The overlap between miscarriage and extreme preterm birth in a limited-resource setting on the Thailand-Myanmar border: a population cohort study

Background : No universal demarcation of gestational age distinguishes miscarriage and stillbirth or extreme preterm birth (exPTB). This study provides a synopsis of outcome between 22 to <28 weeks gestation from a low resource setting. Methods : A retrospective record review of a population on the Thailand-Myanmar border was conducted. Outcomes were classified as miscarriage, late expulsion of products between 22 to < 28 weeks gestation with evidence of non-viability (mostly ultrasound absent fetal heart beat) prior to 22 weeks; or exPTB (stillbirth/live born) between 22 to < 28 weeks gestation when the fetus was viable at ≥22 weeks. Termination of pregnancy and gestational trophoblastic disease were excluded. Results : From 1995-2015, 80.9% (50,046/ 61,829) of registered women had a known pregnancy outcome, of whom 99.8% (49,931) had a known gestational age. Delivery between 22 to <28 weeks gestation included 0.9% (472/49,931) of pregnancies after removing 18 cases (3.8%) who met an exclusion criteria. Most pregnancies had an ultrasound: 72.5% (n=329/454); 43.6% (n=197) were classified as miscarriage and 56.4% (n=257) exPTB. Individual record review of miscarriages estimated that fetal death had occurred at a median of 16 weeks, despite late expulsion between 22 to <28 weeks. With available data (n=252, 5 missing) the proportion of stillbirth was 47.6% (n=120), congenital abnormality 10.5% (24/228, 29 missing) and neonatal death was 98.5% (128/131, 1 missing). Introduction of ultrasound was associated with a 2-times higher odds of classification of outcome as exPTB rather than miscarriage. Conclusion : In this low resource setting few (<1%) pregnancy outcomes occurred in the 22 to <28 weeks gestational window; four in ten were miscarriage (late expulsion) and neonatal mortality approached 100%. In the scale-up to preventable newborns deaths (at least initially) greater benefits will be obtained by focusing on the viable newborns of ≥ 28 weeks gestation.

In this low resource setting few (<1%) pregnancy outcomes Conclusion : occurred in the 22 to <28 weeks gestational window; four in ten were miscarriage (late expulsion) and neonatal mortality approached 100%. In the scale-up to preventable newborns deaths (at least initially) greater benefits will be obtained by focusing on the viable newborns of ≥ 28 weeks gestation.

Introduction
To determine progress towards the Sustainable Developmental Goal (SDG) 3.2 "by 2030, to end preventable deaths of newborns and children under 5 years of age", standardized data collection is critical [1][2][3] . Without a universally accepted definition to distinguish fetal death from miscarriage (spontaneous abortion) versus stillbirth 4,5 (Supplementary File 1) and the wide variation in standard of care available for live born extreme preterm births (exPTB), individual patient care and assessment of SDG progress are affected 6 . Miscarriage remains an awkward condition to define due to variability between and even within countries particularly in relation to laws surrounding termination of pregnancy 7 . There are obvious legal, religious and cultural sensitivities to pregnancy termination which may inhibit large organizations from specifying an upper limit to define miscarriage 8 . The International Classification of Diseases, provides a graded definition of miscarriage (pregnancy loss before 22 completed weeks gestation); and stillbirth (expulsion of a fetus with no signs of life) classified as early fetal death (of a fetus weighing 500 g or more, or aged 22 weeks or more, or with a body length of 25 cm or more); late fetal death (of a fetus of 1000 g or more, or aged 28 weeks or more, or with a body length of 35 cm or more) 9 . These definitions overlap. For example, is a stillborn 1000 g fetus with accurately determined gestation born at 27 +2 weeks +days , a late fetal death because of the birth weight or an early fetal death because of the gestation? For organizations that utilize the WHO definition of stillbirth recommended for global comparison i.e.: "a baby born with no signs of life at or after 28 weeks gestation" 10 confusion arises with live newborns before 28 weeks. It is difficult to explain to health care workers that stillbirth starts at 28 weeks but live birth can start at <28 weeks.
On the Thailand-Myanmar border at Shoklo Malaria Research Unit (SMRU), the WHO definition of stillbirth based upon reaching 28 weeks gestation has been used for the past 32 years (Supplementary File 1). Based on this definition there has been a significant decline in stillbirth (28 to 14 per 1,000 live births) and neonatal mortality (49 to 11 per 1,000 livebirths) from 1993-1996 to 2008-2011 11,12 . Miscarriage rates have been stable over time at 10% (2257/23 118) from 1994 to 2013 13 . The outcomes of pregnancies of 28 weeks or more gestation in this area have been published previously and are not the focus of this analysis 12,14,15 .
One of the most significant factors in this border region has been the 3 decades long struggle with recurring loss of antimalarials to treat patients with multi-drug resistant P. falciparum malaria 16 . The Thailand-Myanmar border was one of the first places in the world to introduce artemisinin-based combination therapy for treatment of malaria in the general population 17 . The artemisinin derivatives were also used initially for treatment of malaria in the 2 nd and 3 rd trimesters of pregnancy or to save the life of the mother with severe malaria in any trimester 18 . This class of drug distinguished itself in pregnancy as it was first reported by Chinese scientists to cause fetal resorption (or early miscarriage) 19 so it was important to be able to determine gestational age during pregnancy and to have a clear definition of trimester and miscarriage. Quinine remained the WHO recommended drug for women in the first trimester of pregnancy and systematic data collection by local health workers has built an evidence base that has made a significant contribution to global knowledge about antimalarial drug use in pregnancy particularly in the first trimester 13,20 .
In the 1990's at SMRU, gestational age during pregnancy was estimated from fundal height by a formula developed for the study population 21,22 or by last menstrual period (LMP), and from late 2001 by ultrasound 23 . In 1995 the only tools available in the field were pregnancy tests, symphysis fundal height measurement (SFH), pinnard and at best (late 1990's) a small hand-held Doppler to detect fetal heart beat. SFH measurements were routinely carried out at antenatal care and quality control measurement exercises were conducted because it was so heavily relied upon. SFH was examined at every visit until the fundus of the uterus could be felt for the first time and thereafter approximately every 2-4 weeks unless more frequent repeated measures were indicated. In these circumstances and in the absence of bleeding or fetal movements, detection of loss of fetal viability was identified late. When ultrasound was introduced in the area, a first trimester or first ANC visit scan was followed up by a repeat scan initially at 18 weeks and later shifted to 22 weeks. Gestational age of a non-viable fetus could be more readily obtained by ultrasound but for SMRU data to be consistent with the former data, the end of gestation of pregnancy has remained constant, as the date of expulsion of the products of conception.
At the same time, work has been standardized for local health workers by use of obstetric guidelines 24 . In the guidelines the working definition of miscarriage is outcome of pregnancy before 28 weeks, while births (stillbirth and live birth) commence from 28 weeks, as does reporting of neonatal deaths. This definition has remained unchanged with the establishment of a special care baby unit in this resource limited setting where assisted ventilation of newborns is not available (i.e birth before 28 weeks is not viable) 12 .
The objective of this study is to bring clarity to the nature of the products of conception that are expulsed in the 22 to <28 weeks gestational age window in a low resource setting, highlighting operational issues and the role of ultrasound 9 .  Figure 1).

