Comments on: A new screening of preterm birth in gestation with short cervix after pessary plus progesterone

Dear Editor,

I have read with extreme interest the following publication by França et al.,(¹) published in Revista Brasileira de Ginecologia e Obstetrícia in 2024.

Quite surprisingly, I found that data used seems to have been already published in a previous paper of 2022.(²) Beside providing an explanation for the use of published material, the authors should also clarify the following methodological flaws that may prompt to question their results and conclusions:

1. The article reuses the exact same dataset from the P5 Trial (ReBec UTN: U1111-1164-2636), conducted in 17 centers between 2015 and 2019, yet presents the analysis as if it constituted new evidence. This is misleading and obscures the fact that the study is not independent.

2. The reported predictive performance (AUC 0.978 for preterm birth <34 weeks) is implausible. Such near-perfect discrimination is unprecedented in obstetric prediction models and strongly suggests severe overfitting, manipulation, or fabrication of data.

3. The so-called validation is invalid. Splitting the same randomized trial cohort into progesterone+pessary and progesterone-only groups does not constitute independent validation but instead recycles the same dataset to artificially inflate performance metrics.

4. Analytic thresholds (false-positive rates of 10% and 20%) were chosen retrospectively and optimized after viewing the data. With 31 variables examined without adjustment for multiplicity, the analysis reeks of post hoc cherry-picking and data dredging.

5. The results are irreconcilable with the original findings of the P5 Trial (Obstetrics and Gynecology, 2022).(²) That RCT showed no significant difference in adverse neonatal outcomes (19.2% vs 20.9%, adjusted RR 0.88; 95% CI 0.69–1.12) and only a modest reduction in preterm birth <34 weeks (9.9% vs 13.9%, adjusted RR 0.66; 95% CI 0.47–0.93). Deriving an AUC of 0.978 from the same dataset is inconsistent with these modest results and raises serious questions about authenticity.

6. The absence of protocol pre-registration for this secondary analysis, lack of external validation, and failure to share the dataset or analysis code prevent verification of the extraordinary claims. Extraordinary results require extraordinary evidence, yet this article provides only opacity and implausible findings.

7. Taken together, these issues go beyond methodological weakness. They raise the possibility of questionable research practices and even scientific misconduct, calling into question the integrity of the article and its conclusions.

References

• ¹ França MS, de Andrade VL Jr, Hatanaka AR, Santos R, Carvalho FH, Costa ML, et al. A new screening of preterm birth in gestation with short cervix after pessary plus progesterone. Rev Bras Ginecol Obstet. 2024;46:e-rbgo39i. doi: 10.61622/rbgo/2024rbgo39i
  » https://doi.org/10.61622/rbgo/2024rbgo39i
• ² Pacagnella RC, Silva TV, Cecatti JG, Passini R Jr, Fanton TF, Borovac-Pinheiro A, et al. Pessary plus progesterone to prevent preterm birth in women with short cervixes: a randomized controlled trial. Obstet Gynecol. 2022;139(1):41-51. doi: 10.1097/AOG.0000000000004634
  » https://doi.org/10.1097/AOG.0000000000004634

Repply to: A new screening of preterm birth in gestation with short cervix after pessary plus progesterone

AuthorshipSCIMAGO INSTITUTIONS RANKINGS

Dear Editor,

We appreciate Dr. Unfer's interest in our publication and the opportunity to clarify the points raised in his letter. Critical discussion is fundamental to scientific progress, especially in complex areas such as preterm birth prediction. However, some of the concerns appear to arise from a misinterpretation of our study.

Our article does not present a new randomized clinical trial nor provide definitive proof of efficacy. It is an exploratory post hoc analysis of a previously published trial, designed to test a predictive model. Its purpose was to generate hypotheses and identify women at higher risk of prematurity, even when receiving preventive treatment. From introduction to conclusion, we explicitly state that our findings may assist in risk identification but still require prospective validation before any clinical use.(¹)

It is well established that a considerable proportion of women experience preterm birth despite prophylactic measures such as progesterone, pessary, or cerclage. These refractory cases illustrate the limitations of current preventive therapies and the need for complementary approaches to identify, early in pregnancy, those most vulnerable to adverse outcomes. Recognizing this subgroup is clinically important: it enables closer monitoring, guides counseling and shared decision-making, and defines priorities for future research targeting women at highest risk.

This motivation underlies the analysis presented in "A new screening of preterm birth in gestation with short cervix after pessary plus progesterone".(¹) The study used data from the P5 Trial,(²) which compared pessary + progesterone versus progesterone alone in women with short cervix. As a secondary post hoc analysis, it sought not to test intervention efficacy but to explore predictors of poor prognosis. This nature exempts the analysis from preregistration. The article clearly states that it is a secondary analysis with a distinct aim, and there was no attempt to conceal this.

The original P5 Trial showed a modest reduction in preterm birth <34 weeks and no significant difference in composite neonatal outcomes. Our analysis does not contradict these results. The high AUC reported does not reflect superior clinical efficacy of pessary use but rather model fitting within a homogeneous population of women with short cervix, in which categorical variable vectorization likely enhanced discrimination. Such performance, while interesting, must be interpreted cautiously because, as highlighted by Dr. Unfer, it may reflect overfitting.

