The Association Between Amniocentesis and Adverse Pregnancy Outcomes in Pregnancies with Normal/Reportable Test Results: An Indication-Based Comparison with Non-Invasive Prenatal Testing

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript is methodologically generally well designed, with particular attention given to the analysis of composite outcomes and to the attempt to disentangle the risk attributable to the procedure itself from the baseline obstetric risk. The authors have also aimed to overcome methodological limitations observed in previously published studies that used NIPT-based control groups. Nevertheless, the manuscript could be improved in the following aspects:

Introduction – Repetitive elements regarding the rationale for selecting the control group should be reduced. In addition, the authors should clearly specify whether the primary focus of the study is adverse pregnancy outcomes or a broader spectrum of neonatal and/or perinatal outcomes.

Methodology – The criteria for “risk matching” should be explained in greater detail. It remains unclear whether this refers merely to the inclusion of groups with similar clinical indications without formal statistical matching. The application of propensity score matching should be considered. Furthermore, the authors should clarify whether a power analysis was performed and whether balance diagnostics were conducted. It is essential to report whether the groups were adequately balanced with respect to key variables. The multivariable model should be expanded beyond maternal age and parity to include additional relevant confounders (e.g., abnormal serum markers, positive personal obstetric history, and other clinically relevant risk factors).

Results – In the analysis of the correlation between gestational age at the time of the procedure and gestational age at delivery, a negative correlation (r = −0.174) was reported. Although statistically significant, this correlation is clinically weak. The authors attribute potential pathophysiological significance to this finding without additional supporting analyses. Moreover, some of the conclusions appear to imply causality, which is not appropriate for a retrospective study design. A graphical presentation of gestational age distribution should be considered. The Discussion section should be streamlined and focused on the key findings.

Author Response

We sincerely thank the reviewer for the positive evaluation of our study. We appreciate the reviewer’s constructive feedback and have carefully revised the manuscript accordingly to improve methodological transparency, clarify the study focus, and refine the analytical approach. Detailed responses to each specific comment are provided below.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Below is my review of the manuscript “Evaluating procedure-related risks of amniocentesis: using non-invasive prenatal testing as a risk-matched comparator.”
Overall I think the paper should go to majr revision, because with this analytic model there’s a big risk of wrong/inflated conclusions. My comments:

1. The authors claim a “risk-matched control group”, but in practice I see just a comparison of two cohorts with similar indications, without any real description of matching (matching/propensity etc) and without showing covariate balance after the “matching”. On top of that, the regression adjustment is limited to maternal age and parity only, which is far from enough. Why? Because:

• the choice of amniocentesis vs NIPT is strongly driven by the screening risk level (continuous risk estimate, MoM of markers, NT, ultrasound markers etc),

• and first-trimester markers themselves (PAPP-A, free β-hCG) are independently linked to PTB and other adverse outcomes (authors cite older stuff like FASTER, but they don’t include these variables in the model).
  So with the current approach the paper shows association, not “procedure-related risk”. If the model stays like this, the language really needs to be toned down (it’s not causal).

2. The authors exclude NIPT cases with fetal fraction <2.8%. This is a huge risk selector, since low FF is associated with worse outcomes (e.g. FGR/PIH in some studies). As a result, the NIPT group is kind of “cleaned” from higher-risk pregnancies, which can artificially enlarge differences against the amnio group.

3. In Methods they say they excluded women “who gave birth in another hospital”, but at the same time for miscarriage/IUD they analyse the full cohorts (n=7668 and n=2044). This could be logical (two datasets: fetal loss in the whole database + perinatal outcomes only for births in the centre), but right now it’s written in a way that reads like a contradiction.

4. In the loss outcomes table (IUD 17/7668 vs 18/2044) the crude OR should be like a few, not ~40 in univariable and 14.8 after adjustment. I suspect they used a different denominator (maybe only deliveries in the centre?) or there is some coding/table error.

5. In Table 5: OR 10.68; CI 7.50–15.21, but p=0.030. That looks plainly wrong — with this OR/CI the p-value should be extremely tiny (<<0.001). So there’s likely an error in calculations or how p-values were exported.

6. For the regression models, given the rare outcomes, they should:

• consider Firth penalized logistic regression / exact methods,

• report absolute risks and risk differences, not only ORs,

• and present a clear variable-selection plan (why only age+parity?).

6. (second point, but keeping it) The correlation r = -0.174 (p=0.001) within the amnio group is weak and very vulnerable to confounding (e.g. earlier amnio in more “problematic” pregnancies or with other risk factors). In Discussion it’s interpreted as a possible procedure effect — with this design that’s too strong.

7. In the Introduction, in the paragraph about first-trimester markers being linked to adverse outcomes, please mention PTB-focused papers, e.g. doi:10.1038/s41598-024-67300-6 (PMID: 39003389; PMCID: PMC11246412).