Setting
At SMRU, place and attendance at birth has shifted from 75% occurring at home with traditional birth attendants with no formal training in 1986, to more than 80% of births occurring in health facilities with skilled attendants in 2015 15 . SMRU is staffed predominantly by locally trained workers for antenatal care and ultrasound 23 , child birth 15,24 and emergencies in adults and neonates 12 . Local medics, midwives and nurses do the majority of the clinical work and expatriate doctors assist local staff with 24-hour back up. Ultrasound quality has been measured previously in this setting 23 and the eight sonographers undergo a small quality control every 6 months by a clinician (5 scans, blinded to gestation, to check for image quality and intra and inter-observer error) 24 . Over 3,000 women register at SMRU antenatal clinics annually and these were well established before birth services were offered. The first birthing unit was opened at Shoklo refugee camp in 1986, with border skirmishes and closure of Shoklo, this was relocated to Maela Refugee camp in 1995, and two more units for marginalized migrant workers were opened in Wang Pha in Dec-2007 and in Maw Ker Thai in April 2010 ( Figure 1).
The seven signal functions for Basic Emergency Obstetric and Newborn Care, including parenteral administration of an oxytocic, antibiotics and anticonvulsants, removal of retained products of conception, assisted vaginal birth including breech birth, resuscitation of the newborn using a bag and mask for infants ≥ 28 weeks gestation, and screened blood transfusions, are provided by local staff. A description of the special care baby unit for neonates has been detailed by Turner et al. 12 , but there is no capacity for intubation and assisted ventilation and prohibitive costs limit newborn referrals. Aminophylline is used in place of caffeine for apneas, again due to costs. Local protocols  For example, if a woman reported absence of fetal movement, or bleeding, or the fundal height did not increase, an ultrasound could be done to determine viability. Measurement of the fetus size at each scan was encouraged. Loss of fetal heart beat could also be an incidental finding when the woman attended for her second scan. Loss of viability with ultrasound could be confirmed by presence of a fetus and absent fetal heartbeat. In some cases, ultrasound confirmed pregnancies persisting to 22 weeks gestation, but a fetus was never observed, e.g. anovulatory gestation or non-classic gestational trophoblastic disease.
In this setting SFH measurement has been important to determine whether a woman was in the first or second trimester which made the difference between being able to use quinine or artesunate to treat uncomplicated malaria. At antenatal care, pregnant women had abdominal palpation at each visit until the SFH was first palpable and measurable above the symphysis, then SFH was checked monthly up to 32 weeks and weekly from 36 weeks gestation. In case of doubt, urine (sometimes serum) pregnancy testing was available. It was and still is common to have the SFH measured multiple times in a single pregnancy 21,26 . Before ultrasound availability, loss of fetal viability could be confirmed by SFH measurements, which were increasing and then levelled off, or decreased with unexpected SFH results being confirmed by a clinician), complaints of loss of fetal movement or never feeling movement, bleeding episodes or expulsion of products. Quality control exercises of SFH involved comparing 20 women per month between SFH measurers -a difference of > 1cm was considered inacceptable and corrections of the firmness of applying the tape measure, correctly placing the end of the tape measure on the upper border of the symphysis pubis and identifying the fundus, were made during these sessions.
Management of fetal loss diverged from high income settings. Before misoprostol was available, induction of pregnancies with confirmed fetal loss was difficult and if there was no vaginal bleeding and the cervix was not open a conservative management style was adopted. Induction with syntocinon, the only available agent for many years, was frequently a prolonged and unsuccessful process. In this low resource context, only surgical emergencies were referred to tertiary hospitals, so non-viable pregnancy loss, with no imminent danger signs did not qualify for referral.

