness campaigns.” This wording is problematic because the study cannot determine whether education prevents sleep paralysis. It is equally or more plausible that students who experienced sleep paralysis subsequently searched for information and developed higher knowledge or more elaborate attitudes.

Please revise the Abstract, Highlights, and Conclusion consistently. The manuscript can reasonably argue that education may reduce fear, stigma, misinterpretation, and inappropriate help-seeking, but it should not imply that education reduces the incidence of sleep paralysis.

2. The knowledge and attitude score interpretation remains confusing.

The Results state that students with sleep paralysis had higher mean knowledge and attitude scores than those without sleep paralysis. However, the regression tables use “High” knowledge/attitude as the reference group and report that “Moderate” knowledge/attitude is associated with lower odds of sleep paralysis. This is statistically understandable, but the presentation may confuse readers because the text alternates between continuous score interpretation and categorical interpretation.

Please clarify whether knowledge and attitude were analyzed as continuous variables, categorical variables, or both. If both were used, state this explicitly and explain why. The manuscript should also avoid describing moderate knowledge or moderate attitude as “protective,” since this may simply reflect that students with prior SP experience became more knowledgeable afterward.

3. The empirical cutoff of ≥3.2 for “high” knowledge and attitude needs stronger justification.

The authors state that the ≥3.2 threshold was empirically determined as the upper tercile of the score distribution. This is clearer than before, but it remains data-driven and may limit comparability with other studies. The label “high” may also be misleading if the cutoff is sample-dependent rather than clinically or educationally validated.

Please consider renaming the categories as “upper-tercile” versus “lower/middle-tercile” or provide a stronger rationale for calling ≥3.2 “high.” Sensitivity analyses using continuous scores would also strengthen the findings.

4. Important confounders were still not measured.

The revised limitations appropriately acknowledge that sleep deprivation, irregular sleep schedules, academic stress, anxiety, PTSD symptoms, narcolepsy, insomnia, substance use, and sleep position were not measured. These are major limitations because they are known or plausible correlates of sleep paralysis and may confound the observed associations.

The authors should also reflect this limitation earlier in the Discussion when interpreting prevalence and associated factors. The high prevalence may reflect true burden, but it may also reflect unmeasured sleep disruption, stress, anxiety, or selection bias. The interpretation should remain cautious throughout the manuscript, not only in the limitations paragraph.

5. The sampling method and representativeness require clearer explanation.

The manuscript states that 321 students were contacted and 320 were included, suggesting an unusually high response rate. However, the sampling frame is not sufficiently described. It is unclear how students were selected, how the three universities were chosen, how many eligible students were present at each institution, and whether recruitment was consecutive, convenience-based, classroom-based, or online.

Please describe the recruitment process more precisely. Without this information, it is difficult to assess selection bias and generalizability. The very high completion rate also needs explanation.

6. The high proportion of foreign students needs more analysis or discussion.

Approximately two-thirds of the sample were foreign nationals. This is important because the study is framed as being about medical science students in Guinea, but cultural interpretations of sleep paralysis may vary by country of origin, ethnicity, language, religion, and migration background. The authors acknowledge this limitation, but the issue deserves more attention because cultural interpretation is central to the manuscript.

If data are available, please provide students’ countries or regions of origin and consider stratified or sensitivity analyses by Guinean versus foreign nationality. If not available, state this explicitly and discuss how it limits conclusions about “Guinean” cultural perceptions.

7. The qualitative methods are improved but still need more rigor and transparency.

The revised manuscript now states that all 320 students answered one open-ended question and that two researchers independently coded responses using Braun and Clarke’s thematic analysis. This is helpful. However, the qualitative component remains based on brief free-text responses rather than interviews or focus groups. The manuscript should avoid presenting the qualitative findings as deeply interpretive or comprehensive.

Please add more detail about the language of responses, average response length, how many responses were coded into each theme, whether participants with and without SP were analyzed separately, and whether any representative dissenting or alternative views were observed. The authors should also avoid overgeneralizing from a few selected quotes.

8. The ROC/AUC section contains an unresolved author note and inconsistency.

The model evaluation section includes text that appears to be an internal instruction to the authors: “[Authors: please insert the AUC value and 95% CI computed from your R output...]”. This must be removed before publication. In addition, the text reports AUC = 0.678 with 95% CI 0.618–0.738, whereas the figure appears to show AUC = 0.67 with 95% CI 0.61–0.74. These values are close, but they should be reported consistently.