8. Formatting issue: title page says “Diagnostics 2025”, but the next pages have the header “Medicina 2025”.

9. The definition miscarriage <24 weeks and IUD ≥24 weeks is locally ok, but they should add justification (viability threshold / national definitions) and keep terminology consistent (miscarriage vs stillbirth vs IUFD).

In general, for me the paper is about association, not truly “procedure-related risks”. Either they need to redesign the modelling/approach, or they need to change the language and assumptions of the manuscript.

Author Response

We sincerely thank the reviewer for the careful and critical evaluation of our manuscript. In response to these important comments, we have substantially revised the manuscript to improve methodological transparency, refine the analytical approach, and avoid causal interpretations. Specifically, we have clarified the study design, revised the terminology regarding “risk matching,” expanded the multivariable models using all available clinically relevant confounders, corrected statistical inconsistencies, and strengthened the limitations section to explicitly acknowledge residual confounding. The manuscript language has also been carefully revised to avoid implying causality and to emphasize that the findings represent associations within an indication-based comparative framework.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I agree with the revisions

Author Response

Answer: We sincerely thank the reviewer for the positive evaluation of our revisions and for the constructive comments provided during the review process. We appreciate the reviewer’s time and valuable feedback, which helped improve the clarity and quality of the manuscript.

Reviewer 2 Report

Comments and Suggestions for Authors

I don’t see a point-to-point response to the review.

Still, I’ll comment on the revised manuscript.

The revision is more metodologically complete (they clarified the outcome definitions and limited the perinatal analysis to cases with delivery data), but it still has several critical issues that undermine both the internal consistency of reporting and the strength of any causal inference: (1) “risk-matching” is claimed but no actual matching is done, (2) high risk of selection bias because only “in-center” deliveries are analysed, (3) inconsistent numbers/denominators and table errors, (4) internal contradictions in the reported aORs (IUD 4.10 vs 14.8), and (5) substantial confounding by indication that is not adequately handled (adjustment only for maternal age and parity).

These are the same points I raised before, and they were not fixed.

1. In the Abstract and throughout the narrative the authors use “risk-matched control group design”, but in Methods it is clear that comparability is mostly based on clinical indication level, without any matching algorithm (propensity score / exact matching / weighting) and without any balance diagnostics. This needs to be consistent: either perform matching (and report it properly), or remove/replace the wording and describe it as an “indication-based comparison”.

2. Perinatal/neonatal outcomes are assessed only for 2063/7668 (26.9%) in the NIPT group and 377/2044 (18.4%) in the amniocentesis group, because only those deliveries occured “in-center”. This can meaningfully distort the results, and the direction of bias is not predictable (e.g., higher-risk patients may be more likely to deliver in a referral center). They mention this in limitations, but it needs hard analyses. Minimum additions:

• a table comparing baseline characteristics (age, parity, indication) between “delivered at our center” vs “delivered elsewhere”, separately for NIPT and amnio groups;

• consider inverse probability weighting (IPW) for the probability of “in-center delivery”, or at least a sensitivity analysis.

3. Miscarriage/IUD outcomes are calculated in the full cohort (n=7668 vs 2044), but parts of the tables and text use headers as if they refer to the delivery subset (2063 vs 377), which creates interpretational mess. In addition, Table 5 footnote reports n=7668/2044, although the table also includes LBW/SGA/APGAR, which according to Methods are available only for “in-center” deliveries.

4. In Results/Table 5, IUD is reported as aOR 4.10 (2.05–8.20), while in Discussion it appears as aOR 14.8 (10.5–47.6). This looks like an editing error or a swapped table version. There must be one consistent value across Abstract, Results, Discussion, and Table 5.

5. Please add a statement in the Introduction that first-trimester screening parameters have been linked not only to aneuploidy risk but also to downstream obstetric and neonatal complications, with citation: https://doi.org/10.3390/jcm13071982.

6. In Table 5 some numbers and CIs are “merged/stuck together” (e.g., in the IUD and LBW rows, univariate column). This is a technical issue, but for peer-review it’s a red flag, because it makes it hard to verify the analysis.

7. Table 1 reports maternal age as 32±6.5, but also shows ≥35 years = 1548 (75%) in the NIPT group (and 77% in amnio). That is statistically almost impossible. This suggests a data extraction or reporting error that must be corrected.

8. The regression models adjust only for age and parity. With indications like “screening risk” and “US markers”, this is clearly insufficient—because the severity of screening risk and the type of markers (and also BMI, smoking, ART, chronic diseases, prior PTB history, etc.) may affect PTB/LBW/SGA independent of the procedure. The authors themselves note the lack of quantitative risk metrics (combined risk, MoM) in Limitations. That is the core of confounding by indication and should be stated more strongly, and the conclusions should be toned down (“associated with”, not “procedure leads to”).