Data extraction and data definitions
All birth records at SMRU are computer based and paperbased records are archived at SMRU head office in Mae Sot. The original paper records initially matched data collection used by Médecins Sans Frontières (MSF) in humanitarian settings. In 1998 computer based registration commenced with adjustments to the MSF form, in late 2001 ultrasound commenced requiring further amendments, and in late 2007 all data was captured in real time in an application developed by Technology Sans Frontières (TSF) with data limitations set to alert users to inaccurate entry. Verification of paper and computer based data has taken place over decades to a computer based record system.
All pregnancy records from 1995 until 2015 in the window period from 22 to <28 weeks gestation were selected and reviewed case by case by for the present study. The starting point of 1995 was selected, since the first local guideline for obstetrics was introduced at this time.
For each record the following evaluation was conducted using a step-wise query process: a) Was there evidence of in utero fetal demise (no FHB) before 22 weeks? If yes, could the gestation of loss be estimated? For example, was fetal anthropometry measured by ultrasound when absence of fetal heart beat was confirmed; or did the SFH measurements stall or decrease (and at how many centimeters) before fetal heart beat could be heard or fetal movements felt, or did the mother report that she never felt fetal movements? Was there evidence that there was never a fetus i.e. that the pregnancy was never viable? Did ultrasound measure only annovulatory pregnancy (blighted ovum)? Records with evidence of in utero fetal demise before 22 weeks but expulsion between 22 to <28 weeks were classified as miscarriage (late expulsion). Pregnancies with evidence of fetal viability at ≥22 weeks that were expulsed between 22 to <28 weeks were classified as exPTB (live or stillborn). Neonatal mortality was defined as death in the first 28 days of a live born neonate of 22 to <28 weeks gestation. Gestational trophoblastic disease and termination of pregnancy were excluded. Congenital abnormalities were coded using the ICD-10 criteria 9 .

Statistical analysis
Gestation was reported by week, for example 22 weeks included women from 22 +0 to 22 +6 weeks, (i.e 22 weeks plus 6 days) of pregnancy. Continuous normally distributed data, such as gestation and birth weight, were described using the mean, and standard deviation (SD) and compared with the Student's t-test. Only the first born twin birth weight was retrieved from the electronic files. Non-parametric data, such as gravidity, were described using median and 25 th -75 th percentiles and compared with the Mann-Whitney U test. Proportions were compared using the Chi-squared test. To assess the role of ultrasound in the final classification as exPTB rather than miscarriage between 22 to <28 weeks gestation, univariable and multivariable logistic regression was used to determine the association between ultrasound use and outcome, adjusted for first ANC attendance in first trimester and delivering with a skilled attendant (confounders identified a priori). Data was analysed using SPSS version 20 (IBM SPSS, Armonk, NY, USA) and Stata version 13 (StatCorp, College Station, TX, USA).

Ethics statement
Ethical approval for retrospective analysis of pregnancy records was given by the Oxford Tropical Research Ethics Committee (OXTREC 28-09) and after discussion with the local Community Advisory Board (TCAB-4/1/2015)

Results
Between 1995 and 2015, 80.9% (50,046/61,829) of women registered to antenatal care had a known pregnancy outcome and were included in the present study. Only a small proportion, 0.2% (115/50,046), of these pregnancies could not be assigned a reliable gestational age ( Figure 2). The proportion of all pregnancy outcomes within the gestational window of 22 to <28 weeks was small: 0.9% (472/49,931) and most of these had an obstetric ultrasound scan and dating: 73.1% (345/472).
There were 3.8% (18/372) excluded from analysis: termination of pregnancy involved 13 cases including: six ultrasound confirmed major fetal abnormality (five anencephalic, one holoprosencephaly), two life-threatening maternal conditions both with uncontrollable severe pre-eclampsia, and five self-induced (one of whom was recently widowed); and five gestational trophoblastic disease. The demographic characteristics of the remaining 454 pregnancy outcomes are summarized in Table 1.
The majority of these pregnancies were dated by ultrasound: 72.5% (329/454). The numbers and proportions of pregnancy outcome for each gestational age week from 22 to <28 are shown in Table 2. There were 6.2% (28/454) twin pregnancies.

Miscarriage
Of the 454 pregnancy outcomes from 22 to <28 weeks gestation 197 (43.5%) were miscarriage (loss of viability before 22 weeks) two of which were twin gestations (23 weeks) ( Loss of viability was not possible to determine for 11 records (nine indicated spontaneous miscarriage and two records could not be located) leaving 186 records. Evidence of fetal demise (absence of fetal heart beat) also known as fetal death in utero, occurred in 60.2% (112/186), with most of these determined by ultrasound (Table 3). The proportion with stalling or decreases