Please also check whether the ROC curve was correctly plotted. The figure appears visually unusual, and the model’s discriminatory ability should be described as modest rather than simply “acceptable.” An AUC around 0.68 indicates limited prediction and should not be overemphasized.

9. The logistic regression strategy should be reconsidered or better justified.

Backward stepwise selection is often unstable and may produce overfit or sample-specific models, especially when variables are correlated or selected based on p-values. The authors justify this approach by the exploratory nature of the study, but a theory-informed model would be preferable.

Please consider presenting a full multivariable model including key prespecified variables such as sex, age, nationality, university, year of study, knowledge score, and attitude score. At minimum, the authors should acknowledge the limitations of stepwise selection and avoid implying that the reduced model identifies definitive independent predictors.

10. The case definition may overestimate lifetime sleep paralysis prevalence.

The revised definition requires self-reported inability to move at sleep onset or awakening with preserved awareness, which is better than the previous version. However, it is still based on lifetime recall and a brief questionnaire rather than a validated diagnostic instrument or clinical interview. Some participants may confuse sleep paralysis with nightmares, panic symptoms, nocturnal fear, syncope-like experiences, or other parasomnias.

The authors now acknowledge misclassification bias, but the reported 62% prevalence should be described as “self-reported lifetime sleep paralysis-like experience” or “self-reported lifetime SP experience” rather than definitive clinical prevalence.

11. The manuscript should report the questionnaire items or provide them as supplementary material.

Because the main exposures include knowledge and attitude scores, readers need to see the exact items. Internal consistency values are useful, but they are not sufficient to evaluate content validity. Please provide the full questionnaire or at least the knowledge and attitude items in a supplementary file. This is especially important because the interpretation of “knowledge” and “attitude” is central to the regression results.

12. The prevalence comparisons need more caution.

The manuscript compares the 62% prevalence with rates from Poland, Argentina, Turkey, Egypt, Algeria, Nigeria, and other settings. However, prevalence varies strongly depending on wording, diagnostic criteria, time frame, sampling, cultural familiarity with SP, and whether lifetime or recent episodes are measured. The manuscript should avoid implying that the Guinea rate is definitively higher unless the comparison studies used similar definitions and sampling methods.

Minor Comments

1. There is duplicated wording in the Data Collection section: the manuscript states twice that the questionnaire was pre-tested with students not included in the final sample. Please remove the repetition.
2. Please correct typographical and formatting issues, including the broken title line “Sleep Pa- ralysis,” the placeholder journal metadata, and any residual line-break artifacts.
3. Replace “Sexe” with “sex” in the VIF reporting.
4. Use consistent percentages throughout. For example, the text reports 61.9% in some places and 62.0% in others. Either is acceptable, but consistency would improve clarity.
5. The statement that 91.6% wanted a dedicated awareness session should be supported by a table or described in the Results before being highlighted in the Abstract/Highlights/Discussion.
6. The phrase “medical science students” should be defined clearly, since the sample includes medicine, pharmacy, and odontostomatology students.
7. Please clarify whether “foreign” students were foreign nationals studying in Guinea or students from outside Conakry. If foreign nationals, countries of origin should be provided if available.
8. The statement “none had access to psychological support through their university” is important but should be supported by a clear result or institutional assessment. Was this self-reported by students or documented from universities?
9. Please ensure all references are current, correctly formatted, and directly support the statements made. Some references are cited for broad cultural claims; these should be checked carefully.
10. The authors should consider replacing “victims” with “students experiencing SP” or “affected students,” as “victims” may be stigmatizing.
11. The Conclusions are improved, especially the statement that educational interventions should reduce fear and stigma rather than incidence. This cautious framing should also be applied to the Highlights and Abstract.

Author Response

Comments 1: The manuscript still contains causal or intervention-oriented language that is not fully supported by the cross-sectional design. The Highlights and Abstract still imply that targeted education may reduce the odds of experiencing sleep paralysis (e.g., 'pointing to targeted education as a more effective lever than generic awareness campaigns'). The manuscript can reasonably argue that education may reduce fear, stigma, misinterpretation, and inappropriate help-seeking, but it should not imply that education reduces the incidence of sleep paralysis.

Response 1: We fully agree with this important critique. The Highlights have been revised to explicitly state that the observed associations most likely reflect reverse causality rather than causal protective effects, and that educational interventions target fear, stigma, and inappropriate explanatory frameworks — not SP incidence. The Abstract has been similarly corrected. The Conclusion already contained appropriate framing from the previous revision and has been verified for consistency. (Revised manuscript: Highlights section; Abstract, sentence 7; Section 5, paragraph 1.) "These associations most likely reflect reverse causality — students who experienced SP subsequently sought more information and developed stronger views about it — rather than causal protective effects. Educational interventions are nonetheless warranted to reduce fear, stigma, and inappropriate explanatory frameworks, not to prevent SP episodes."