9. The correlation between GA at amniocentesis and GA at delivery is negative (r=-0.174), yet the text interprets it as “earlier amniocentesis might increase the risk of earlier delivery”. That does not follow from a negative r (negative r would rather suggest: later amnio ↔ earlier delivery). This interpretation needs to be corrected.

Overall, I still think the study has publication potential, but the authors have revised the manuscript in an insufficient way, and the major methodological/reporting issues remain.

Author Response

Please find the attachment.

Author Response File: Author Response.pdf

Round 3

Reviewer 2 Report

Comments and Suggestions for Authors

I still have a few comments on the manuscript:

1. In the Methods, the authors state that they excluded, among others, “cases who gave birth in another hospital”, but a few sentences later they say that miscarriage and IUD were analyzed in the whole cohort (7668 vs 2044), regardless of place of delivery, while only the perinatal/neonatal outcomes were restricted to “in-center” births (2063 vs 377). This is logically inconsistent. The Methods section should clearly describe two analytical populations, not just one list of exclusion criteria.

2. In the flowchart for the NIPT group, there are 17 stillbirths, 21 miscarriages, 5510 births at another hospital, 17 maternal additional disease, and 26 missing data, yet the final number is 2063. If these are separate, sequential exclusions as shown in the diagram, the math gives 2077, not 2063.

3. In the flowchart, the starting boxes are labeled “NIPT with normal results” and “Amniocentesis with normal results”. This limits generalizability and changes the interpretation of the comparison. The title and abstract should clearly state that the analysis concerns pregnancies with normal/reportable results, not all women referred for diagnostic testing.

4. In the Methods, the authors write that “Newborns with APGAR scores below 7 at both the 1st and 5th minutes were considered to have low APGAR scores”, while in the Results/Tables they analyze 1st-minute APGAR <7 and 5th-minute APGAR <7 separately, and the composite includes either one of these components. So the current definition in the Methods is incorrect, or at least confusing.

5. Table 5 is still methodologically unclear. There is a separate row “No **” next to “Miscarriage” and “Intrauterine fetal demise”. For a standard logistic regression, such a “No” row does not really make presentational sense and looks as if a multicategory analysis was being mixed with a binary one.

6. In the Introduction, the authors say that this design “ensured a strong comparison ... while controlling for potential confounding factors”, and in the strengths/limitations section they write that the project was “minimizing potential confounding and enhancing the reliability of the findings”. This is too strong, considering that they themselves acknowledge missing data on severity of screening risk, biochemical markers, smoking, ART, chronic diseases, etc., and the models are adjusted only for age and parity.

Overall, these are rather smaller issues to correct now.

Author Response

Dear Editor,

We sincerely thank you and the reviewers for the careful evaluation of our manuscript and for the constructive comments that helped us improve the quality and clarity of our study. In this revised version, we carefully addressed all reviewer comments and revised the manuscript accordingly. We clarified the study design and terminology, corrected inconsistencies in tables and numerical reporting, improved the description of the study population and outcome definitions, and strengthened the discussion and limitations regarding potential confounding and selection bias. All changes have been detailed in the point-by-point response document and incorporated into the revised manuscript.

We appreciate the time and effort invested by the reviewers and hope that the revised manuscript will now be suitable for publication.

Sincerely,

Prof. Burak Bayraktar, M.D.

Bayraktar, Burak - Web of Science Core Collection

‪Burak Bayraktar - ‪Google Akademik

Burak Bayraktar (0000-0001-6233-4207)

 

Reviewer 2:

I still have a few comments on the manuscript:

1. In the Methods, the authors state that they excluded, among others, “cases who gave birth in another hospital”, but a few sentences later they say that miscarriage and IUD were analyzed in the whole cohort (7668 vs 2044), regardless of place of delivery, while only the perinatal/neonatal outcomes were restricted to “in-center” births (2063 vs 377). This is logically inconsistent. The Methods section should clearly describe two analytical populations, not just one list of exclusion criteria.

Answer: In the revised manuscript, we clarified that two analytical populations were used. The overall cohort consisted of all eligible pregnancies undergoing amniocentesis (n = 2044) or NIPT (n = 7668), and this population was used for the analysis of pregnancy loss outcomes (miscarriage and intrauterine fetal demise), which were captured through the institutional registry regardless of the place of delivery. To avoid misunderstanding, the sentence in the exclusion criteria referring to deliveries in another hospital has been revised, and the Methods section now clearly describes these two analytical populations.

In Materials and Methods:

Exclusion criteria included multiple pregnancies, fetal structural or genetic anomalies, suspected fetal infection, parental genetic abnormalities, maternal chronic disease, and cases with missing data.” Pregnancies delivered at other hospitals were not excluded from the overall cohort; however, perinatal and neonatal outcome analyses were restricted to cases with delivery data available at our center. Perinatal and neonatal outcomes were evaluated for 2063 pregnant women in the NIPT group and 377 pregnant women in the amniocentesis group whose birth data were available.