Survival of Newborns
Missing data NND Most twins pregnancies were extreme PTB except for 2 which were miscarriage at 23 weeks b Born alive and probably died but data not available in the record c not sure exact days NND (only 700g and born at home, brought to clinic) d One died at day 33; one was still alive at 40 months of age in SFH was higher before ultrasound was introduced (Table 3). Annovulatory pregnancy (blighted ovum) was also a reason for expulsion of products of pregnancy from 22 to < 28 weeks of pregnancy in this setting: 16.8% (19/113) when ultrasound was available ( Amongst the 257 pregnancies ending in exPTB, 1.9% (5/257) had missing data on whether the infant was still-or live born (Table 2), and for the remaining cases, 47.6% (120/252) were recorded as stillbirths (including first born twin), and 52.4% (132/252) were born alive. Most women birthed vaginally, 98.0% (253/257), with 66.5% (171/257) in SMRU clinic. There were 1.6% (4/252) delivered by caesarean sections. These four cases were in singleton pregnancies, three at 27 weeks and one at 26 weeks, three of whom had placental pathologies (two placenta praevia and one placental abruption) and one with preterm labour and transverse presentation. These four births all ended in stillbirth with a birth weight available for one case (900g).
Congenital abnormality involved 10.6% (24/227, 30 missing) of exPTB and half of these congenital abnormality cases 50.0% (12/24) were stillborn (Table 4).  (Table 2). One newborn died at day 33 and the remaining child survived and was still alive when last seen at 40 months of age, with a normal neurodevelopment. The surviving female child was born at 27 +5 weeks, and two ultrasounds, including an early scan at 8 weeks and a later scan at 18 weeks, assured the gestation. The mother was a 32-year-old refugee with a gravidity of two and parity of one, with no history of PTB, who went into spontaneous labour and received a single dose of nifedipine and dexamethasone less than one hour before delivery. Delivery was supervised by skilled birth attendants and after a normal vaginal birth of a 890g baby, the Apgar scores were six and seven, at one and five minutes, with no resuscitation necessary. The neonate was provided with supportive care (with oxygen delivered by nasal prongs; temperature control, phototherapy, breast milk) because that was all that was available, and discharged home after ten weeks at 1061g.  (Table 5).

Discussion
Cautious interpretation of this data is required given the limitations of this dataset because 19% of women who registered to antenatal care had no pregnancy outcome reported. In a recent Lancet report of Myanmar Demographic Health Surveillance in 2016, less than 30% of rural women in Myanmar delivered in recognized institutions 27 . In the context of this postconflict, cross border, low resource setting, the fact that 81% of outcomes were reported, with ultrasound available for nearly ¾, supports the use of the data to bring clarity to the nature of the products of conception expulsed in the 22 to <28 weeks gestational age window. As well the median number of six antenatal consultations (Table 1) is high for pregnancies ending before 28 weeks gestation in a limited-resource setting.
The window of gestation from 22 to <28 weeks involved less than 1% of all pregnancy outcomes and 4 in 10 outcomes were miscarriage (i.e. non-viable before reaching 22 weeks). Ultrasound was associated with a 2 times higher odds of the outcome being classified as an exPTB rather than a miscarriage. This suggests that ultrasound adds clarity to the outcome of pregnancy in this 22 to <28 week window, mostly because it detects the presence/absence of a fetal heart beat or in this setting even the presence of annovulatory pregnancy -both of which are not easily identified with more limited tools (pregnancy test, SFH and pinnard/hand held Doppler). Under the WHO/ICD 10 definition these miscarriages could be (wrongly) defined as exPTB based on gestation alone, which would falsely inflate stillbirth rates 4 (Supplementary File 1).
Amongst the outcomes classified as exPTB, nearly one-half were stillborn, 98.5% of the babies born alive were neonatal deaths of which two-thirds occurred on day one, and there was a high proportion (as expected) of congenital abnormalities. A high congenital abnormality rate is expected amongst this age group in a setting that does not actively screen for abnormalities 28 . Two infants emerged from the neonatal period, one died at day 33 and one female of 27+5 weeks gestation survived infancy (end of the first year). Not only are the outcomes of pregnancies of 22 to <28 weeks gestation in this setting dismal, the proportion of pregnancies that are involved, relative to all pregnancies with a known outcome, is small (0.9%). It could be argued that if more than palliative care was offered to the newborns at SMRU, outcomes may have been different but with no means to provide additional support such as intubation, incubators, surfactant, parenteral nutrition etc, improvements would remain marginal. The high proportion of stillbirths (47.6% (120/252) contrasts with rates under 2% previously reported in this population in the 28 to <34 week' gestation window 12 . Stillbirth rates may be inflated using gestational age rather than birthweight cut-offs 29 and also because nearly all these infants were born vaginally, including breech births, whereas caesarean section may have been offered in HIC. These different proportions are important because they indicate that in low resource setting, the maximum benefits of interventions to prevent newborn death, is in the group of 28 weeks (and above) gestation 30 . Overall, this data supports the WHO definition of 28 weeks to define birth (live birth and stillbirth) 31 , and <28 weeks as a pragmatic definition of miscarriage 32 In low resource settings there are many reasons why women may have a late outcome (22 to <28 weeks gestation) of a miscarriage (non-viable before 22 weeks) (Table 3). Coming in to the clinic for an induction when there is no pain or bleeding (no obvious problem that is felt by the woman herself) may be perceived as being of greater consequence than not being able to plant or harvest crops, or not being able to receive daily wages which the family depend upon. Information on fetal movement was rarely volunteered so there may be cultural reasons for apparent tolerance to loss of fetal viability such as desire for fetus papyraceus, which, surprisingly, is culturally fortunate 33 .
Registration of births is incomplete in many LIC 34 and dating pregnancies reliably a particularly challenging issue 35 . Some of the higher end birth weights may be due to inaccurate dating but meta-analysis 14 also suggests 1000g as a cut-off misses gestations in the exPTB window. On the Thailand-Myanmar border, SMRU has integrated basic ultrasound delivered by local sonographers to routine ANC and this has been accepted by women 36 . Efforts have been directed towards having local staff skilled in routine gestational age scanning 23 , standard care at antenatal clinics and at birth 15 , and in newborn care 12 . Record keeping has been based on the cut-off point of 28 weeks gestation for birth and miscarriage and while that has been a strength in directing human resources in the delivery room and in special care baby unit to viable neonates it has also resulted in weaker reporting of pregnancy outcomes from 22 to <28 weeks, which is an obvious limitation of this data set. Another limitation of the analysis is the 21 year period of the cohort. This could however be viewed differently as it puts into perspective the small group of pregnancies involved: approximately six live born exPTBs per year compared to >2,000 births of 28 weeks or more per year. While there have been changes over time including the introduction of a special care baby unit and ultrasound, there has been no change in the assisted ventilatory support of newborns, which is not available.
In a low resource setting the outcome of pregnancy between 22 to <28 weeks gestation involves <1% of all outcomes, a high proportion of miscarriage (late expulsion), with one-in-two of the exPTBs being stillborn and amongst livebirths, a neonatal mortality approaching 100%. The distinction between miscarriage and exPTB in this gestational window is improved by ultrasound but this is unlikely to result in improvement in survival due to significant resource constraints. The WHO cut-point to define stillbirth and miscarriage is pragmatic and useful in low resource settings in the scale-up towards reducing preventable newborn deaths as it allows a greater focus on newborns more likely to survive with a gestation of 28 weeks or more.