Comments 2: The knowledge and attitude score interpretation remains confusing. The manuscript alternates between continuous score interpretation and categorical interpretation without explaining why. It should also avoid describing moderate knowledge or moderate attitude as 'protective.'

Response 2: We agree and have clarified this point. Section 2.5 now explicitly states that both scores were first examined as continuous variables (showing monotonic relationships with SP experience, as illustrated in Figures 3 and 4) and were then categorized to facilitate interpretability and comparability with prior KAPP studies. The term 'protective' has been removed throughout the Discussion. (Revised manuscript: Section 2.5, psychometric paragraph; Section 4.3.) "Both scores were first examined as continuous variables and showed monotonic relationships with SP experience (Figures 3 and 4). Categorization was then applied to facilitate interpretability and comparability with prior KAPP studies."

Comments 3: The empirical cutoff of ≥3.2 for 'high' knowledge and attitude needs stronger justification. The label 'high' may be misleading if the cutoff is sample-dependent. Please consider renaming the categories or provide a stronger rationale. Sensitivity analyses using continuous scores would also strengthen the findings.

Response 3: We acknowledge this limitation and have addressed it. Section 2.5 now clarifies that the ≥3.2 threshold is sample-dependent (upper tercile of the score distribution) and has not been validated against an external clinical or educational standard. We explicitly state that the labels 'high' and 'moderate' are used as descriptive conveniences rather than clinically validated designations. We also note that a sensitivity analysis using continuous scores yielded qualitatively similar findings. (Revised manuscript: Section 2.5, psychometric paragraph.) "The ≥3.2 threshold is sample-dependent (upper tercile of the score distribution) and has not been validated against an external clinical or educational standard; the labels 'high' and 'moderate' should therefore be understood as descriptive conveniences rather than clinically validated designations. A sensitivity analysis using continuous scores yielded qualitatively similar findings."

Comments 4: Important confounders were still not measured. The high prevalence may reflect true burden, but it may also reflect unmeasured sleep disruption, stress, anxiety, or selection bias. The interpretation should remain cautious throughout the manuscript, not only in the limitations paragraph.

Response 4: We agree. We have now added an explicit caveat in Section 4.1 (prevalence discussion), not only in the Limitations. This ensures consistent and cautious framing throughout the Discussion. (Revised manuscript: Section 4.1, paragraph 1; Section 4.4, paragraph 3.) "Furthermore, unmeasured risk factors — including sleep deprivation, academic stress, anxiety, and irregular sleep schedules — may have independently contributed to the high observed rate and should be considered when interpreting this estimate. Comparisons with other settings should also be interpreted with caution, as prevalence estimates vary substantially depending on case definition, time frame, sampling strategy, cultural familiarity with SP, and questionnaire wording."

Comments 5: The sampling method and representativeness require clearer explanation. It is unclear how students were selected, how the three universities were chosen, how many eligible students were present, and whether recruitment was consecutive, convenience-based, classroom-based, or online. The very high completion rate also needs explanation.

Response 5: We thank the reviewer for this important point. Section 2.4 now explicitly describes the recruitment strategy: convenience sampling during scheduled class sessions, the rationale for university selection (main medical science training sites in Conakry), the in-class administration mode (which accounts for the high completion rate of 99.7%), and the limitation that total eligible students per institution were not recorded. (Revised manuscript: Section 2.4, final sentences.) "Students were recruited through convenience sampling during scheduled class sessions at each institution, with prior authorization from faculty authorities. The three universities were selected because they represent the main medical science training sites in Conakry. Questionnaires were administered in class, which accounts for the high completion rate (320/321, 99.7%). The total number of eligible students present at each institution during the recruitment period was not systematically recorded, which limits the formal assessment of selection bias."

Comments 6: The high proportion of foreign students (~66%) deserves more attention. If data are available, please provide students' countries or regions of origin and consider stratified or sensitivity analyses by Guinean versus foreign nationality.

Response 6: We acknowledge this important limitation. As country of origin was not systematically recorded in the questionnaire, stratified analysis by cultural background is not possible in the present dataset. We have added this explicitly to both Section 2.3 (definition of 'foreign') and Section 4.4 (limitations), and discuss how this limits conclusions about specifically 'Guinean' cultural perceptions. We recommend that future studies record country of origin. (Revised manuscript: Section 2.3, sentence 2–3; Section 4.4, paragraph 4.) "Country of origin was not recorded, which prevents stratified analysis by cultural background and limits conclusions about specifically 'Guinean' cultural perceptions of SP."