The study, however, adds incremental information to the field. It reinforces that prematurity is a multifactorial condition that cannot be predicted by cervical length alone. Combining demographic, obstetric, and morphological variables improved risk discrimination. The model identified previous preterm birth, early detection of short cervix (<19 weeks), very short cervix (≤14.7 mm), and white ethnicity as main risk factors. Protective factors included no prior curettage, singleton gestation, and curved cervical length >21 mm, findings consistent with established biological plausibility regarding cervical integrity, obstetric history, and anatomic configuration.

These parameters merit consideration in future screening strategies so that more vulnerable women can be recognized early and followed closely. Nevertheless, the algorithm described should not be used universally. Predictive variables identified in one cohort may not hold in others, since cervical length distribution, ethnicity, and environmental influences vary among populations.

Validation in our study constitutes derived internal validation using two approaches: comparison with cervical length <15 mm and application of model coefficients to the progesterone-only group. This derived internal validation is not equivalent to independent external validation. The drop in model performance in the progesterone-only group demonstrates that accuracy decreases in broader or more heterogeneous samples. Precisely why extrapolation to clinical practice is premature and could lead to bias or misinterpretation. Prospective multicenter validation across diverse settings is essential to confirm reproducibility and ensure that apparent performance is not driven by homogeneity of the original cohort.

Regarding methodology, we used fixed false-positive rates (FPR 10% and 20%), following established precedents in obstetric screening.(³-⁵) This approach allows balanced comparison of detection rates and reflects international standards for predictive modeling. Fixed FPRs are common in prenatal screening because they balance identification of high-risk patients with avoidance of excessive false alarms that may prompt unnecessary procedures.

On the matter of data and code sharing, we acknowledge this as a limitation of many papers, but this does not constitute misconduct. We support initiatives such as TRIPOD-AI(⁶) that promote responsible transparency and reproducibility and agree that progress will depend on institutional and editorial policies that encourage data sharing without compromising the ethical protection of patients or reinforcing a colonialist stance on scientific data.

It is important to reaffirm that post hoc analyses such as ours are not intended to guide immediate clinical decisions or to establish definitive evidence. Their purpose is to generate hypotheses and stimulate scientific discussion. While methodological criticisms about overfitting and limited validation are legitimate, allegations of data manipulation or fabrication are unfounded. Our analysis represents a statistical exploration with preliminary findings, not a re-evaluation of treatment efficacy. The results should therefore be viewed as hypothesis-generating, pending confirmation in prospectively designed independent cohorts.

In conclusion, the model proposed is not intended for immediate clinical implementation, but it reinforces the multifactorial nature of the condition and demonstrates that multivariate modeling may help identify high-risk women even among those receiving prophylaxis. The appropriate next step is to externally validate the model, conduct prospective, multicenter studies including more diverse populations and determine whether recognition of this subgroup enables tailored interventions capable of reducing preterm birth. Only after such rigorous evaluation should broader clinical use be considered.

References

• ¹ França MS, de Andrade VL Jr, Hatanaka AR, Santos R, Carvalho FH, Costa ML, et al. A new screening of preterm birth in gestation with short cervix after pessary plus progesterone. Rev Bras Ginecol Obstet. 2024;46:e-rbgo39i. doi: 10.61622/rbgo/2024rbgo39i
  » https://doi.org/10.61622/rbgo/2024rbgo39i
• ² Pacagnella RC, Silva TV, Cecatti JG, Passini R Jr, Fanton TF, Borovac-Pinheiro A, et al. Pessary plus progesterone to prevent preterm birth in women with short cervixes: a randomized controlled trial. Obstet Gynecol. 2022;139(1):41-51. doi: 10.1097/AOG.0000000000004634
  » https://doi.org/10.1097/AOG.0000000000004634
• ³ Mercer BM, Goldenberg RL, Das A, Moawad AH, Iams JD, Meis PJ, et al. The preterm prediction study: a clinical risk assessment system. Am J Obstet Gynecol. 1996;174(6):1885-93. doi: 10.1016/S0002-9378(96)70225-9
  » https://doi.org/10.1016/S0002-9378(96)70225-9
• ⁴ Chiu CP, Feng Q, Chaemsaithong P, Sahota DS, Lau YY, Yeung YK, et al. Prediction of spontaneous preterm birth and preterm prelabor rupture of membranes using maternal factors, obstetric history and biomarkers of placental function at 11–13 weeks. Ultrasound Obstet Gynecol. 2022;60(2):192-9. doi: 10.1002/uog.24917
  » https://doi.org/10.1002/uog.24917
• ⁵ Gudicha DW, Romero R, Kabiri D, Hernandez-Andrade E, Pacora P, Erez O, et al. Personalized assessment of cervical length improves prediction of spontaneous preterm birth: a standard and a percentile calculator. Am J Obstet Gynecol. 2021;224(3):288.e1-17. doi: 10.1016/j.ajog.2020.09.002
  » https://doi.org/10.1016/j.ajog.2020.09.002
• ⁶ Collins GS, Moons KG, Dhiman P, Riley RD, Beam AL, Van Calster B, et al. TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods. BMJ. 2024;385:e078378. doi: 10.1136/bmj-2023-078378
  » https://doi.org/10.1136/bmj-2023-078378

Publication Dates

• Publication in this collection
  05 Dec 2025
• Date of issue
  2025

History

• Received
  15 Sept 2025
• Accepted
  08 Oct 2025