Data availability
Due to ethical and security considerations, the data that supports the findings in this study can be accessed only through the

Halit Pinar
Warren Alpert Medical School, Brown University, Providence, RI, USA I reviewed the 3rd revision of the manuscript titled "Miscarriage, stillbirth and neonatal mortality in the extreme preterm birth window of gestation in a limited-resource setting on the Thailand-Myanmar border: a population cohort study'. I think the authors have addressed the concerns to the maximum of their capability within the confines of the study design and data set and I approve for publication.
No competing interests were disclosed. Competing Interests: The latest version of the paper "The overlap between miscarriage and extreme preterm birth in a limited-resource setting on the Thailand-Myanmar border; a population cohort study " version 3 is a significant improvement on previous 2 versions of this paper.

I have read this submission. I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.
However, I think it is worth mentioning that although it is sensible and logical to report as SBs those fetuses that are at least 28 weeks or 1000g, for international comparisons of PMRs, it is still worth gathering data for all fetuses (including SBs that are between 500-999g and/or between 22-28 weeks gestation, -even though this paper is not about PMRs per se.)

Is the study design appropriate and is the work technically sound? Yes
Are sufficient details of methods and analysis provided to allow replication by others?

Results
The outcome of 19% of booked pregnancies was not known. This is an important problem as pregnancies losses may be over-represented in these unknown outcome pregnancies.
The stillbirth rate is very high especially at the lower gestations. For example, over 40% of births at up 26 weeks were assessed to be stillbirths. There is a frequent of limited assessment for live birth at the lowest gestations. It is common to assess that a newborn is stillbirth because resuscitation is not implemented when few signs of live birth are present. Signs of live birth are not reported. Resuscitation efforts are not reported.
It is concerning that birth weight was not available for 42% of the births? Other basic data also were frequently not available. This raises concerns about the quality of the data.
Gestational age may be underestimated as one SD is above 1000 g which is unlikely for a population of 22-27 week pregnancies.
Data on provision of care are not provided well. It is unclear to what extent live births were provided full support (other than the therapies stated not to be available).

Discussion
In the Discussion, the authors appear to implicate that palliative care was generally provided. Details on this should be provided in Results.
The Discussion does not address most major limitations including the 19% of the pregnancies without outcome data, the post-natal interventions rate of use, and the rate of provision of palliative care.
The Discussion is lengthy and yet does not address critical aspects of the limitations so the study can be pit in a better perspective and better interpret its contribution to the literature.

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility? No

Are the conclusions drawn adequately supported by the results? No
No competing interests were disclosed. Competing Interests:

I have read this submission. I believe that I have an appropriate level of expertise to confirm that I have read this submission. I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.
Author

1.The Methods section of the abstract does not distinguish well the classification used. It seems there were two groups but this is not clear from the structure of the sentence. How was non-viability versus viability assessed?
Amendments to abstract significant to correct this.

2.It is stated that non-viability was determined at a median GA of 16 weeks. It is unclear in what cases would non-viability of births at 22-28 weeks can be determined at 16 weeks.
The sentence "Miscarriage (late expulsion) between 22 to <28 weeks was observed with non-viability occurring at an estimated median gestation of 16 weeks." …says miscarriage -not birth: and has now been amended "Individual patient analysis of all miscarriages, estimated non-viability to have occurred at a median of 16 weeks despite late expulsion between 22 to <28 weeks." Details are provided in the methods but the space limitation of the abstract prevents further elucidation of the reviewers query scenario e.g. A woman may present for the first time at 6 months (24 weeks GA) by LMP, she is sure of her dates but the ultrasound shows absent FHB in a fetus with a FL of 16 weeks (the HC may or may not be possible to obtain depending on why the fetus is so small) e.g FDIU since 16 weeks or IUGR only growing to a 16 week size. The expulsion occurs from 22 to <28 weeks pregnant based on LMP.