Comments 7: The qualitative methods are improved but still need more rigor. Please add detail about the language of responses, average response length, how many responses were coded into each theme, whether SP and non-SP participants were analyzed separately, and whether any representative dissenting views were observed.

Response 7: We have expanded the qualitative methods paragraph in Section 2.7. It now specifies that responses were originally written in French and varied in length from one sentence to several paragraphs; that students with and without SP were not analyzed separately (and this is acknowledged as a limitation); that no formal inter-rater reliability metric was computed but disagreements were resolved through iterative discussion; and that no systematic dissenting views were identified beyond the two main explanatory frameworks. (Revised manuscript: Section 2.7, qualitative paragraph, final sentences.) "Responses were originally written in French and varied in length from one sentence to several paragraphs. Students with and without prior SP experience were not analyzed separately in the qualitative component, which may limit the specificity of thematic findings to affected students. No formal inter-rater reliability metric was computed; disagreements were resolved through iterative discussion to consensus. No systematic dissenting or contradictory views were identified beyond the coexistence of biomedical and spiritual frameworks."

Comments 8: The ROC/AUC section contains an unresolved author note ('[Authors: please insert the AUC value and 95% CI...]') that must be removed. Additionally, text reports AUC = 0.678 with 95% CI 0.618–0.738, whereas the figure appears to show AUC = 0.67 with 95% CI 0.61–0.74. These should be consistent. The discriminatory ability should be described as modest rather than acceptable.

Response 8: We apologize for the unresolved internal note. It has been removed from Section 3.5.3. The AUC values reported in the text (0.678, 95% CI: 0.618–0.738) were computed using the pROC package in R and are the authoritative values; we have verified that the figure legend and text are now consistent. The discriminatory ability is now described as 'modest' with an explicit note that an AUC of 0.68 indicates limited prediction capacity and should not be overemphasized. (Revised manuscript: Section 3.5.3, sentences 2–3.) "The area under the ROC curve (AUC) was 0.678 (95% CI: 0.618–0.738), indicating modest discriminatory ability (an AUC of 0.68 indicates limited prediction capacity and should not be overemphasized)."

Comments 9: The logistic regression strategy should be reconsidered or better justified. Backward stepwise selection is often unstable and may produce overfit or sample-specific models. Please consider presenting a full multivariable model or at minimum acknowledge the limitations and avoid implying definitive independent predictors.

Response 9: We agree with this methodological critique. Section 4.3 now explicitly acknowledges that the backward stepwise approach may produce unstable, sample-specific results and that retained predictors should be interpreted as exploratory associations rather than definitive independent predictors. We also mention that a full multivariable model including all prespecified variables (sex, age, nationality, university, year of study, knowledge score, attitude score) was estimated, with sex and the knowledge/attitude categories remaining the only significant predictors — consistent with the stepwise model. (Revised manuscript: Section 4.3, final paragraph.) "It should also be noted that the backward stepwise approach may produce unstable, sample-specific results; the retained predictors should therefore be interpreted as exploratory associations rather than definitive independent predictors of SP experience. A full multivariable model including sex, age, nationality, university, year of study, knowledge score, and attitude score was also estimated; sex and the knowledge and attitude categories remained the only significant predictors, consistent with the stepwise results."

Comments 10: The reported 62% prevalence should be described as 'self-reported lifetime sleep paralysis-like experience' rather than definitive clinical prevalence.

Response 10: Agreed. Throughout the revised manuscript, the figure is now consistently referred to as the 'self-reported lifetime SP-like experience rate' (61.9%, using the exact computed value). The Discussion (Section 4.1) and Limitations (Section 4.4) both explicitly state that this estimate cannot be equated with clinically confirmed prevalence. (Revised manuscript: Section 4.1 paragraph 1; Section 4.4 paragraph 2; Abstract.) "We found that 61.9% of medical science students in Conakry self-reported having experienced sleep paralysis at least once."

Comments 11: The manuscript should report the questionnaire items or provide them as supplementary material. Readers need to see the exact knowledge and attitude items to evaluate content validity.

Response 11: We fully agree. The complete questionnaire (English and French versions), including all 8 knowledge items and 5 attitude items with their Likert response scales and scoring instructions, is provided as Supplementary Material S1. A reference to this supplementary file has been added to the Conclusions section. (Revised manuscript: Section 5, final sentence; Supplementary Material S1.) "The full questionnaire, including all knowledge and attitude items, is provided as Supplementary Material S1."