In Table 4 and 5 footnotes:

Miscarriage and intrauterine fetal demise analyses were performed in the full cohort (NIPT n = 7668; amniocentesis n = 2044).

1. In the flowchart for the NIPT group, there are 17 stillbirths, 21 miscarriages, 5510 births at another hospital, 17 maternal additional disease, and 26 missing data, yet the final number is 2063. If these are separate, sequential exclusions as shown in the diagram, the math gives 2077, not 2063.

Answer: We thank the reviewer for carefully checking the flowchart. In fact, the categories shown in the flowchart were derived from overlapping case classifications rather than strictly sequential exclusions. We have therefore revised the flowchart and its legend to clearly distinguish the overall cohort, pregnancy-loss outcomes assessed in the full cohort, and the subset with available in-center delivery data used for perinatal and neonatal analyses. This revision eliminates the apparent numerical inconsistency and better reflects the actual analytical structure of the study.

Figure 1. The flow chart of the study

1. In the flowchart, the starting boxes are labeled “NIPT with normal results” and “Amniocentesis with normal results”. This limits generalizability and changes the interpretation of the comparison. The title and abstract should clearly state that the analysis concerns pregnancies with normal/reportable results, not all women referred for diagnostic testing.

Answer: We thank the reviewer for this important observation. The reviewer is correct that the study population was restricted to pregnancies with normal/reportable test results after exclusion of structural or genetic abnormalities, and this should be more explicitly reflected in the manuscript. In response, we have revised the title/abstract wording and clarified in the Methods section that the analysis was limited to pregnancies with normal/reportable results and without structural or genetic anomalies. This clarification improves the precision of the study population definition and the interpretation of the findings.

In Title:

Association between amniocentesis and adverse pregnancy outcomes in pregnancies with normal/reportable test results: an indication-based comparison with non-invasive prenatal testing

In Abstract:

The analysis was restricted to pregnancies with normal/reportable test results and without structural or genetic anomalies.

In Materials and Methods:

… who underwent reportable/normal NIPT results..

1. In the Methods, the authors write that “Newborns with APGAR scores below 7 at both the 1st and 5th minutes were considered to have low APGAR scores”, while in the Results/Tables they analyze 1st-minute APGAR <7 and 5th-minute APGAR <7 separately, and the composite includes either one of these components. So the current definition in the Methods is incorrect, or at least confusing.

Answer: The Methods section has been revised for clarity. In the analyses we evaluated APGAR scores at the 1st and 5th minutes separately, and the composite neonatal outcome included an APGAR score <7 at either time point. The Methods section has now been corrected to reflect this definition.

In Materials and Methods:

APGAR scores were evaluated separately at the 1st and 5th minutes. Newborns with APGAR scores <7 at the 1st minute or at the 5th minute were considered to have low APGAR scores for the respective analysis.

1. Table 5 is still methodologically unclear. There is a separate row “No **” next to “Miscarriage” and “Intrauterine fetal demise”. For a standard logistic regression, such a “No” row does not really make presentational sense and looks as if a multicategory analysis was being mixed with a binary one.

Answer: The presentation of the “No” row in Table 5 was methodologically inappropriate for logistic regression analysis, as it represents the reference category. Accordingly, the “No” row has been removed and the table has been revised to present only the outcome categories analyzed in the logistic regression models. This modification improves the clarity and methodological consistency of the table.

1. In the Introduction, the authors say that this design “ensured a strong comparison ... while controlling for potential confounding factors”, and in the strengths/limitations section they write that the project was “minimizing potential confounding and enhancing the reliability of the findings”. This is too strong, considering that they themselves acknowledge missing data on severity of screening risk, biochemical markers, smoking, ART, chronic diseases, etc., and the models are adjusted only for age and parity.

Answer: Specifically, statements suggesting that confounding was minimized or controlled have been replaced with wording indicating that the study design allowed comparison between groups with similar clinical indications, while acknowledging that residual confounding cannot be excluded due to the retrospective design and the absence of detailed data on several potential confounders.

In Introduction:

“This design allowed a comparison of outcomes between women undergoing amniocentesis and those opting for NIPT within a cohort with similar clinical indications, partially addressing potential confounding factors.”

In Limitation:

This design allowed for a more balanced comparison between groups with similar clinical indications. However, residual confounding cannot be completely excluded due to the retrospective design and the absence of detailed data on several potential confounders.

 

Overall, these are rather smaller issues to correct now.

Answer: We thank the reviewer for the careful re-evaluation of our manuscript and for the positive assessment of the revised version. All comments have been carefully addressed, and the manuscript has been revised accordingly to further improve clarity and methodological transparency.

Author Response File: Author Response.pdf