Methods 3.What was the quality control for data collection?
The records have been quality controlled for data entry since 1986 when the antenatal care program commenced, firstly by author FN and later by RM. The data has been quality controlled for completeness and correctness of entry as this has been the most intensely studied population (dataset) in terms of malaria in pregnancy worldwide. Gestational age has been a significant issue for decades as antimalarial drug exposures and pregnancy outcomes have been so important in this population where drug resistance P.falciparum is amongst the highest in the world. In addition significant efforts with gestational age were important to understand the association of artemisinin derivatives (the backbone of antimalarial treatment worldwide today) with fetal resorption. 4.Were clinicians trained in data collection? treatment worldwide today) with fetal resorption. 4.Were clinicians trained in data collection? Yes 5.Were data collectors prospectively assessing the quality of the data? Yes 6.What QI efforts were made to improve data collection? Amendment at methods: section of data extraction 7.Were the definitions of stillbirth versus neonatal death prospectively applied throughout the whole period of data collection? Yes. See supplementary file 1 for clarification. The introduction has also been modified to conform with the reviewers suggestion.

8.What signs of live birth were assessed?
In clinic based birth heart rate was assessed. In clinic and home births respiratory effort, and movement were assessed (or asked to the birth attendant). 9.How was HR assessed? Auscultation 10.What resuscitation measures used in HR was not audible? < 28 weeks none. This is not without reason in our experience: while the baby may live it does not live for very long as no accessory support can be provided. ≥ 28 weeks: 5 inflation breaths and then action dependent on HR according to the neonatal drill which is published elsewhere (Janet S et al 2018, e0190419 which is ref 45 in . PLoS One 13: v2..0). Clarified in the methods, in section "Setting", last paragraph. 11.Were data systematically collected on bag and mask ventilation, incubator use, IV fluids, oxygen supplementation, etc? Yes, these data were systematically collected. We have clarified the text as follows: "If extreme PTB occurs (<28 weeks) infants are provided with palliative care. Parents are involved and counseled in the process . No resuscitative efforts were offered to live born infants of 22 to <28 weeks' gestation as no further sophisticated care i.e. intubation, incubators, surfactant, parenteral nutrition etc. was available, unlike infants of ≥28 weeks' gestation."[ref 45 in v2.0] The two references Turner and Janet (both open access) are important as they describe everything that does happen (e.g.bag and mask ventilation etc) when an infant is of 28 weeks' gestation or more, in this low resource environment. Results 12.The outcome of 19% of booked pregnancies was not known. This is an important problem as pregnancies losses may be over-represented in these unknown outcome pregnancies.
We agree with the reviewer and we have added Amendment to the first sentence of the results (and abstract). This limitation of the data as the first paragraph to the discussion. However we do not think that this limitation detracts from our main conclusion i.e. the need to focus on viable fetuses that have reached a GA of 28 weeks. 13.The stillbirth rate is very high especially at the lower gestations. For example, over 40% of births at up 26 weeks were assessed to be stillbirths. There is a frequent of limited assessment for live birth at the lowest gestations. It is common to assess that a newborn is stillbirth because resuscitation is not implemented when few signs of live birth are present.
Thank you and this has been added as a limitation.
the reviewers comment (Before the section on palliative care…) the reviewers comment (Before the section on palliative care…) to further explain the high rate of stillbirth as many of these births would have been by emergency C.section in a HIC setting, not vaginal birth.(After the section on palliative care…)

Signs of live birth are not reported.
Amended, see methods, data extraction and data definitions; 2 bullet point. .

Resuscitation efforts are not reported
See point 11. 14.It is concerning that birth weight was not available for 42% of the births? Other basic data also were frequently not available. This raises concerns about the quality of the data.
We agree this is a limitation. As the 28 week cut-point has been used to define birth and miscarriage it has been an ongoing process to motivate staff to weigh infants born prior to this gestation. Parents also see no reason to bring a stillborn or perinatal death infant to the clinic for a weight.

15.Gestational age may be underestimated as one SD is above 1000 g which is unlikely for a population of 22-27 week pregnancies.
There are no SD above 1000g, only the upper range. These are possible from the local centiles produced for this population from ultrasound dated pregnancies (Rijken MJ et al. (2014) Quantifying Low Birth Weight, Preterm Birth and Small-for-Gestational-Age Effects of Malaria in Pregnancy: A Population Cohort Study. PLoS ONE 9(7): e100247. doi:10.1371/journal.pone.0100247). They may represent incorrect EGA but birthweight give lower stillbirth rates than gestational age cut-offs and this has been added to the discussion.
14.Data on provision of care are not provided well. It is unclear to what extent live births were provided full support (other than the therapies stated not to be available).
Amended. This is the same as point 9. There was no resuscitation, intubation, incubators, surfactant, parenteral nutrition etc. Thank you for the opportunity of reviewing this manuscript entitled "Miscarriage, stillbirth and neonatal mortality in the extreme preterm birth window of gestation in a limited-resource setting on the Thailand-Myanmar border: a population cohort study." The work done is impressive and has the potential to provide insight in how we: (a) view and manage the 'grey zone' in low resource setting, (b) support mothers and their families who present for care when in this 'grey zone', and (c) should allocate the limited resources in reducing neonatal mortality in low resource setting.