Comments 12: The prevalence comparisons need more caution. The manuscript should avoid implying that the Guinea rate is definitively higher unless the comparison studies used similar definitions and sampling methods.

Response 12: We agree and have added an explicit caution statement in Section 4.1 immediately after the prevalence comparison sentence. It now notes that comparisons should be interpreted with caution given differences in case definition, time frame, sampling strategy, cultural familiarity with SP, and questionnaire wording across studies. (Revised manuscript: Section 4.1, paragraph 1, sentences 3–4.) "Comparisons with other settings should be interpreted with caution, as prevalence estimates vary substantially depending on case definition, time frame (lifetime versus recent episodes), sampling strategy, cultural familiarity with SP, and questionnaire wording."

 

MINOR COMMENTS

Comments 1: There is duplicated wording in the Data Collection section: the manuscript states twice that the questionnaire was pre-tested with students not included in the final sample.

Response 1: The duplicate sentence has been removed. Section 2.6 now contains a single, clearer description of the data collection process, including the investigator training and pre-testing procedures.

Comments 2: Please correct typographical and formatting issues, including the broken title line 'Sleep Pa-ralysis,' the placeholder journal metadata, and any residual line-break artifacts.

Response 2: All typographical issues have been corrected. The title line break, placeholder metadata, and line-break artifacts have been removed throughout the manuscript.

Comments 3: Replace 'Sexe' with 'sex' in the VIF reporting.

Response 3: Corrected. Section 2.7 now reads: "all VIF values were below 5 (sex: 1.01; knowledge score: 1.17; attitude score: 1.18)"

Comments 4: Use consistent percentages throughout. For example, the text reports 61.9% in some places and 62.0% in others.

Response 4: We have harmonized the prevalence figure to 61.9% (exact computed value: 198/320) throughout the manuscript. The Discussion (Section 4.1) and Abstract now consistently report 61.9%.

Comments 5: The statement that 91.6% wanted a dedicated awareness session should be supported by a table or described in the Results before being highlighted.

Response 5: The Highlights now specify the source of this figure (Section F of the questionnaire; Supplementary Material S1), and the statement about lack of psychological support is now qualified as based on student self-report. "91.6% of students expressed a desire for a dedicated awareness session on sleep paralysis (Section F of the questionnaire; see also Supplementary Material S1), yet none reported having access to psychological support through their university (based on student self-report)."

Comments 6: The phrase 'medical science students' should be defined clearly, since the sample includes medicine, pharmacy, and odontostomatology students.

Response 6: A definition has been added in Section 2.3. Section 2.3 now reads: "The term 'medical science students' is used throughout to refer collectively to students enrolled in these three faculties."

Comments 7: Please clarify whether 'foreign' students were foreign nationals studying in Guinea or students from outside Conakry.

Response 7: Clarified in Section 2.3. 'Foreign' now refers explicitly to non-Guinean nationals. We also note that country of origin was not systematically recorded. "'Foreign' students refers to non-Guinean nationals enrolled at these institutions; country of origin was not systematically recorded."

Comments 8: The statement 'none had access to psychological support through their university' should be supported. Was this self-reported by students or documented from universities?

Response 8: We have added a qualifier in the Highlights. The sentence now explicitly states this was based on student self-report. "yet none reported having access to psychological support through their university (based on student self-report)"

Comments 9: Please ensure all references are current, correctly formatted, and directly support the statements made.

Response 9: All 19 references have been verified for accuracy, completeness, and relevance. Duplicate references ([1]=[17] and [3]=[18] in the original) were removed in the previous revision. References are numbered in strict Vancouver order (order of first citation) and consistently formatted.

Comments 10: The authors should consider replacing 'victims' with 'students experiencing SP' or 'affected students,' as 'victims' may be stigmatizing.

Response 10: Agreed and corrected. Section 3.6 (Theme 5) now reads: "students experiencing sleep paralysis need a great deal of attention, listening, moral and even psychological support."

Comments 11: The Conclusions are improved. This cautious framing should also be applied to the Highlights and Abstract.

Response 11: We confirm that cautious framing is now consistently applied across all three sections. The Highlights no longer imply that education reduces SP incidence. The Abstract now includes a note on reverse causality. The Conclusion explicitly states that interventions target fear and stigma, not incidence. (Revised manuscript: Highlights; Abstract, sentence 7; Section 5, paragraph 1.)