Title
The title is very long and needs to be shortened. Moreover, it needs to better capture the findings of the study.

Introduction
The definition of stillbirth, miscarriage, preterm birth, and neonatal death has been identified as problematic given overlaps in gestational cut-points. Hence, it is imperative that these terms be defined at the outset so that there is conceptual clarity both with respect to variables being examined and the intended purpose of the study. This would also help with interpretation of findings.
There are discrepancies with respect to the definition of miscarriage in the introduction and methodology (e.g., bottom of page 4). Moreover, issues are highlighted about accuracy of gestational assessment, lack of ultrasound diagnosis of congenital abnormality which further impacts definitions of terms particularly with inclusion of data before 2001. It is therefore not clear what the study hopes to accomplish by examining pregnancy outcomes between 24 to < 28 weeks' gestational age over a 20 year period.
The second paragraph: the information pertaining to preterm births, neonatal deaths, and rationale for the high rates seems misplaced. Explain the relevance of the information shared to the aims of the study. Furthermore, explain the consequence(s) of poor reporting of stillbirth and neonatal deaths.
The third paragraph: relevance of statistics related to mortality secondary to malaria is not clear.
The third paragraph: relevance of statistics related to mortality secondary to malaria is not clear.
The introduction should more clearly articulate the conceptual approach (i.e., operationalize definitions), explain the rationale of the study, and situate information related to preterm birth, neonatal deaths, maternal mortality within the context of study aims.

Methodology Data extraction and data definition
Data were reviewed from 1995 to 2015 which is problematic as care patterns have changed significantly in this time frame (details shared in second and third paragraph). For example, the authors explain that there has been a shift from home birth to hospital births and improvement in care during the antenatal period. These changes are likely to reduce fetal mortality, as well as maternal and neonatal mortality. Thus, it is not clear why this time span was used for the purpose of this study. Moreover, ultrasound was introduced in 2001 which means that accurate assessment of gestational age may be an issue prior to this time. Ultrasound has also changed the way in which loss of fetal viability was confirmed (e.g., fundal height measurements). Given the intent to describe viability and mortality in the window of 24 to <28 weeks' gestational age it is important to have consistency in the way definitions are operationalized throughout the course of the study.
Data Quality -birth records are computer based; however is initial data recorded on paper and then transferred to the computer system? Please comment on the accuracy of data and strategies employed in ensuring quality data. Based on the results it appears that there are potential issues with data quality as in 4.4% of the 204 pregnancies between 24 and < 28 weeks' gestational age there was missing data; birth weight data was available for 57.6% of neonates. This is a limitation and your discussion should explain how this impacts interpretation of findings.
Ultrasound assessment -was this done by one person? If multiple assessors please comment on inter-rater reliability of ultrasound dating.
Page 4 -please indicate the actual average of transporting women by car to the Thai hospital. Is this a potential limitation of the study? Statistical analysis "Univariable and multivariable logistic regression was used to assess the association between estimation of gestational age by ultrasound, homebirth and year of birth, and birth outcome (birth rather than miscarriage)." A better description needs to be provided about the analytic plan and how it relates to the objectives of the study. The analytic plan should identify how variable were identified to be important to include in the multivariable logistic regression analysis. How were variables entered in the multivariable logistic regression analysis? Explain the rationale for the decision(s). It appears that the outcome of interest is birth however only 57.8% of these were live birth. What outcome was used? Furthermore, only one infant survived. What is the minimal sample size required for your multivariable regression analysis?

Results
Pregnancy outcomes were unknown for 19.1% of the registered women -what was the issue? Only 50.6% of the women had their first antenatal care visit in trimester one which has implications for accuracy of gestational dating using ultrasound (most accurate in the first trimester).
The inclusion of termination of pregnancy is confusing particularly since two were for maternal conditions at 24 and < 28 weeks' gestation which in high income countries would be managed by inducing labour.

Discussion
The results do not demonstrate the bias detailed in the opening paragraph. Is there an opportunity here to compare historical cohorts (before introduction of ultrasound and after introduction of ultrasound)?
No competing interests were disclosed. The study titled 'Miscarriage, stillbirth and neonatal mortality in the extreme preterm birth window of gestation in a limited-resource setting on the Thailand-Myanmar border: a population cohort study' is a well thought out and interesting study and has a lot of potential to be better and more useful to the reader.
The following are some points I am hoping will be useful in the authors 'efforts to improve the manuscript.

Recommendation 1:
The main objectives of the study are not clear. If it is challenging the definition (or lack thereof) of the miscarriages and its overlap with extreme premature deliveries, please say so more clearly. Obviously, this overlap is very confusing and just to demonstrate the conceptual difficulties one faces when trying to analyze a data set like this, is a worthy cause.
Spend more time with these conceptual inconsistencies so the readers who are not familiar with these concepts will have a better understanding of your objectives and results.

Recommendation 2:
Third paragraph in the Introduction: The background information given is interesting but very short. In addition, it is not clear what purpose it serves in the general context of the manuscript. Recommendation: It would be ideal to provide more detailed information that is pertinent to the etiology of maternal diseases, miscarriages, stillbirth etc. in that region to help the reader to better understand the characteristics of the patient population. that region to help the reader to better understand the characteristics of the patient population.

Recommendation 3:
It appears there is a discrepancy between the definition of miscarriages at the end of the first paragraph of page 4 and the definition given under the header of miscarriage in the second column.
The first definition: When a fetus dies <24 weeks but is delivered between 24-28 weeks.
The second definition is used in the first sentence just under the subtitle of 'miscarriage': 114 women out of 318 women miscarried from 24<28 weeks' gestation and the distribution is given according to the gestational age. In this section it appears that for a fetus to be called he/she does not have to die<24 weeks and be delivered after 28 weeks. In addition, in our experience it not very common for the mother to keep the dead fetus for two or more weeks (in this case up to 4 weeks).

Recommendation 4:
Second column on page 6. The correct term for 'surface' examination is 'external' examination.

Recommendation 5:
I would recommend exclusion of terminations. That adds more confusion to the calculations and it is considered a very different group.

Question 1:
The true consequence of using different GA for the definition of fetal death/stillborn vs miscarriage (except causing confusion) is not clear. population cohort study' is a well thought out and interesting study and has a lot of potential to be better and more useful to the reader. Agree with potential to be better. The following are some points I am hoping will be useful in the authors 'efforts to improve the manuscript.

Recommendation 1:
The main objectives of the study are not clear. If it is challenging the definition (or lack thereof) of the miscarriages and its overlap with extreme premature deliveries, please say so more clearly. Obviously, this overlap is very confusing and just to demonstrate the conceptual difficulties one faces when trying to analyze a data set like this, is a worthy cause. Thank you for your patience with the manuscript -we are trying to articulate i.e. that this window (22 to <28 weeks) can include miscarriage, a large proportion.
Spend more time with these conceptual inconsistencies so the readers who are not familiar with these concepts will have a better understanding of your objectives and results. We agree that we did not succeed to give clarity on the main objective. We have rephrased the entire manuscript. The aim of this study is to contribute to the sparse body of evidence from limited resource settings that support the use of the 28 weeks' gestational age cut-point for stillbirth. We also argue for the usefulness of <28 weeks' gestational age to define miscarriage.

Recommendation 2:
Third paragraph in the Introduction: The background information given is interesting but very short. In addition, it is not clear what purpose it serves in the general context of the manuscript. Recommendation: It would be ideal to provide more detailed information that is pertinent to the etiology of maternal diseases, miscarriages, stillbirth etc. in that region to help the reader to better understand the characteristics of the patient population. Thank you for this comment -the introduction has been rewritten to address this point and Recommendation 1.

Recommendation 3:
It appears there is a discrepancy between the definition of miscarriages at the end of the first paragraph of page 4 and the definition given under the header of miscarriage in the second column. The first definition: When a fetus dies <24 weeks but is delivered between 24-28 weeks. The second definition is used in the first sentence just under the subtitle of 'miscarriage': 114 women out of 318 women miscarried from 24<28 weeks' gestation and the distribution is given according to the gestational age. In this section it appears that for a fetus to be called he/she does not have to die<24 weeks and be delivered after 28 weeks. The definitions have been laid out more clearly. For this analysis the definitions of miscarriage is absence of viability before 22 weeks but delivery in the 22 to <28 week window and extreme preterm birth is presence of viability and delivery at 22 to <28 weeks.
In addition, in our experience it not very common for the mother to keep the dead fetus for two or more weeks (in this case up to 4 weeks). This is a very helpful comment and it is exactly what we want to highlight for an in issue relevant to low resource settings. We have an average of 6 weeks before expulsion. low resource settings. We have an average of 6 weeks before expulsion. We suspect this must be a source of confusion in other low resource settings as these types of pregnancy outcomes are hard to document without a reliable method of measuring pregnancy viability e.g. ultrasound or multiple. We are also trying to highlight that the experience in low resource settings may not match the bulk of published evidence available.
Recommendation 4: Second column on page 6. The correct term for 'surface' examination is 'external' examination.

Amended as suggested
Recommendation 5: I would recommend exclusion of terminations. That adds more confusion to the calculations and it is considered a very different group. Amended as suggested Question 1: The true consequence of using different GA for the definition of fetal death/stillborn vs miscarriage (except causing confusion) is not clear. We have spent more time to explain this in the introduction. The practice of obstetrics in resource limited settings is not as efficient. Miscarriage can occur at 22-28 weeks (the window of extreme preterm birth) and in the setting described in this manuscript where significant effort has been placed on assessment of gestational age, we present data to prove that point. The WHO definition of 28 week cut-point that defines a stillbirth is attractive and pragmatic for low resource settings.

Question 2:
Since you have the numbers, what happens if you analyze the data where a pregnancy loss up to 20 weeks gestational age will be classified as a miscarriage and any delivery ≥20 weeks is classified either as a liveborn or stillborn? We have gone back to the records to classify outcomes from 22 to <28 weeks gestation as miscarriage, livebirth, stillbirth, neonatal mortality. The records of 20 and 21 weeks start to have too much missing data to contribute reliably to the argument.

I have read this submission. I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant
No competing interests were disclosed. reservations, as outlined above. Competing Interests: Close No competing interests were disclosed. Competing Interests: